fix

add ipnet crates
fix
2026-01-17 09:40:32 +00:00 · 2025-12-27 01:22:03 +08:00 · 2025-12-27 01:08:52 +08:00 · 2025-12-25 11:35:40 +08:00 · 2025-12-25 01:11:05 +08:00 · 2025-12-25 00:19:03 +08:00
104 changed files with 4180 additions and 3573 deletions
--- a/.github/workflows/audit.yml
+++ b/.github/workflows/audit.yml
@@ -40,7 +40,7 @@ env:
 jobs:
  security-audit:
    name: Security Audit
-    runs-on: ubicloud-standard-4
+    runs-on: ubicloud-standard-2
    timeout-minutes: 15
    steps:
      - name: Checkout repository
@@ -65,7 +65,7 @@ jobs:

  dependency-review:
    name: Dependency Review
-    runs-on: ubicloud-standard-4
+    runs-on: ubicloud-standard-2
    if: github.event_name == 'pull_request'
    permissions:
      contents: read
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -83,7 +83,7 @@ jobs:
  # Build strategy check - determine build type based on trigger
  build-check:
    name: Build Strategy Check
-    runs-on: ubicloud-standard-4
+    runs-on: ubicloud-standard-2
    outputs:
      should_build: ${{ steps.check.outputs.should_build }}
      build_type: ${{ steps.check.outputs.build_type }}
@@ -167,19 +167,19 @@ jobs:
      matrix:
        include:
          # Linux builds
-          - os: ubicloud-standard-4
+          - os: ubicloud-standard-2
            target: x86_64-unknown-linux-musl
            cross: false
            platform: linux
-          - os: ubicloud-standard-4
+          - os: ubicloud-standard-2
            target: aarch64-unknown-linux-musl
            cross: true
            platform: linux
-          - os: ubicloud-standard-4
+          - os: ubicloud-standard-2
            target: x86_64-unknown-linux-gnu
            cross: false
            platform: linux
-          - os: ubicloud-standard-4
+          - os: ubicloud-standard-2
            target: aarch64-unknown-linux-gnu
            cross: true
            platform: linux
@@ -532,7 +532,7 @@ jobs:
    name: Build Summary
    needs: [ build-check, build-rustfs ]
    if: always() && needs.build-check.outputs.should_build == 'true'
-    runs-on: ubicloud-standard-4
+    runs-on: ubicloud-standard-2
    steps:
      - name: Build completion summary
        shell: bash
@@ -584,7 +584,7 @@ jobs:
    name: Create GitHub Release
    needs: [ build-check, build-rustfs ]
    if: startsWith(github.ref, 'refs/tags/') && needs.build-check.outputs.build_type != 'development'
-    runs-on: ubicloud-standard-4
+    runs-on: ubicloud-standard-2
    permissions:
      contents: write
    outputs:
@@ -670,7 +670,7 @@ jobs:
    name: Upload Release Assets
    needs: [ build-check, build-rustfs, create-release ]
    if: startsWith(github.ref, 'refs/tags/') && needs.build-check.outputs.build_type != 'development'
-    runs-on: ubicloud-standard-4
+    runs-on: ubicloud-standard-2
    permissions:
      contents: write
      actions: read
@@ -751,7 +751,7 @@ jobs:
    name: Update Latest Version
    needs: [ build-check, upload-release-assets ]
    if: startsWith(github.ref, 'refs/tags/')
-    runs-on: ubicloud-standard-4
+    runs-on: ubicloud-standard-2
    steps:
      - name: Update latest.json
        env:
@@ -801,7 +801,7 @@ jobs:
    name: Publish Release
    needs: [ build-check, create-release, upload-release-assets ]
    if: startsWith(github.ref, 'refs/tags/') && needs.build-check.outputs.build_type != 'development'
-    runs-on: ubicloud-standard-4
+    runs-on: ubicloud-standard-2
    permissions:
      contents: write
    steps:
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -69,7 +69,7 @@ concurrency:
 env:
  CARGO_TERM_COLOR: always
  RUST_BACKTRACE: 1
-  CARGO_BUILD_JOBS: 8
+  CARGO_BUILD_JOBS: 2

 jobs:

@@ -78,7 +78,7 @@ jobs:
    permissions:
      actions: write
      contents: read
-    runs-on: ubicloud-standard-4
+    runs-on: ubicloud-standard-2
    outputs:
      should_skip: ${{ steps.skip_check.outputs.should_skip }}
    steps:
@@ -93,7 +93,7 @@ jobs:

  typos:
    name: Typos
-    runs-on: ubicloud-standard-4
+    runs-on: ubicloud-standard-2
    steps:
      - uses: actions/checkout@v6
      - uses: dtolnay/rust-toolchain@stable
@@ -136,7 +136,7 @@ jobs:
    name: End-to-End Tests
    needs: skip-check
    if: needs.skip-check.outputs.should_skip != 'true'
-    runs-on: ubicloud-standard-4
+    runs-on: ubicloud-standard-2
    timeout-minutes: 30
    steps:
      - name: Checkout repository
@@ -166,7 +166,7 @@ jobs:
        run: |
          touch rustfs/build.rs
          # Limit concurrency to prevent OOM
-          cargo build -p rustfs --bins --jobs 4
+          cargo build -p rustfs --bins --jobs 2

      - name: Run end-to-end tests
        run: |
--- a/.github/workflows/docker.yml
+++ b/.github/workflows/docker.yml
@@ -72,7 +72,7 @@ jobs:
  # Check if we should build Docker images
  build-check:
    name: Docker Build Check
-    runs-on: ubicloud-standard-4
+    runs-on: ubicloud-standard-2
    outputs:
      should_build: ${{ steps.check.outputs.should_build }}
      should_push: ${{ steps.check.outputs.should_push }}
@@ -264,7 +264,7 @@ jobs:
    name: Build Docker Images
    needs: build-check
    if: needs.build-check.outputs.should_build == 'true'
-    runs-on: ubicloud-standard-4
+    runs-on: ubicloud-standard-2
    timeout-minutes: 60
    steps:
      - name: Checkout repository
@@ -404,7 +404,7 @@ jobs:
    name: Docker Build Summary
    needs: [ build-check, build-docker ]
    if: always() && needs.build-check.outputs.should_build == 'true'
-    runs-on: ubicloud-standard-4
+    runs-on: ubicloud-standard-2
    steps:
      - name: Docker build completion summary
        run: |
--- a/.github/workflows/e2e-mint.yml
+++ b/.github/workflows/e2e-mint.yml
@@ -37,7 +37,7 @@ env:

 jobs:
  mint-single:
-    runs-on: ubicloud-standard-4
+    runs-on: ubicloud-standard-2
    timeout-minutes: 40
    steps:
      - name: Checkout
@@ -114,7 +114,7 @@ jobs:
  mint-multi:
    if: github.event_name == 'workflow_dispatch' && github.event.inputs.run-multi == 'true'
    needs: mint-single
-    runs-on: ubicloud-standard-4
+    runs-on: ubicloud-standard-2
    timeout-minutes: 60
    steps:
      - name: Checkout
--- a/.github/workflows/e2e-s3tests.yml
+++ b/.github/workflows/e2e-s3tests.yml
@@ -58,7 +58,7 @@ defaults:
 jobs:
  s3tests-single:
    if: github.event.inputs.test-mode == 'single'
-    runs-on: ubicloud-standard-4
+    runs-on: ubicloud-standard-2
    timeout-minutes: 120
    steps:
      - uses: actions/checkout@v6
@@ -212,7 +212,7 @@ jobs:

  s3tests-multi:
    if: github.event_name == 'workflow_dispatch' && github.event.inputs.test-mode == 'multi'
-    runs-on: ubicloud-standard-4
+    runs-on: ubicloud-standard-2
    timeout-minutes: 150
    steps:
      - uses: actions/checkout@v6
--- a/.github/workflows/helm-package.yml
+++ b/.github/workflows/helm-package.yml
@@ -27,7 +27,7 @@ env:

 jobs:
  build-helm-package:
-    runs-on: ubicloud-standard-4
+    runs-on: ubicloud-standard-2
    # Only run on successful builds triggered by tag pushes (version format: x.y.z or x.y.z-suffix)
    if: |
      github.event.workflow_run.conclusion == 'success' && 
@@ -63,7 +63,7 @@ jobs:
          retention-days: 1

  publish-helm-package:
-    runs-on: ubicloud-standard-4
+    runs-on: ubicloud-standard-2
    needs: [ build-helm-package ]

    steps:
--- a/.github/workflows/performance.yml
+++ b/.github/workflows/performance.yml
@@ -40,7 +40,7 @@ env:
 jobs:
  performance-profile:
    name: Performance Profiling
-    runs-on: ubicloud-standard-4
+    runs-on: ubicloud-standard-2
    timeout-minutes: 30
    steps:
      - name: Checkout repository
@@ -115,7 +115,7 @@ jobs:

  benchmark:
    name: Benchmark Tests
-    runs-on: ubicloud-standard-4
+    runs-on: ubicloud-standard-2
    timeout-minutes: 45
    steps:
      - name: Checkout repository
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -216,9 +216,12 @@ dependencies = [

 [[package]]
 name = "arc-swap"
-version = "1.7.1"
+version = "1.8.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "69f7f8c3906b62b754cd5326047894316021dcfe5a194c8ea52bdd94934a3457"
+checksum = "51d03449bb8ca2cc2ef70869af31463d1ae5ccc8fa3e334b307203fbf815207e"
+dependencies = [
+ "rustversion",
+]

 [[package]]
 name = "argon2"
@@ -515,9 +518,9 @@ dependencies = [

 [[package]]
 name = "async-lock"
-version = "3.4.1"
+version = "3.4.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "5fd03604047cee9b6ce9de9f70c6cd540a0520c813cbd49bae61f33ab80ed1dc"
+checksum = "290f7f2596bd5b78a9fec8088ccd89180d7f9f55b94b0576823bbbdc72ee8311"
 dependencies = [
 "event-listener",
 "event-listener-strategy",
@@ -691,9 +694,9 @@ dependencies = [

 [[package]]
 name = "aws-sdk-s3"
-version = "1.118.0"
+version = "1.119.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d3e6b7079f85d9ea9a70643c9f89f50db70f5ada868fa9cfe08c1ffdf51abc13"
+checksum = "1d65fddc3844f902dfe1864acb8494db5f9342015ee3ab7890270d36fbd2e01c"
 dependencies = [
 "aws-credential-types",
 "aws-runtime",
@@ -942,9 +945,9 @@ dependencies = [

 [[package]]
 name = "aws-smithy-runtime"
-version = "1.9.6"
+version = "1.9.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "65fda37911905ea4d3141a01364bc5509a0f32ae3f3b22d6e330c0abfb62d247"
+checksum = "a392db6c583ea4a912538afb86b7be7c5d8887d91604f50eb55c262ee1b4a5f5"
 dependencies = [
 "aws-smithy-async",
 "aws-smithy-http",
@@ -4843,9 +4846,9 @@ dependencies = [

 [[package]]
 name = "libz-rs-sys"
-version = "0.5.4"
+version = "0.5.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "15413ef615ad868d4d65dce091cb233b229419c7c0c4bcaa746c0901c49ff39c"
+checksum = "c10501e7805cee23da17c7790e59df2870c0d4043ec6d03f67d31e2b53e77415"
 dependencies = [
 "zlib-rs",
 ]
@@ -6725,9 +6728,9 @@ checksum = "ba39f3699c378cd8970968dcbff9c43159ea4cfbd88d43c00b22f2ef10a435d2"

 [[package]]
 name = "reqwest"
-version = "0.12.26"
+version = "0.12.28"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3b4c14b2d9afca6a60277086b0cc6a6ae0b568f6f7916c943a8cdc79f8be240f"
+checksum = "eddd3ca559203180a307f12d114c268abf583f59b03cb906fd0b3ff8646c1147"
 dependencies = [
 "base64",
 "bytes",
@@ -6851,22 +6854,19 @@ dependencies = [

 [[package]]
 name = "rmp"
-version = "0.8.14"
+version = "0.8.15"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "228ed7c16fa39782c3b3468e974aec2795e9089153cd08ee2e9aefb3613334c4"
+checksum = "4ba8be72d372b2c9b35542551678538b562e7cf86c3315773cae48dfbfe7790c"
 dependencies = [
- "byteorder",
 "num-traits",
- "paste",
 ]

 [[package]]
 name = "rmp-serde"
-version = "1.3.0"
+version = "1.3.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "52e599a477cf9840e92f2cde9a7189e67b42c57532749bf90aea6ec10facd4db"
+checksum = "72f81bee8c8ef9b577d1681a70ebbc962c232461e397b22c208c43c04b67a155"
 dependencies = [
- "byteorder",
 "rmp",
 "serde",
 ]
@@ -7040,6 +7040,7 @@ dependencies = [
 "hex-simd",
 "http 1.4.0",
 "http-body 1.0.1",
+ "http-body-util",
 "hyper 1.8.1",
 "hyper-util",
 "jemalloc_pprof",
@@ -7302,6 +7303,7 @@ dependencies = [
 "tonic",
 "tower",
 "tracing",
+ "tracing-subscriber",
 "url",
 "urlencoding",
 "uuid",
@@ -7437,6 +7439,7 @@ dependencies = [
 name = "rustfs-notify"
 version = "0.0.5"
 dependencies = [
+ "arc-swap",
 "async-trait",
 "axum",
 "chrono",
@@ -7655,6 +7658,7 @@ dependencies = [
 "hmac 0.13.0-rc.3",
 "http 1.4.0",
 "hyper 1.8.1",
+ "ipnet",
 "libc",
 "local-ip-address",
 "lz4",
@@ -7751,9 +7755,9 @@ dependencies = [

 [[package]]
 name = "rustix"
-version = "1.1.2"
+version = "1.1.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "cd15f8a2c5551a84d56efdc1cd049089e409ac19a3072d5037a17fd70719ff3e"
+checksum = "146c9e247ccc180c1f61615433868c99f3de3ae256a30a43b49f67c2d9171f34"
 dependencies = [
 "bitflags 2.10.0",
 "errno",
@@ -7868,8 +7872,8 @@ checksum = "62049b2877bf12821e8f9ad256ee38fdc31db7387ec2d3b3f403024de2034aea"

 [[package]]
 name = "s3s"
-version = "0.12.0-rc.6"
-source = "git+https://github.com/s3s-project/s3s.git?branch=main#df2434d7ad2f0b774e68f25cae90c053dcb84f24"
+version = "0.13.0-alpha"
+source = "git+https://github.com/s3s-project/s3s.git?branch=main#f6198bbf49abe60066fe47cbbefcb7078863b3e9"
 dependencies = [
 "arrayvec",
 "async-trait",
@@ -8121,15 +8125,15 @@ dependencies = [

 [[package]]
 name = "serde_json"
-version = "1.0.145"
+version = "1.0.147"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "402a6f66d8c709116cf22f558eab210f5a50187f702eb4d7e5ef38d9a7f1c79c"
+checksum = "6af14725505314343e673e9ecb7cd7e8a36aa9791eb936235a3567cc31447ae4"
 dependencies = [
 "itoa",
 "memchr",
- "ryu",
 "serde",
 "serde_core",
+ "zmij",
 ]

 [[package]]
@@ -8849,14 +8853,14 @@ dependencies = [

 [[package]]
 name = "tempfile"
-version = "3.23.0"
+version = "3.24.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2d31c77bdf42a745371d260a26ca7163f1e0924b64afa0b688e61b5a9fa02f16"
+checksum = "655da9c7eb6305c55742045d5a8d2037996d61d8de95806335c7c86ce0f82e9c"
 dependencies = [
 "fastrand",
 "getrandom 0.3.4",
 "once_cell",
- "rustix 1.1.2",
+ "rustix 1.1.3",
 "windows-sys 0.61.2",
 ]

@@ -10246,7 +10250,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "32e45ad4206f6d2479085147f02bc2ef834ac85886624a23575ae137c8aa8156"
 dependencies = [
 "libc",
- "rustix 1.1.2",
+ "rustix 1.1.3",
 ]

 [[package]]
@@ -10423,9 +10427,15 @@ dependencies = [

 [[package]]
 name = "zlib-rs"
-version = "0.5.4"
+version = "0.5.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "51f936044d677be1a1168fae1d03b583a285a5dd9d8cbf7b24c23aa1fc775235"
+checksum = "40990edd51aae2c2b6907af74ffb635029d5788228222c4bb811e9351c0caad3"
+
+[[package]]
+name = "zmij"
+version = "0.1.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9e404bcd8afdaf006e529269d3e85a743f9480c3cef60034d77860d02964f3ba"

 [[package]]
 name = "zopfli"
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -109,7 +109,8 @@ hyper-rustls = { version = "0.27.7", default-features = false, features = ["nati
 hyper-util = { version = "0.1.19", features = ["tokio", "server-auto", "server-graceful"] }
 http = "1.4.0"
 http-body = "1.0.1"
-reqwest = { version = "0.12.26", default-features = false, features = ["rustls-tls-webpki-roots", "charset", "http2", "system-proxy", "stream", "json", "blocking"] }
+http-body-util = "0.1.3"
+reqwest = { version = "0.12.28", default-features = false, features = ["rustls-tls-webpki-roots", "charset", "http2", "system-proxy", "stream", "json", "blocking"] }
 socket2 = "0.6.1"
 tokio = { version = "1.48.0", features = ["fs", "rt-multi-thread"] }
 tokio-rustls = { version = "0.26.4", default-features = false, features = ["logging", "tls12", "ring"] }
@@ -121,6 +122,7 @@ tonic-prost = { version = "0.14.2" }
 tonic-prost-build = { version = "0.14.2" }
 tower = { version = "0.5.2", features = ["timeout"] }
 tower-http = { version = "0.6.8", features = ["cors"] }
+ipnet = "2.11.0"

 # Serialization and Data Formats
 bytes = { version = "1.11.0", features = ["serde"] }
@@ -131,10 +133,10 @@ form_urlencoded = "1.2.2"
 prost = "0.14.1"
 quick-xml = "0.38.4"
 rmcp = { version = "0.12.0" }
-rmp = { version = "0.8.14" }
-rmp-serde = { version = "1.3.0" }
+rmp = { version = "0.8.15" }
+rmp-serde = { version = "1.3.1" }
 serde = { version = "1.0.228", features = ["derive"] }
-serde_json = { version = "1.0.145", features = ["raw_value"] }
+serde_json = { version = "1.0.147", features = ["raw_value"] }
 serde_urlencoded = "0.7.1"
 schemars = "1.1.0"

@@ -163,13 +165,13 @@ time = { version = "0.3.44", features = ["std", "parsing", "formatting", "macros

 # Utilities and Tools
 anyhow = "1.0.100"
-arc-swap = "1.7.1"
+arc-swap = "1.8.0"
 astral-tokio-tar = "0.5.6"
 atoi = "2.0.0"
 atomic_enum = "0.3.0"
 aws-config = { version = "1.8.12" }
 aws-credential-types = { version = "1.2.11" }
-aws-sdk-s3 = { version = "1.117.0", default-features = false, features = ["sigv4a", "rustls", "rt-tokio"] }
+aws-sdk-s3 = { version = "1.119.0", default-features = false, features = ["sigv4a", "rustls", "rt-tokio"] }
 aws-smithy-types = { version = "1.3.5" }
 base64 = "0.22.1"
 base64-simd = "0.8.0"
@@ -222,7 +224,7 @@ regex = { version = "1.12.2" }
 rumqttc = { version = "0.25.1" }
 rust-embed = { version = "8.9.0" }
 rustc-hash = { version = "2.1.1" }
-s3s = { version = "0.12.0-rc.6", features = ["minio"], git = "https://github.com/s3s-project/s3s.git", branch = "main" }
+s3s = { version = "0.13.0-alpha", features = ["minio"], git = "https://github.com/s3s-project/s3s.git", branch = "main" }
 serial_test = "3.2.0"
 shadow-rs = { version = "1.5.0", default-features = false }
 siphasher = "1.0.1"
@@ -235,7 +237,7 @@ strum = { version = "0.27.2", features = ["derive"] }
 sysctl = "0.7.1"
 sysinfo = "0.37.2"
 temp-env = "0.3.6"
-tempfile = "3.23.0"
+tempfile = "3.24.0"
 test-case = "3.3.1"
 thiserror = "2.0.17"
 tracing = { version = "0.1.44" }
--- a/45
+++ b/45
@@ -9,30 +9,53 @@ CONTAINER_NAME ?= rustfs-dev
 DOCKERFILE_PRODUCTION = Dockerfile
 DOCKERFILE_SOURCE = Dockerfile.source

+# Fatal check
+# Checks all required dependencies and exits with error if not found
+# (e.g., cargo, rustfmt)
+check-%:
+	@command -v $* >/dev/null 2>&1 || { \
+		echo >&2 "❌ '$*' is not installed."; \
+		exit 1; \
+	}
+
+# Warning-only check
+# Checks for optional dependencies and issues a warning if not found
+# (e.g., cargo-nextest for enhanced testing)
+warn-%:
+	@command -v $* >/dev/null 2>&1 || { \
+		echo >&2 "⚠️ '$*' is not installed."; \
+	}
+
+# For checking dependencies use check-<dep-name> or warn-<dep-name>
+.PHONY: core-deps fmt-deps test-deps
+core-deps: check-cargo
+fmt-deps: check-rustfmt
+test-deps: warn-cargo-nextest
+
 # Code quality and formatting targets
 .PHONY: fmt
-fmt:
+fmt: core-deps fmt-deps
 	@echo "🔧 Formatting code..."
 	cargo fmt --all

 .PHONY: fmt-check
-fmt-check:
+fmt-check: core-deps fmt-deps
 	@echo "📝 Checking code formatting..."
 	cargo fmt --all --check

 .PHONY: clippy
-clippy:
+clippy: core-deps
 	@echo "🔍 Running clippy checks..."
 	cargo clippy --fix --allow-dirty
 	cargo clippy --all-targets --all-features -- -D warnings

 .PHONY: check
-check:
+check: core-deps
 	@echo "🔨 Running compilation check..."
 	cargo check --all-targets

 .PHONY: test
-test:
+test: core-deps test-deps
 	@echo "🧪 Running tests..."
 	@if command -v cargo-nextest >/dev/null 2>&1; then \
 		cargo nextest run --all --exclude e2e_test; \
@@ -42,16 +65,16 @@ test:
 	fi
 	cargo test --all --doc

-.PHONY: pre-commit
-pre-commit: fmt clippy check test
-	@echo "✅ All pre-commit checks passed!"
-
 .PHONY: setup-hooks
 setup-hooks:
 	@echo "🔧 Setting up git hooks..."
 	chmod +x .git/hooks/pre-commit
 	@echo "✅ Git hooks setup complete!"

+.PHONY: pre-commit
+pre-commit: fmt clippy check test
+	@echo "✅ All pre-commit checks passed!"
+
 .PHONY: e2e-server
 e2e-server:
 	sh $(shell pwd)/scripts/run.sh
@@ -186,8 +209,6 @@ docker-dev-push:
 		--push \
 		.

-
-
 # Local production builds using direct buildx (alternative to docker-buildx.sh)
 .PHONY: docker-buildx-production-local
 docker-buildx-production-local:
@@ -247,8 +268,6 @@ dev-env-stop:
 .PHONY: dev-env-restart
 dev-env-restart: dev-env-stop dev-env-start

-
-
 # ========================================================================================
 # Build Utilities
 # ========================================================================================
--- a/README.md
+++ b/README.md
@@ -174,7 +174,7 @@ nix run

 ### Accessing RustFS

-5.  **Access the Console**: Open your web browser and navigate to `http://localhost:9000` to access the RustFS console.
+5.  **Access the Console**: Open your web browser and navigate to `http://localhost:9001` to access the RustFS console.
      * Default credentials: `rustfsadmin` / `rustfsadmin`
 6.  **Create a Bucket**: Use the console to create a new bucket for your objects.
 7.  **Upload Objects**: You can upload files directly through the console or use S3-compatible APIs/clients to interact with your RustFS instance.
--- a/crates/ahm/src/heal/manager.rs
+++ b/crates/ahm/src/heal/manager.rs
@@ -468,14 +468,17 @@ impl HealManager {
        let active_heals = self.active_heals.clone();
        let cancel_token = self.cancel_token.clone();
        let storage = self.storage.clone();
-
-        info!(
-            "start_auto_disk_scanner: Starting auto disk scanner with interval: {:?}",
-            config.read().await.heal_interval
-        );
+        let mut duration = {
+            let config = config.read().await;
+            config.heal_interval
+        };
+        if duration < Duration::from_secs(1) {
+            duration = Duration::from_secs(1);
+        }
+        info!("start_auto_disk_scanner: Starting auto disk scanner with interval: {:?}", duration);

        tokio::spawn(async move {
-            let mut interval = interval(config.read().await.heal_interval);
+            let mut interval = interval(duration);

            loop {
                tokio::select! {
--- a/crates/ahm/src/scanner/data_scanner.rs
+++ b/crates/ahm/src/scanner/data_scanner.rs
@@ -30,7 +30,7 @@ use rustfs_ecstore::{
    bucket::versioning::VersioningApi,
    bucket::versioning_sys::BucketVersioningSys,
    data_usage::{aggregate_local_snapshots, compute_bucket_usage, store_data_usage_in_backend},
-    disk::{Disk, DiskAPI, DiskStore, RUSTFS_META_BUCKET, WalkDirOptions},
+    disk::{DiskAPI, DiskStore, RUSTFS_META_BUCKET, WalkDirOptions},
    set_disk::SetDisks,
    store_api::ObjectInfo,
 };
@@ -1977,7 +1977,7 @@ impl Scanner {
                    } else {
                        // Apply lifecycle actions
                        if let Some(lifecycle_config) = &lifecycle_config {
-                            if let Disk::Local(_local_disk) = &**disk {
+                            if disk.is_local() {
                                let vcfg = BucketVersioningSys::get(bucket).await.ok();

                                let mut scanner_item = ScannerItem {
--- a/crates/ahm/tests/data_usage_fallback_test.rs
+++ b/crates/ahm/tests/data_usage_fallback_test.rs
@@ -21,10 +21,11 @@ use rustfs_ecstore::bucket::metadata_sys::{BucketMetadataSys, GLOBAL_BucketMetad
 use rustfs_ecstore::endpoints::EndpointServerPools;
 use rustfs_ecstore::store::ECStore;
 use rustfs_ecstore::store_api::{ObjectIO, PutObjReader, StorageAPI};
-use std::sync::Arc;
+use std::sync::{Arc, Once};
 use tempfile::TempDir;
 use tokio::sync::RwLock;
 use tokio_util::sync::CancellationToken;
+use tracing::Level;

 /// Build a minimal single-node ECStore over a temp directory and populate objects.
 async fn create_store_with_objects(count: usize) -> (TempDir, std::sync::Arc<ECStore>) {
@@ -74,8 +75,22 @@ async fn create_store_with_objects(count: usize) -> (TempDir, std::sync::Arc<ECS
    (temp_dir, store)
 }

+static INIT: Once = Once::new();
+
+fn init_tracing(filter_level: Level) {
+    INIT.call_once(|| {
+        let _ = tracing_subscriber::fmt()
+            .with_env_filter(tracing_subscriber::EnvFilter::from_default_env())
+            .with_max_level(filter_level)
+            .with_timer(tracing_subscriber::fmt::time::UtcTime::rfc_3339())
+            .with_thread_names(true)
+            .try_init();
+    });
+}
+
 #[tokio::test]
 async fn fallback_builds_full_counts_over_100_objects() {
+    init_tracing(Level::ERROR);
    let (_tmp, store) = create_store_with_objects(1000).await;
    let scanner = Scanner::new(None, None);

--- a/crates/ahm/tests/heal_integration_test.rs
+++ b/crates/ahm/tests/heal_integration_test.rs
@@ -38,9 +38,13 @@ use walkdir::WalkDir;
 static GLOBAL_ENV: OnceLock<(Vec<PathBuf>, Arc<ECStore>, Arc<ECStoreHealStorage>)> = OnceLock::new();
 static INIT: Once = Once::new();

-fn init_tracing() {
+pub fn init_tracing() {
    INIT.call_once(|| {
-        let _ = tracing_subscriber::fmt::try_init();
+        let _ = tracing_subscriber::fmt()
+            .with_env_filter(tracing_subscriber::EnvFilter::from_default_env())
+            .with_timer(tracing_subscriber::fmt::time::UtcTime::rfc_3339())
+            .with_thread_names(true)
+            .try_init();
    });
 }

@@ -356,7 +360,7 @@ mod serial_tests {

        // Create heal manager with faster interval
        let cfg = HealConfig {
-            heal_interval: Duration::from_secs(2),
+            heal_interval: Duration::from_secs(1),
            ..Default::default()
        };
        let heal_manager = HealManager::new(heal_storage.clone(), Some(cfg));
--- a/crates/audit/src/factory.rs
+++ b/crates/audit/src/factory.rs
@@ -60,8 +60,9 @@ impl TargetFactory for WebhookTargetFactory {
        let endpoint = config
            .lookup(WEBHOOK_ENDPOINT)
            .ok_or_else(|| TargetError::Configuration("Missing webhook endpoint".to_string()))?;
-        let endpoint_url = Url::parse(&endpoint)
-            .map_err(|e| TargetError::Configuration(format!("Invalid endpoint URL: {e} (value: '{endpoint}')")))?;
+        let parsed_endpoint = endpoint.trim();
+        let endpoint_url = Url::parse(parsed_endpoint)
+            .map_err(|e| TargetError::Configuration(format!("Invalid endpoint URL: {e} (value: '{parsed_endpoint}')")))?;

        let args = WebhookArgs {
            enable: true, // If we are here, it's already enabled.
--- a/crates/common/src/lib.rs
+++ b/crates/common/src/lib.rs
@@ -19,6 +19,10 @@ pub mod globals;
 pub mod heal_channel;
 pub mod last_minute;
 pub mod metrics;
+mod readiness;
+
+pub use globals::*;
+pub use readiness::{GlobalReadiness, SystemStage};

 // is ','
 pub static DEFAULT_DELIMITER: u8 = 44;
--- a/crates/common/src/readiness.rs
+++ b/crates/common/src/readiness.rs
@@ -0,0 +1,136 @@
+// Copyright 2024 RustFS Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use std::sync::atomic::{AtomicU8, Ordering};
+
+/// Represents the various stages of system startup
+#[repr(u8)]
+pub enum SystemStage {
+    Booting = 0,
+    StorageReady = 1, // Disks online, Quorum met
+    IamReady = 2,     // Users and Policies loaded into cache
+    FullReady = 3,    // System ready to serve all traffic
+}
+
+/// Global readiness tracker for the service
+/// This struct uses atomic operations to track the readiness status of various components
+/// of the service in a thread-safe manner.
+pub struct GlobalReadiness {
+    status: AtomicU8,
+}
+
+impl Default for GlobalReadiness {
+    fn default() -> Self {
+        Self::new()
+    }
+}
+
+impl GlobalReadiness {
+    /// Create a new GlobalReadiness instance with initial status as Starting
+    /// # Returns
+    /// A new instance of GlobalReadiness
+    pub fn new() -> Self {
+        Self {
+            status: AtomicU8::new(SystemStage::Booting as u8),
+        }
+    }
+
+    /// Update the system to a new stage
+    ///
+    /// # Arguments
+    /// * `step` - The SystemStage step to mark as ready
+    pub fn mark_stage(&self, step: SystemStage) {
+        self.status.fetch_max(step as u8, Ordering::SeqCst);
+    }
+
+    /// Check if the service is fully ready
+    /// # Returns
+    /// `true` if the service is fully ready, `false` otherwise
+    pub fn is_ready(&self) -> bool {
+        self.status.load(Ordering::SeqCst) == SystemStage::FullReady as u8
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use std::sync::Arc;
+    use std::thread;
+
+    #[test]
+    fn test_initial_state() {
+        let readiness = GlobalReadiness::new();
+        assert!(!readiness.is_ready());
+        assert_eq!(readiness.status.load(Ordering::SeqCst), SystemStage::Booting as u8);
+    }
+
+    #[test]
+    fn test_mark_stage_progression() {
+        let readiness = GlobalReadiness::new();
+        readiness.mark_stage(SystemStage::StorageReady);
+        assert!(!readiness.is_ready());
+        assert_eq!(readiness.status.load(Ordering::SeqCst), SystemStage::StorageReady as u8);
+
+        readiness.mark_stage(SystemStage::IamReady);
+        assert!(!readiness.is_ready());
+        assert_eq!(readiness.status.load(Ordering::SeqCst), SystemStage::IamReady as u8);
+
+        readiness.mark_stage(SystemStage::FullReady);
+        assert!(readiness.is_ready());
+    }
+
+    #[test]
+    fn test_no_regression() {
+        let readiness = GlobalReadiness::new();
+        readiness.mark_stage(SystemStage::FullReady);
+        readiness.mark_stage(SystemStage::IamReady); // Should not regress
+        assert!(readiness.is_ready());
+    }
+
+    #[test]
+    fn test_concurrent_marking() {
+        let readiness = Arc::new(GlobalReadiness::new());
+        let mut handles = vec![];
+
+        for _ in 0..10 {
+            let r = Arc::clone(&readiness);
+            handles.push(thread::spawn(move || {
+                r.mark_stage(SystemStage::StorageReady);
+                r.mark_stage(SystemStage::IamReady);
+                r.mark_stage(SystemStage::FullReady);
+            }));
+        }
+
+        for h in handles {
+            h.join().unwrap();
+        }
+
+        assert!(readiness.is_ready());
+    }
+
+    #[test]
+    fn test_is_ready_only_at_full_ready() {
+        let readiness = GlobalReadiness::new();
+        assert!(!readiness.is_ready());
+
+        readiness.mark_stage(SystemStage::StorageReady);
+        assert!(!readiness.is_ready());
+
+        readiness.mark_stage(SystemStage::IamReady);
+        assert!(!readiness.is_ready());
+
+        readiness.mark_stage(SystemStage::FullReady);
+        assert!(readiness.is_ready());
+    }
+}
--- a/crates/config/src/constants/runtime.rs
+++ b/crates/config/src/constants/runtime.rs
@@ -39,3 +39,10 @@ pub const DEFAULT_MAX_IO_EVENTS_PER_TICK: usize = 1024;
 /// Event polling default (Tokio default 61)
 pub const DEFAULT_EVENT_INTERVAL: u32 = 61;
 pub const DEFAULT_RNG_SEED: Option<u64> = None; // None means random
+
+/// Threshold for small object seek support in megabytes.
+///
+/// When an object is smaller than this size, rustfs will provide seek support.
+///
+/// Default is set to 10MB.
+pub const DEFAULT_OBJECT_SEEK_SUPPORT_THRESHOLD: usize = 10 * 1024 * 1024;
--- a/crates/config/src/notify/mod.rs
+++ b/crates/config/src/notify/mod.rs
@@ -51,6 +51,18 @@ pub const ENV_NOTIFY_TARGET_STREAM_CONCURRENCY: &str = "RUSTFS_NOTIFY_TARGET_STR
 /// Adjust this value based on your system's capabilities and expected load.
 pub const DEFAULT_NOTIFY_TARGET_STREAM_CONCURRENCY: usize = 20;

+/// Name of the environment variable that configures send concurrency.
+/// Controls how many send operations are processed in parallel by the notification system.
+/// Defaults to [`DEFAULT_NOTIFY_SEND_CONCURRENCY`] if not set.
+/// Example: `RUSTFS_NOTIFY_SEND_CONCURRENCY=64`.
+pub const ENV_NOTIFY_SEND_CONCURRENCY: &str = "RUSTFS_NOTIFY_SEND_CONCURRENCY";
+
+/// Default concurrency for send operations in the notification system
+/// This value is used if the environment variable `RUSTFS_NOTIFY_SEND_CONCURRENCY` is not set.
+/// It defines how many send operations can be processed in parallel by the notification system at any given time.
+/// Adjust this value based on your system's capabilities and expected load.
+pub const DEFAULT_NOTIFY_SEND_CONCURRENCY: usize = 64;
+
 #[allow(dead_code)]
 pub const NOTIFY_SUB_SYSTEMS: &[&str] = &[NOTIFY_MQTT_SUB_SYS, NOTIFY_WEBHOOK_SUB_SYS];

--- a/crates/e2e_test/src/reliant/head_deleted_object_versioning_test.rs
+++ b/crates/e2e_test/src/reliant/head_deleted_object_versioning_test.rs
@@ -0,0 +1,138 @@
+// Copyright 2024 RustFS Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+//! Test for HeadObject on deleted objects with versioning enabled
+//!
+//! This test reproduces the issue where getting a deleted object returns
+//! 200 OK instead of 404 NoSuchKey when versioning is enabled.
+
+#![cfg(test)]
+
+use aws_config::meta::region::RegionProviderChain;
+use aws_sdk_s3::Client;
+use aws_sdk_s3::config::{Credentials, Region};
+use aws_sdk_s3::error::SdkError;
+use aws_sdk_s3::types::{BucketVersioningStatus, VersioningConfiguration};
+use bytes::Bytes;
+use serial_test::serial;
+use std::error::Error;
+use tracing::info;
+
+const ENDPOINT: &str = "http://localhost:9000";
+const ACCESS_KEY: &str = "rustfsadmin";
+const SECRET_KEY: &str = "rustfsadmin";
+const BUCKET: &str = "test-head-deleted-versioning-bucket";
+
+async fn create_aws_s3_client() -> Result<Client, Box<dyn Error>> {
+    let region_provider = RegionProviderChain::default_provider().or_else(Region::new("us-east-1"));
+    let shared_config = aws_config::defaults(aws_config::BehaviorVersion::latest())
+        .region(region_provider)
+        .credentials_provider(Credentials::new(ACCESS_KEY, SECRET_KEY, None, None, "static"))
+        .endpoint_url(ENDPOINT)
+        .load()
+        .await;
+
+    let client = Client::from_conf(
+        aws_sdk_s3::Config::from(&shared_config)
+            .to_builder()
+            .force_path_style(true)
+            .build(),
+    );
+    Ok(client)
+}
+
+/// Setup test bucket, creating it if it doesn't exist, and enable versioning
+async fn setup_test_bucket(client: &Client) -> Result<(), Box<dyn Error>> {
+    match client.create_bucket().bucket(BUCKET).send().await {
+        Ok(_) => {}
+        Err(SdkError::ServiceError(e)) => {
+            let e = e.into_err();
+            let error_code = e.meta().code().unwrap_or("");
+            if !error_code.eq("BucketAlreadyExists") && !error_code.eq("BucketAlreadyOwnedByYou") {
+                return Err(e.into());
+            }
+        }
+        Err(e) => {
+            return Err(e.into());
+        }
+    }
+
+    // Enable versioning
+    client
+        .put_bucket_versioning()
+        .bucket(BUCKET)
+        .versioning_configuration(
+            VersioningConfiguration::builder()
+                .status(BucketVersioningStatus::Enabled)
+                .build(),
+        )
+        .send()
+        .await?;
+
+    Ok(())
+}
+
+/// Test that HeadObject on a deleted object returns NoSuchKey when versioning is enabled
+#[tokio::test]
+#[serial]
+#[ignore = "requires running RustFS server at localhost:9000"]
+async fn test_head_deleted_object_versioning_returns_nosuchkey() -> Result<(), Box<dyn std::error::Error>> {
+    let _ = tracing_subscriber::fmt()
+        .with_max_level(tracing::Level::INFO)
+        .with_test_writer()
+        .try_init();
+
+    info!("🧪 Starting test_head_deleted_object_versioning_returns_nosuchkey");
+
+    let client = create_aws_s3_client().await?;
+    setup_test_bucket(&client).await?;
+
+    let key = "test-head-deleted-versioning.txt";
+    let content = b"Test content for HeadObject with versioning";
+
+    // Upload and verify
+    client
+        .put_object()
+        .bucket(BUCKET)
+        .key(key)
+        .body(Bytes::from_static(content).into())
+        .send()
+        .await?;
+
+    // Delete the object (creates a delete marker)
+    client.delete_object().bucket(BUCKET).key(key).send().await?;
+
+    // Try to head the deleted object (latest version is delete marker)
+    let head_result = client.head_object().bucket(BUCKET).key(key).send().await;
+
+    assert!(head_result.is_err(), "HeadObject on deleted object should return an error");
+
+    match head_result.unwrap_err() {
+        SdkError::ServiceError(service_err) => {
+            let s3_err = service_err.into_err();
+            assert!(
+                s3_err.meta().code() == Some("NoSuchKey")
+                    || s3_err.meta().code() == Some("NotFound")
+                    || s3_err.meta().code() == Some("404"),
+                "Error should be NoSuchKey or NotFound, got: {s3_err:?}"
+            );
+            info!("✅ HeadObject correctly returns NoSuchKey/NotFound");
+        }
+        other_err => {
+            panic!("Expected ServiceError but got: {other_err:?}");
+        }
+    }
+
+    Ok(())
+}
--- a/crates/e2e_test/src/reliant/mod.rs
+++ b/crates/e2e_test/src/reliant/mod.rs
@@ -14,6 +14,7 @@

 mod conditional_writes;
 mod get_deleted_object_test;
+mod head_deleted_object_versioning_test;
 mod lifecycle;
 mod lock;
 mod node_interact_test;
--- a/crates/ecstore/Cargo.toml
+++ b/crates/ecstore/Cargo.toml
@@ -113,6 +113,7 @@ faster-hex = { workspace = true }
 tokio = { workspace = true, features = ["rt-multi-thread", "macros"] }
 criterion = { workspace = true, features = ["html_reports"] }
 temp-env = { workspace = true }
+tracing-subscriber = { workspace = true }

 [build-dependencies]
 shadow-rs = { workspace = true, features = ["build", "metadata"] }
--- a/crates/ecstore/src/admin_server_info.rs
+++ b/crates/ecstore/src/admin_server_info.rs
@@ -23,7 +23,7 @@ use crate::{
 };

 use crate::data_usage::load_data_usage_cache;
-use rustfs_common::{globals::GLOBAL_LOCAL_NODE_NAME, heal_channel::DriveState};
+use rustfs_common::{GLOBAL_LOCAL_NODE_NAME, heal_channel::DriveState};
 use rustfs_madmin::{
    BackendDisks, Disk, ErasureSetInfo, ITEM_INITIALIZING, ITEM_OFFLINE, ITEM_ONLINE, InfoMessage, ServerProperties,
 };
--- a/crates/ecstore/src/cache_value/metacache_set.rs
+++ b/crates/ecstore/src/cache_value/metacache_set.rs
@@ -16,7 +16,7 @@ use crate::disk::error::DiskError;
 use crate::disk::{self, DiskAPI, DiskStore, WalkDirOptions};
 use futures::future::join_all;
 use rustfs_filemeta::{MetaCacheEntries, MetaCacheEntry, MetacacheReader, is_io_eof};
-use std::{future::Future, pin::Pin, sync::Arc};
+use std::{future::Future, pin::Pin};
 use tokio::spawn;
 use tokio_util::sync::CancellationToken;
 use tracing::{error, info, warn};
@@ -71,14 +71,14 @@ pub async fn list_path_raw(rx: CancellationToken, opts: ListPathRawOptions) -> d

    let mut jobs: Vec<tokio::task::JoinHandle<std::result::Result<(), DiskError>>> = Vec::new();
    let mut readers = Vec::with_capacity(opts.disks.len());
-    let fds = Arc::new(opts.fallback_disks.clone());
+    let fds = opts.fallback_disks.iter().flatten().cloned().collect::<Vec<_>>();

    let cancel_rx = CancellationToken::new();

    for disk in opts.disks.iter() {
        let opdisk = disk.clone();
        let opts_clone = opts.clone();
-        let fds_clone = fds.clone();
+        let mut fds_clone = fds.clone();
        let cancel_rx_clone = cancel_rx.clone();
        let (rd, mut wr) = tokio::io::duplex(64);
        readers.push(MetacacheReader::new(rd));
@@ -113,21 +113,20 @@ pub async fn list_path_raw(rx: CancellationToken, opts: ListPathRawOptions) -> d
            }

            while need_fallback {
-                // warn!("list_path_raw: while need_fallback start");
-                let disk = match fds_clone.iter().find(|d| d.is_some()) {
-                    Some(d) => {
-                        if let Some(disk) = d.clone() {
-                            disk
-                        } else {
-                            warn!("list_path_raw: fallback disk is none");
-                            break;
-                        }
-                    }
-                    None => {
-                        warn!("list_path_raw: fallback disk is none2");
-                        break;
+                let disk_op = {
+                    if fds_clone.is_empty() {
+                        None
+                    } else {
+                        let disk = fds_clone.remove(0);
+                        if disk.is_online().await { Some(disk.clone()) } else { None }
                    }
                };
+
+                let Some(disk) = disk_op else {
+                    warn!("list_path_raw: fallback disk is none");
+                    break;
+                };
+
                match disk
                    .as_ref()
                    .walk_dir(
--- a/crates/ecstore/src/checksum.rs
+++ b/crates/ecstore/src/checksum.rs
@@ -1,350 +0,0 @@
-#![allow(clippy::map_entry)]
-// Copyright 2024 RustFS Team
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-#![allow(unused_imports)]
-#![allow(unused_variables)]
-#![allow(unused_mut)]
-#![allow(unused_assignments)]
-#![allow(unused_must_use)]
-#![allow(clippy::all)]
-
-use lazy_static::lazy_static;
-use rustfs_checksums::ChecksumAlgorithm;
-use std::collections::HashMap;
-
-use crate::client::{api_put_object::PutObjectOptions, api_s3_datatypes::ObjectPart};
-use crate::{disk::DiskAPI, store_api::GetObjectReader};
-use rustfs_utils::crypto::{base64_decode, base64_encode};
-use s3s::header::{
-    X_AMZ_CHECKSUM_ALGORITHM, X_AMZ_CHECKSUM_CRC32, X_AMZ_CHECKSUM_CRC32C, X_AMZ_CHECKSUM_SHA1, X_AMZ_CHECKSUM_SHA256,
-};
-
-use enumset::{EnumSet, EnumSetType, enum_set};
-
-#[derive(Debug, EnumSetType, Default)]
-#[enumset(repr = "u8")]
-pub enum ChecksumMode {
-    #[default]
-    ChecksumNone,
-    ChecksumSHA256,
-    ChecksumSHA1,
-    ChecksumCRC32,
-    ChecksumCRC32C,
-    ChecksumCRC64NVME,
-    ChecksumFullObject,
-}
-
-lazy_static! {
-    static ref C_ChecksumMask: EnumSet<ChecksumMode> = {
-        let mut s = EnumSet::all();
-        s.remove(ChecksumMode::ChecksumFullObject);
-        s
-    };
-    static ref C_ChecksumFullObjectCRC32: EnumSet<ChecksumMode> =
-        enum_set!(ChecksumMode::ChecksumCRC32 | ChecksumMode::ChecksumFullObject);
-    static ref C_ChecksumFullObjectCRC32C: EnumSet<ChecksumMode> =
-        enum_set!(ChecksumMode::ChecksumCRC32C | ChecksumMode::ChecksumFullObject);
-}
-const AMZ_CHECKSUM_CRC64NVME: &str = "x-amz-checksum-crc64nvme";
-
-impl ChecksumMode {
-    //pub const CRC64_NVME_POLYNOMIAL: i64 = 0xad93d23594c93659;
-
-    pub fn base(&self) -> ChecksumMode {
-        let s = EnumSet::from(*self).intersection(*C_ChecksumMask);
-        match s.as_u8() {
-            1_u8 => ChecksumMode::ChecksumNone,
-            2_u8 => ChecksumMode::ChecksumSHA256,
-            4_u8 => ChecksumMode::ChecksumSHA1,
-            8_u8 => ChecksumMode::ChecksumCRC32,
-            16_u8 => ChecksumMode::ChecksumCRC32C,
-            32_u8 => ChecksumMode::ChecksumCRC64NVME,
-            _ => panic!("enum err."),
-        }
-    }
-
-    pub fn is(&self, t: ChecksumMode) -> bool {
-        *self & t == t
-    }
-
-    pub fn key(&self) -> String {
-        //match c & checksumMask {
-        match self {
-            ChecksumMode::ChecksumCRC32 => {
-                return X_AMZ_CHECKSUM_CRC32.to_string();
-            }
-            ChecksumMode::ChecksumCRC32C => {
-                return X_AMZ_CHECKSUM_CRC32C.to_string();
-            }
-            ChecksumMode::ChecksumSHA1 => {
-                return X_AMZ_CHECKSUM_SHA1.to_string();
-            }
-            ChecksumMode::ChecksumSHA256 => {
-                return X_AMZ_CHECKSUM_SHA256.to_string();
-            }
-            ChecksumMode::ChecksumCRC64NVME => {
-                return AMZ_CHECKSUM_CRC64NVME.to_string();
-            }
-            _ => {
-                return "".to_string();
-            }
-        }
-    }
-
-    pub fn can_composite(&self) -> bool {
-        let s = EnumSet::from(*self).intersection(*C_ChecksumMask);
-        match s.as_u8() {
-            2_u8 => true,
-            4_u8 => true,
-            8_u8 => true,
-            16_u8 => true,
-            _ => false,
-        }
-    }
-
-    pub fn can_merge_crc(&self) -> bool {
-        let s = EnumSet::from(*self).intersection(*C_ChecksumMask);
-        match s.as_u8() {
-            8_u8 => true,
-            16_u8 => true,
-            32_u8 => true,
-            _ => false,
-        }
-    }
-
-    pub fn full_object_requested(&self) -> bool {
-        let s = EnumSet::from(*self).intersection(*C_ChecksumMask);
-        match s.as_u8() {
-            //C_ChecksumFullObjectCRC32 as u8 => true,
-            //C_ChecksumFullObjectCRC32C as u8 => true,
-            32_u8 => true,
-            _ => false,
-        }
-    }
-
-    pub fn key_capitalized(&self) -> String {
-        self.key()
-    }
-
-    pub fn raw_byte_len(&self) -> usize {
-        let u = EnumSet::from(*self).intersection(*C_ChecksumMask).as_u8();
-        if u == ChecksumMode::ChecksumCRC32 as u8 || u == ChecksumMode::ChecksumCRC32C as u8 {
-            4
-        } else if u == ChecksumMode::ChecksumSHA1 as u8 {
-            use sha1::Digest;
-            sha1::Sha1::output_size() as usize
-        } else if u == ChecksumMode::ChecksumSHA256 as u8 {
-            use sha2::Digest;
-            sha2::Sha256::output_size() as usize
-        } else if u == ChecksumMode::ChecksumCRC64NVME as u8 {
-            8
-        } else {
-            0
-        }
-    }
-
-    pub fn hasher(&self) -> Result<Box<dyn rustfs_checksums::http::HttpChecksum>, std::io::Error> {
-        match /*C_ChecksumMask & **/self {
-            ChecksumMode::ChecksumCRC32 => {
-                return Ok(ChecksumAlgorithm::Crc32.into_impl());
-            }
-            ChecksumMode::ChecksumCRC32C => {
-                return Ok(ChecksumAlgorithm::Crc32c.into_impl());
-            }
-            ChecksumMode::ChecksumSHA1 => {
-                return Ok(ChecksumAlgorithm::Sha1.into_impl());
-            }
-            ChecksumMode::ChecksumSHA256 => {
-                return Ok(ChecksumAlgorithm::Sha256.into_impl());
-            }
-            ChecksumMode::ChecksumCRC64NVME => {
-                return Ok(ChecksumAlgorithm::Crc64Nvme.into_impl());
-            }
-            _ => return Err(std::io::Error::other("unsupported checksum type")),
-        }
-    }
-
-    pub fn is_set(&self) -> bool {
-        let s = EnumSet::from(*self).intersection(*C_ChecksumMask);
-        s.len() == 1
-    }
-
-    pub fn set_default(&mut self, t: ChecksumMode) {
-        if !self.is_set() {
-            *self = t;
-        }
-    }
-
-    pub fn encode_to_string(&self, b: &[u8]) -> Result<String, std::io::Error> {
-        if !self.is_set() {
-            return Ok("".to_string());
-        }
-        let mut h = self.hasher()?;
-        h.update(b);
-        let hash = h.finalize();
-        Ok(base64_encode(hash.as_ref()))
-    }
-
-    pub fn to_string(&self) -> String {
-        //match c & checksumMask {
-        match self {
-            ChecksumMode::ChecksumCRC32 => {
-                return "CRC32".to_string();
-            }
-            ChecksumMode::ChecksumCRC32C => {
-                return "CRC32C".to_string();
-            }
-            ChecksumMode::ChecksumSHA1 => {
-                return "SHA1".to_string();
-            }
-            ChecksumMode::ChecksumSHA256 => {
-                return "SHA256".to_string();
-            }
-            ChecksumMode::ChecksumNone => {
-                return "".to_string();
-            }
-            ChecksumMode::ChecksumCRC64NVME => {
-                return "CRC64NVME".to_string();
-            }
-            _ => {
-                return "<invalid>".to_string();
-            }
-        }
-    }
-
-    // pub fn check_sum_reader(&self, r: GetObjectReader) -> Result<Checksum, std::io::Error> {
-    //     let mut h = self.hasher()?;
-    //     Ok(Checksum::new(self.clone(), h.sum().as_bytes()))
-    // }
-
-    // pub fn check_sum_bytes(&self, b: &[u8]) -> Result<Checksum, std::io::Error> {
-    //     let mut h = self.hasher()?;
-    //     Ok(Checksum::new(self.clone(), h.sum().as_bytes()))
-    // }
-
-    pub fn composite_checksum(&self, p: &mut [ObjectPart]) -> Result<Checksum, std::io::Error> {
-        if !self.can_composite() {
-            return Err(std::io::Error::other("cannot do composite checksum"));
-        }
-        p.sort_by(|i, j| {
-            if i.part_num < j.part_num {
-                std::cmp::Ordering::Less
-            } else if i.part_num > j.part_num {
-                std::cmp::Ordering::Greater
-            } else {
-                std::cmp::Ordering::Equal
-            }
-        });
-        let c = self.base();
-        let crc_bytes = Vec::<u8>::with_capacity(p.len() * self.raw_byte_len() as usize);
-        let mut h = self.hasher()?;
-        h.update(crc_bytes.as_ref());
-        let hash = h.finalize();
-        Ok(Checksum {
-            checksum_type: self.clone(),
-            r: hash.as_ref().to_vec(),
-            computed: false,
-        })
-    }
-
-    pub fn full_object_checksum(&self, p: &mut [ObjectPart]) -> Result<Checksum, std::io::Error> {
-        todo!();
-    }
-}
-
-#[derive(Default)]
-pub struct Checksum {
-    checksum_type: ChecksumMode,
-    r: Vec<u8>,
-    computed: bool,
-}
-
-#[allow(dead_code)]
-impl Checksum {
-    fn new(t: ChecksumMode, b: &[u8]) -> Checksum {
-        if t.is_set() && b.len() == t.raw_byte_len() {
-            return Checksum {
-                checksum_type: t,
-                r: b.to_vec(),
-                computed: false,
-            };
-        }
-        Checksum::default()
-    }
-
-    #[allow(dead_code)]
-    fn new_checksum_string(t: ChecksumMode, s: &str) -> Result<Checksum, std::io::Error> {
-        let b = match base64_decode(s.as_bytes()) {
-            Ok(b) => b,
-            Err(err) => return Err(std::io::Error::other(err.to_string())),
-        };
-        if t.is_set() && b.len() == t.raw_byte_len() {
-            return Ok(Checksum {
-                checksum_type: t,
-                r: b,
-                computed: false,
-            });
-        }
-        Ok(Checksum::default())
-    }
-
-    fn is_set(&self) -> bool {
-        self.checksum_type.is_set() && self.r.len() == self.checksum_type.raw_byte_len()
-    }
-
-    fn encoded(&self) -> String {
-        if !self.is_set() {
-            return "".to_string();
-        }
-        base64_encode(&self.r)
-    }
-
-    #[allow(dead_code)]
-    fn raw(&self) -> Option<Vec<u8>> {
-        if !self.is_set() {
-            return None;
-        }
-        Some(self.r.clone())
-    }
-}
-
-pub fn add_auto_checksum_headers(opts: &mut PutObjectOptions) {
-    opts.user_metadata
-        .insert("X-Amz-Checksum-Algorithm".to_string(), opts.auto_checksum.to_string());
-    if opts.auto_checksum.full_object_requested() {
-        opts.user_metadata
-            .insert("X-Amz-Checksum-Type".to_string(), "FULL_OBJECT".to_string());
-    }
-}
-
-pub fn apply_auto_checksum(opts: &mut PutObjectOptions, all_parts: &mut [ObjectPart]) -> Result<(), std::io::Error> {
-    if opts.auto_checksum.can_composite() && !opts.auto_checksum.is(ChecksumMode::ChecksumFullObject) {
-        let crc = opts.auto_checksum.composite_checksum(all_parts)?;
-        opts.user_metadata = {
-            let mut hm = HashMap::new();
-            hm.insert(opts.auto_checksum.key(), crc.encoded());
-            hm
-        }
-    } else if opts.auto_checksum.can_merge_crc() {
-        let crc = opts.auto_checksum.full_object_checksum(all_parts)?;
-        opts.user_metadata = {
-            let mut hm = HashMap::new();
-            hm.insert(opts.auto_checksum.key_capitalized(), crc.encoded());
-            hm.insert("X-Amz-Checksum-Type".to_string(), "FULL_OBJECT".to_string());
-            hm
-        }
-    }
-
-    Ok(())
-}
--- a/crates/ecstore/src/chunk_stream.rs
+++ b/crates/ecstore/src/chunk_stream.rs
@@ -1,270 +0,0 @@
-// Copyright 2024 RustFS Team
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-// use crate::error::StdError;
-// use bytes::Bytes;
-// use futures::pin_mut;
-// use futures::stream::{Stream, StreamExt};
-// use std::future::Future;
-// use std::pin::Pin;
-// use std::task::{Context, Poll};
-// use transform_stream::AsyncTryStream;
-
-// pub type SyncBoxFuture<'a, T> = Pin<Box<dyn Future<Output = T> + Send + Sync + 'a>>;
-
-// pub struct ChunkedStream<'a> {
-//     /// inner
-//     inner: AsyncTryStream<Bytes, StdError, SyncBoxFuture<'a, Result<(), StdError>>>,
-
-//     remaining_length: usize,
-// }
-
-// impl<'a> ChunkedStream<'a> {
-//     pub fn new<S>(body: S, content_length: usize, chunk_size: usize, need_padding: bool) -> Self
-//     where
-//         S: Stream<Item = Result<Bytes, StdError>> + Send + Sync + 'a,
-//     {
-//         let inner = AsyncTryStream::<_, _, SyncBoxFuture<'a, Result<(), StdError>>>::new(|mut y| {
-//             #[allow(clippy::shadow_same)] // necessary for `pin_mut!`
-//             Box::pin(async move {
-//                 pin_mut!(body);
-//                 // Data left over from the previous call
-//                 let mut prev_bytes = Bytes::new();
-//                 let mut read_size = 0;
-
-//                 loop {
-//                     let data: Vec<Bytes> = {
-//                         // Read a fixed-size chunk
-//                         match Self::read_data(body.as_mut(), prev_bytes, chunk_size).await {
-//                             None => break,
-//                             Some(Err(e)) => return Err(e),
-//                             Some(Ok((data, remaining_bytes))) => {
-//                                 // debug!(
-//                                 //     "content_length:{},read_size:{}, read_data data:{}, remaining_bytes: {} ",
-//                                 //     content_length,
-//                                 //     read_size,
-//                                 //     data.len(),
-//                                 //     remaining_bytes.len()
-//                                 // );
-
-//                                 prev_bytes = remaining_bytes;
-//                                 data
-//                             }
-//                         }
-//                     };
-
-//                     for bytes in data {
-//                         read_size += bytes.len();
-//                         // debug!("read_size {}, content_length {}", read_size, content_length,);
-//                         y.yield_ok(bytes).await;
-//                     }
-
-//                     if read_size + prev_bytes.len() >= content_length {
-//                         // debug!(
-//                         //     "Finished reading: read_size:{} + prev_bytes.len({}) == content_length {}",
-//                         //     read_size,
-//                         //     prev_bytes.len(),
-//                         //     content_length,
-//                         // );
-
-//                         // Pad with zeros?
-//                         if !need_padding {
-//                             y.yield_ok(prev_bytes).await;
-//                             break;
-//                         }
-
-//                         let mut bytes = vec![0u8; chunk_size];
-//                         let (left, _) = bytes.split_at_mut(prev_bytes.len());
-//                         left.copy_from_slice(&prev_bytes);
-
-//                         y.yield_ok(Bytes::from(bytes)).await;
-
-//                         break;
-//                     }
-//                 }
-
-//                 // debug!("chunked stream exit");
-
-//                 Ok(())
-//             })
-//         });
-//         Self {
-//             inner,
-//             remaining_length: content_length,
-//         }
-//     }
-//     /// read data and return remaining bytes
-//     async fn read_data<S>(
-//         mut body: Pin<&mut S>,
-//         prev_bytes: Bytes,
-//         data_size: usize,
-//     ) -> Option<Result<(Vec<Bytes>, Bytes), StdError>>
-//     where
-//         S: Stream<Item = Result<Bytes, StdError>> + Send,
-//     {
-//         let mut bytes_buffer = Vec::new();
-
-//         // Run only once
-//         let mut push_data_bytes = |mut bytes: Bytes| {
-//             // debug!("read from body {} split per {}, prev_bytes: {}", bytes.len(), data_size, prev_bytes.len());
-
-//             if bytes.is_empty() {
-//                 return None;
-//             }
-
-//             if data_size == 0 {
-//                 return Some(bytes);
-//             }
-
-//             // Merge with the previous data
-//             if !prev_bytes.is_empty() {
-//                 let need_size = data_size.wrapping_sub(prev_bytes.len());
-//                 // debug!(
-//                 //     "Previous leftover {}, take {} now, total: {}",
-//                 //     prev_bytes.len(),
-//                 //     need_size,
-//                 //     prev_bytes.len() + need_size
-//                 // );
-//                 if bytes.len() >= need_size {
-//                     let data = bytes.split_to(need_size);
-//                     let mut combined = Vec::new();
-//                     combined.extend_from_slice(&prev_bytes);
-//                     combined.extend_from_slice(&data);
-
-//                     // debug!(
-//                     //     "Fetched more bytes than needed: {}, merged result {}, remaining bytes {}",
-//                     //     need_size,
-//                     //     combined.len(),
-//                     //     bytes.len(),
-//                     // );
-
-//                     bytes_buffer.push(Bytes::from(combined));
-//                 } else {
-//                     let mut combined = Vec::new();
-//                     combined.extend_from_slice(&prev_bytes);
-//                     combined.extend_from_slice(&bytes);
-
-//                     // debug!(
-//                     //     "Fetched fewer bytes than needed: {}, merged result {}, remaining bytes {}, return immediately",
-//                     //     need_size,
-//                     //     combined.len(),
-//                     //     bytes.len(),
-//                     // );
-
-//                     return Some(Bytes::from(combined));
-//                 }
-//             }
-
-//             // If the fetched data exceeds the chunk, slice the required size
-//             if data_size <= bytes.len() {
-//                 let n = bytes.len() / data_size;
-
-//                 for _ in 0..n {
-//                     let data = bytes.split_to(data_size);
-
-//                     // println!("bytes_buffer.push: {}, remaining: {}", data.len(), bytes.len());
-//                     bytes_buffer.push(data);
-//                 }
-
-//                 Some(bytes)
-//             } else {
-//                 // Insufficient data
-//                 Some(bytes)
-//             }
-//         };
-
-//         // Remaining data
-//         let remaining_bytes = 'outer: {
-//             // // Exit if the previous data was sufficient
-//             // if let Some(remaining_bytes) = push_data_bytes(prev_bytes) {
-//             //     println!("Consuming leftovers");
-//             //     break 'outer remaining_bytes;
-//             // }
-
-//             loop {
-//                 match body.next().await? {
-//                     Err(e) => return Some(Err(e)),
-//                     Ok(bytes) => {
-//                         if let Some(remaining_bytes) = push_data_bytes(bytes) {
-//                             break 'outer remaining_bytes;
-//                         }
-//                     }
-//                 }
-//             }
-//         };
-
-//         Some(Ok((bytes_buffer, remaining_bytes)))
-//     }
-
-//     fn poll(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Option<Result<Bytes, StdError>>> {
-//         let ans = Pin::new(&mut self.inner).poll_next(cx);
-//         if let Poll::Ready(Some(Ok(ref bytes))) = ans {
-//             self.remaining_length = self.remaining_length.saturating_sub(bytes.len());
-//         }
-//         ans
-//     }
-
-//     // pub fn exact_remaining_length(&self) -> usize {
-//     //     self.remaining_length
-//     // }
-// }
-
-// impl Stream for ChunkedStream<'_> {
-//     type Item = Result<Bytes, StdError>;
-
-//     fn poll_next(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Option<Self::Item>> {
-//         self.poll(cx)
-//     }
-
-//     fn size_hint(&self) -> (usize, Option<usize>) {
-//         (0, None)
-//     }
-// }
-
-// #[cfg(test)]
-// mod test {
-
-//     use super::*;
-
-//     #[tokio::test]
-//     async fn test_chunked_stream() {
-//         let chunk_size = 4;
-
-//         let data1 = vec![1u8; 7777]; // 65536
-//         let data2 = vec![1u8; 7777]; // 65536
-
-//         let content_length = data1.len() + data2.len();
-
-//         let chunk1 = Bytes::from(data1);
-//         let chunk2 = Bytes::from(data2);
-
-//         let chunk_results: Vec<Result<Bytes, _>> = vec![Ok(chunk1), Ok(chunk2)];
-
-//         let stream = futures::stream::iter(chunk_results);
-
-//         let mut chunked_stream = ChunkedStream::new(stream, content_length, chunk_size, true);
-
-//         loop {
-//             let ans1 = chunked_stream.next().await;
-//             if ans1.is_none() {
-//                 break;
-//             }
-
-//             let bytes = ans1.unwrap().unwrap();
-//             assert!(bytes.len() == chunk_size)
-//         }
-
-//         // assert_eq!(ans1.unwrap(), chunk1_data.as_slice());
-//     }
-// }
--- a/crates/ecstore/src/client/hook_reader.rs
+++ b/crates/ecstore/src/client/hook_reader.rs
@@ -1,59 +0,0 @@
-#![allow(clippy::map_entry)]
-// Copyright 2024 RustFS Team
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-use std::{collections::HashMap, sync::Arc};
-
-use crate::{
-    disk::{
-        error::{is_unformatted_disk, DiskError},
-        format::{DistributionAlgoVersion, FormatV3},
-        new_disk, DiskAPI, DiskInfo, DiskOption, DiskStore,
-    },
-    store_api::{
-        BucketInfo, BucketOptions, CompletePart, DeleteBucketOptions, DeletedObject, GetObjectReader, HTTPRangeSpec,
-        ListMultipartsInfo, ListObjectVersionsInfo, ListObjectsV2Info, MakeBucketOptions, MultipartInfo, MultipartUploadResult,
-        ObjectIO, ObjectInfo, ObjectOptions, ObjectToDelete, PartInfo, PutObjReader, StorageAPI,
-    },
-    credentials::{Credentials, SignatureType,},
-    api_put_object_multipart::UploadPartParams,
-};
-
-use http::HeaderMap;
-use tokio_util::sync::CancellationToken;
-use tracing::warn;
-use tracing::{error, info};
-use url::Url;
-
-struct HookReader {
-    source: GetObjectReader,
-    hook:   GetObjectReader,
-}
-
-impl HookReader {
-    pub fn new(source: GetObjectReader, hook: GetObjectReader) -> HookReader {
-        HookReader {
-            source,
-            hook,
-        }
-    }
-
-    fn seek(&self, offset: i64, whence: i64) -> Result<i64> {
-        todo!();
-    }
-
-    fn read(&self, b: &[u8]) -> Result<i64> {
-        todo!();
-    }
-}
--- a/crates/ecstore/src/disk/disk_store.rs
+++ b/crates/ecstore/src/disk/disk_store.rs
@@ -0,0 +1,770 @@
+// Copyright 2024 RustFS Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use crate::disk::{
+    CheckPartsResp, DeleteOptions, DiskAPI, DiskError, DiskInfo, DiskInfoOptions, DiskLocation, Endpoint, Error,
+    FileInfoVersions, ReadMultipleReq, ReadMultipleResp, ReadOptions, RenameDataResp, Result, UpdateMetadataOpts, VolumeInfo,
+    WalkDirOptions, local::LocalDisk,
+};
+use bytes::Bytes;
+use rustfs_filemeta::{FileInfo, ObjectPartInfo, RawFileInfo};
+use rustfs_utils::string::parse_bool_with_default;
+use std::{
+    path::PathBuf,
+    sync::{
+        Arc,
+        atomic::{AtomicI64, AtomicU32, Ordering},
+    },
+    time::Duration,
+};
+use tokio::{sync::RwLock, time};
+use tokio_util::sync::CancellationToken;
+use tracing::{debug, info, warn};
+use uuid::Uuid;
+
+/// Disk health status constants
+const DISK_HEALTH_OK: u32 = 0;
+const DISK_HEALTH_FAULTY: u32 = 1;
+
+pub const ENV_RUSTFS_DRIVE_ACTIVE_MONITORING: &str = "RUSTFS_DRIVE_ACTIVE_MONITORING";
+pub const ENV_RUSTFS_DRIVE_MAX_TIMEOUT_DURATION: &str = "RUSTFS_DRIVE_MAX_TIMEOUT_DURATION";
+pub const CHECK_EVERY: Duration = Duration::from_secs(15);
+pub const SKIP_IF_SUCCESS_BEFORE: Duration = Duration::from_secs(5);
+pub const CHECK_TIMEOUT_DURATION: Duration = Duration::from_secs(5);
+
+lazy_static::lazy_static! {
+    static ref TEST_OBJ: String = format!("health-check-{}", Uuid::new_v4());
+    static ref TEST_DATA: Bytes = Bytes::from(vec![42u8; 2048]);
+    static ref TEST_BUCKET: String = ".rustfs.sys/tmp".to_string();
+}
+
+pub fn get_max_timeout_duration() -> Duration {
+    std::env::var(ENV_RUSTFS_DRIVE_MAX_TIMEOUT_DURATION)
+        .map(|v| Duration::from_secs(v.parse::<u64>().unwrap_or(30)))
+        .unwrap_or(Duration::from_secs(30))
+}
+
+/// DiskHealthTracker tracks the health status of a disk.
+/// Similar to Go's diskHealthTracker.
+#[derive(Debug)]
+pub struct DiskHealthTracker {
+    /// Atomic timestamp of last successful operation
+    pub last_success: AtomicI64,
+    /// Atomic timestamp of last operation start
+    pub last_started: AtomicI64,
+    /// Atomic disk status (OK or Faulty)
+    pub status: AtomicU32,
+    /// Atomic number of waiting operations
+    pub waiting: AtomicU32,
+}
+
+impl DiskHealthTracker {
+    /// Create a new disk health tracker
+    pub fn new() -> Self {
+        let now = std::time::SystemTime::now()
+            .duration_since(std::time::UNIX_EPOCH)
+            .unwrap()
+            .as_nanos() as i64;
+
+        Self {
+            last_success: AtomicI64::new(now),
+            last_started: AtomicI64::new(now),
+            status: AtomicU32::new(DISK_HEALTH_OK),
+            waiting: AtomicU32::new(0),
+        }
+    }
+
+    /// Log a successful operation
+    pub fn log_success(&self) {
+        let now = std::time::SystemTime::now()
+            .duration_since(std::time::UNIX_EPOCH)
+            .unwrap()
+            .as_nanos() as i64;
+        self.last_success.store(now, Ordering::Relaxed);
+    }
+
+    /// Check if disk is faulty
+    pub fn is_faulty(&self) -> bool {
+        self.status.load(Ordering::Relaxed) == DISK_HEALTH_FAULTY
+    }
+
+    /// Set disk as faulty
+    pub fn set_faulty(&self) {
+        self.status.store(DISK_HEALTH_FAULTY, Ordering::Relaxed);
+    }
+
+    /// Set disk as OK
+    pub fn set_ok(&self) {
+        self.status.store(DISK_HEALTH_OK, Ordering::Relaxed);
+    }
+
+    pub fn swap_ok_to_faulty(&self) -> bool {
+        self.status
+            .compare_exchange(DISK_HEALTH_OK, DISK_HEALTH_FAULTY, Ordering::Relaxed, Ordering::Relaxed)
+            .is_ok()
+    }
+
+    /// Increment waiting operations counter
+    pub fn increment_waiting(&self) {
+        self.waiting.fetch_add(1, Ordering::Relaxed);
+    }
+
+    /// Decrement waiting operations counter
+    pub fn decrement_waiting(&self) {
+        self.waiting.fetch_sub(1, Ordering::Relaxed);
+    }
+
+    /// Get waiting operations count
+    pub fn waiting_count(&self) -> u32 {
+        self.waiting.load(Ordering::Relaxed)
+    }
+
+    /// Get last success timestamp
+    pub fn last_success(&self) -> i64 {
+        self.last_success.load(Ordering::Relaxed)
+    }
+}
+
+impl Default for DiskHealthTracker {
+    fn default() -> Self {
+        Self::new()
+    }
+}
+
+/// Health check context key for tracking disk operations
+#[derive(Debug, Clone)]
+struct HealthDiskCtxKey;
+
+#[derive(Debug)]
+struct HealthDiskCtxValue {
+    last_success: Arc<AtomicI64>,
+}
+
+impl HealthDiskCtxValue {
+    fn log_success(&self) {
+        let now = std::time::SystemTime::now()
+            .duration_since(std::time::UNIX_EPOCH)
+            .unwrap()
+            .as_nanos() as i64;
+        self.last_success.store(now, Ordering::Relaxed);
+    }
+}
+
+/// LocalDiskWrapper wraps a DiskStore with health tracking capabilities.
+/// This is similar to Go's xlStorageDiskIDCheck.
+#[derive(Debug, Clone)]
+pub struct LocalDiskWrapper {
+    /// The underlying disk store
+    disk: Arc<LocalDisk>,
+    /// Health tracker
+    health: Arc<DiskHealthTracker>,
+    /// Whether health checking is enabled
+    health_check: bool,
+    /// Cancellation token for monitoring tasks
+    cancel_token: CancellationToken,
+    /// Disk ID for stale checking
+    disk_id: Arc<RwLock<Option<Uuid>>>,
+}
+
+impl LocalDiskWrapper {
+    /// Create a new LocalDiskWrapper
+    pub fn new(disk: Arc<LocalDisk>, health_check: bool) -> Self {
+        // Check environment variable for health check override
+        // Default to true if not set, but only enable if both param and env are true
+        let env_health_check = std::env::var(ENV_RUSTFS_DRIVE_ACTIVE_MONITORING)
+            .map(|v| parse_bool_with_default(&v, true))
+            .unwrap_or(true);
+
+        let ret = Self {
+            disk,
+            health: Arc::new(DiskHealthTracker::new()),
+            health_check: health_check && env_health_check,
+            cancel_token: CancellationToken::new(),
+            disk_id: Arc::new(RwLock::new(None)),
+        };
+
+        ret.start_monitoring();
+
+        ret
+    }
+
+    pub fn get_disk(&self) -> Arc<LocalDisk> {
+        self.disk.clone()
+    }
+
+    /// Start the disk monitoring if health_check is enabled
+    pub fn start_monitoring(&self) {
+        if self.health_check {
+            let health = Arc::clone(&self.health);
+            let cancel_token = self.cancel_token.clone();
+            let disk = Arc::clone(&self.disk);
+
+            tokio::spawn(async move {
+                Self::monitor_disk_writable(disk, health, cancel_token).await;
+            });
+        }
+    }
+
+    /// Stop the disk monitoring
+    pub async fn stop_monitoring(&self) {
+        self.cancel_token.cancel();
+    }
+
+    /// Monitor disk writability periodically
+    async fn monitor_disk_writable(disk: Arc<LocalDisk>, health: Arc<DiskHealthTracker>, cancel_token: CancellationToken) {
+        // TODO: config interval
+
+        let mut interval = time::interval(CHECK_EVERY);
+
+        loop {
+            tokio::select! {
+                _ = cancel_token.cancelled() => {
+                    return;
+                }
+                _ = interval.tick() => {
+                    if cancel_token.is_cancelled() {
+                        return;
+                    }
+
+                    if health.status.load(Ordering::Relaxed) != DISK_HEALTH_OK {
+                        continue;
+                    }
+
+                    let last_success_nanos = health.last_success.load(Ordering::Relaxed);
+                    let elapsed = Duration::from_nanos(
+                        (std::time::SystemTime::now()
+                            .duration_since(std::time::UNIX_EPOCH)
+                            .unwrap()
+                            .as_nanos() as i64 - last_success_nanos) as u64
+                    );
+
+                    if elapsed < SKIP_IF_SUCCESS_BEFORE {
+                        continue;
+                    }
+
+                    tokio::time::sleep(Duration::from_secs(1)).await;
+
+
+                    debug!("health check: performing health check");
+                    if Self::perform_health_check(disk.clone(), &TEST_BUCKET, &TEST_OBJ, &TEST_DATA, true, CHECK_TIMEOUT_DURATION).await.is_err() && health.swap_ok_to_faulty() {
+                        // Health check failed, disk is considered faulty
+
+                        health.increment_waiting(); // Balance the increment from failed operation
+
+                        let health_clone = Arc::clone(&health);
+                        let disk_clone = disk.clone();
+                        let cancel_clone = cancel_token.clone();
+
+                        tokio::spawn(async move {
+                            Self::monitor_disk_status(disk_clone, health_clone, cancel_clone).await;
+                        });
+                    }
+                }
+            }
+        }
+    }
+
+    /// Perform a health check by writing and reading a test file
+    async fn perform_health_check(
+        disk: Arc<LocalDisk>,
+        test_bucket: &str,
+        test_filename: &str,
+        test_data: &Bytes,
+        check_faulty_only: bool,
+        timeout_duration: Duration,
+    ) -> Result<()> {
+        // Perform health check with timeout
+        let health_check_result = tokio::time::timeout(timeout_duration, async {
+            // Try to write test data
+            disk.write_all(test_bucket, test_filename, test_data.clone()).await?;
+
+            // Try to read back the data
+            let read_data = disk.read_all(test_bucket, test_filename).await?;
+
+            // Verify data integrity
+            if read_data.len() != test_data.len() {
+                warn!(
+                    "health check: test file data length mismatch: expected {} bytes, got {}",
+                    test_data.len(),
+                    read_data.len()
+                );
+                if check_faulty_only {
+                    return Ok(());
+                }
+                return Err(DiskError::FaultyDisk);
+            }
+
+            // Clean up
+            disk.delete(
+                test_bucket,
+                test_filename,
+                DeleteOptions {
+                    recursive: false,
+                    immediate: false,
+                    undo_write: false,
+                    old_data_dir: None,
+                },
+            )
+            .await?;
+
+            Ok(())
+        })
+        .await;
+
+        match health_check_result {
+            Ok(result) => match result {
+                Ok(()) => Ok(()),
+                Err(e) => {
+                    debug!("health check: failed: {:?}", e);
+
+                    if e == DiskError::FaultyDisk {
+                        return Err(e);
+                    }
+
+                    if check_faulty_only { Ok(()) } else { Err(e) }
+                }
+            },
+            Err(_) => {
+                // Timeout occurred
+                warn!("health check: timeout after {:?}", timeout_duration);
+                Err(DiskError::FaultyDisk)
+            }
+        }
+    }
+
+    /// Monitor disk status and try to bring it back online
+    async fn monitor_disk_status(disk: Arc<LocalDisk>, health: Arc<DiskHealthTracker>, cancel_token: CancellationToken) {
+        const CHECK_EVERY: Duration = Duration::from_secs(5);
+
+        let mut interval = time::interval(CHECK_EVERY);
+
+        loop {
+            tokio::select! {
+                _ = cancel_token.cancelled() => {
+                    return;
+                }
+                _ = interval.tick() => {
+                    if cancel_token.is_cancelled() {
+                        return;
+                    }
+
+                    match Self::perform_health_check(disk.clone(), &TEST_BUCKET, &TEST_OBJ, &TEST_DATA, false, CHECK_TIMEOUT_DURATION).await {
+                        Ok(_) => {
+                            info!("Disk {} is back online", disk.to_string());
+                            health.set_ok();
+                            health.decrement_waiting();
+                            return;
+                        }
+                        Err(e) => {
+                            warn!("Disk {} still faulty: {:?}", disk.to_string(), e);
+                        }
+                    }
+                }
+            }
+        }
+    }
+
+    async fn check_id(&self, want_id: Option<Uuid>) -> Result<()> {
+        if want_id.is_none() {
+            return Ok(());
+        }
+
+        let stored_disk_id = self.disk.get_disk_id().await?;
+
+        if stored_disk_id != want_id {
+            return Err(Error::other(format!("Disk ID mismatch wanted {:?}, got {:?}", want_id, stored_disk_id)));
+        }
+
+        Ok(())
+    }
+
+    /// Check if disk ID is stale
+    async fn check_disk_stale(&self) -> Result<()> {
+        let Some(current_disk_id) = *self.disk_id.read().await else {
+            return Ok(());
+        };
+
+        let stored_disk_id = match self.disk.get_disk_id().await? {
+            Some(id) => id,
+            None => return Ok(()), // Empty disk ID is allowed during initialization
+        };
+
+        if current_disk_id != stored_disk_id {
+            return Err(DiskError::DiskNotFound);
+        }
+
+        Ok(())
+    }
+
+    /// Set the disk ID
+    pub async fn set_disk_id_internal(&self, id: Option<Uuid>) -> Result<()> {
+        let mut disk_id = self.disk_id.write().await;
+        *disk_id = id;
+        Ok(())
+    }
+
+    /// Get the current disk ID
+    pub async fn get_current_disk_id(&self) -> Option<Uuid> {
+        *self.disk_id.read().await
+    }
+
+    /// Track disk health for an operation.
+    /// This method should wrap disk operations to ensure health checking.
+    pub async fn track_disk_health<T, F, Fut>(&self, operation: F, timeout_duration: Duration) -> Result<T>
+    where
+        F: FnOnce() -> Fut,
+        Fut: std::future::Future<Output = Result<T>>,
+    {
+        // Check if disk is faulty
+        if self.health.is_faulty() {
+            warn!("disk {} health is faulty, returning error", self.to_string());
+            return Err(DiskError::FaultyDisk);
+        }
+
+        // Check if disk is stale
+        self.check_disk_stale().await?;
+
+        // Record operation start
+        let now = std::time::SystemTime::now()
+            .duration_since(std::time::UNIX_EPOCH)
+            .unwrap()
+            .as_nanos() as i64;
+        self.health.last_started.store(now, Ordering::Relaxed);
+        self.health.increment_waiting();
+
+        if timeout_duration == Duration::ZERO {
+            let result = operation().await;
+            self.health.decrement_waiting();
+            if result.is_ok() {
+                self.health.log_success();
+            }
+            return result;
+        }
+        // Execute the operation with timeout
+        let result = tokio::time::timeout(timeout_duration, operation()).await;
+
+        match result {
+            Ok(operation_result) => {
+                // Log success and decrement waiting counter
+                if operation_result.is_ok() {
+                    self.health.log_success();
+                }
+                self.health.decrement_waiting();
+                operation_result
+            }
+            Err(_) => {
+                // Timeout occurred, mark disk as potentially faulty and decrement waiting counter
+                self.health.decrement_waiting();
+                warn!("disk operation timeout after {:?}", timeout_duration);
+                Err(DiskError::other(format!("disk operation timeout after {:?}", timeout_duration)))
+            }
+        }
+    }
+}
+
+#[async_trait::async_trait]
+impl DiskAPI for LocalDiskWrapper {
+    fn to_string(&self) -> String {
+        self.disk.to_string()
+    }
+
+    async fn is_online(&self) -> bool {
+        let Ok(Some(disk_id)) = self.disk.get_disk_id().await else {
+            return false;
+        };
+
+        let Some(current_disk_id) = *self.disk_id.read().await else {
+            return false;
+        };
+
+        current_disk_id == disk_id
+    }
+
+    fn is_local(&self) -> bool {
+        self.disk.is_local()
+    }
+
+    fn host_name(&self) -> String {
+        self.disk.host_name()
+    }
+
+    fn endpoint(&self) -> Endpoint {
+        self.disk.endpoint()
+    }
+
+    async fn close(&self) -> Result<()> {
+        self.stop_monitoring().await;
+        self.disk.close().await
+    }
+
+    async fn get_disk_id(&self) -> Result<Option<Uuid>> {
+        self.disk.get_disk_id().await
+    }
+
+    async fn set_disk_id(&self, id: Option<Uuid>) -> Result<()> {
+        self.set_disk_id_internal(id).await
+    }
+
+    fn path(&self) -> PathBuf {
+        self.disk.path()
+    }
+
+    fn get_disk_location(&self) -> DiskLocation {
+        self.disk.get_disk_location()
+    }
+
+    async fn disk_info(&self, opts: &DiskInfoOptions) -> Result<DiskInfo> {
+        if opts.noop && opts.metrics {
+            let mut info = DiskInfo::default();
+            // Add health metrics
+            info.metrics.total_waiting = self.health.waiting_count();
+            if self.health.is_faulty() {
+                return Err(DiskError::FaultyDisk);
+            }
+            return Ok(info);
+        }
+
+        if self.health.is_faulty() {
+            return Err(DiskError::FaultyDisk);
+        }
+
+        let result = self.disk.disk_info(opts).await?;
+
+        if let Some(current_disk_id) = *self.disk_id.read().await
+            && Some(current_disk_id) != result.id
+        {
+            return Err(DiskError::DiskNotFound);
+        };
+
+        Ok(result)
+    }
+
+    async fn make_volume(&self, volume: &str) -> Result<()> {
+        self.track_disk_health(|| async { self.disk.make_volume(volume).await }, get_max_timeout_duration())
+            .await
+    }
+
+    async fn make_volumes(&self, volumes: Vec<&str>) -> Result<()> {
+        self.track_disk_health(|| async { self.disk.make_volumes(volumes).await }, get_max_timeout_duration())
+            .await
+    }
+
+    async fn list_volumes(&self) -> Result<Vec<VolumeInfo>> {
+        self.track_disk_health(|| async { self.disk.list_volumes().await }, Duration::ZERO)
+            .await
+    }
+
+    async fn stat_volume(&self, volume: &str) -> Result<VolumeInfo> {
+        self.track_disk_health(|| async { self.disk.stat_volume(volume).await }, get_max_timeout_duration())
+            .await
+    }
+
+    async fn delete_volume(&self, volume: &str) -> Result<()> {
+        self.track_disk_health(|| async { self.disk.delete_volume(volume).await }, Duration::ZERO)
+            .await
+    }
+
+    async fn walk_dir<W: tokio::io::AsyncWrite + Unpin + Send>(&self, opts: WalkDirOptions, wr: &mut W) -> Result<()> {
+        self.track_disk_health(|| async { self.disk.walk_dir(opts, wr).await }, Duration::ZERO)
+            .await
+    }
+
+    async fn delete_version(
+        &self,
+        volume: &str,
+        path: &str,
+        fi: FileInfo,
+        force_del_marker: bool,
+        opts: DeleteOptions,
+    ) -> Result<()> {
+        self.track_disk_health(
+            || async { self.disk.delete_version(volume, path, fi, force_del_marker, opts).await },
+            get_max_timeout_duration(),
+        )
+        .await
+    }
+
+    async fn delete_versions(&self, volume: &str, versions: Vec<FileInfoVersions>, opts: DeleteOptions) -> Vec<Option<Error>> {
+        // Check if disk is faulty before proceeding
+        if self.health.is_faulty() {
+            return vec![Some(DiskError::FaultyDisk); versions.len()];
+        }
+
+        // Check if disk is stale
+        if let Err(e) = self.check_disk_stale().await {
+            return vec![Some(e); versions.len()];
+        }
+
+        // Record operation start
+        let now = std::time::SystemTime::now()
+            .duration_since(std::time::UNIX_EPOCH)
+            .unwrap()
+            .as_nanos() as i64;
+        self.health.last_started.store(now, Ordering::Relaxed);
+        self.health.increment_waiting();
+
+        // Execute the operation
+        let result = self.disk.delete_versions(volume, versions, opts).await;
+
+        self.health.decrement_waiting();
+        let has_err = result.iter().any(|e| e.is_some());
+        if !has_err {
+            // Log success and decrement waiting counter
+            self.health.log_success();
+        }
+
+        result
+    }
+
+    async fn delete_paths(&self, volume: &str, paths: &[String]) -> Result<()> {
+        self.track_disk_health(|| async { self.disk.delete_paths(volume, paths).await }, get_max_timeout_duration())
+            .await
+    }
+
+    async fn write_metadata(&self, org_volume: &str, volume: &str, path: &str, fi: FileInfo) -> Result<()> {
+        self.track_disk_health(
+            || async { self.disk.write_metadata(org_volume, volume, path, fi).await },
+            get_max_timeout_duration(),
+        )
+        .await
+    }
+
+    async fn update_metadata(&self, volume: &str, path: &str, fi: FileInfo, opts: &UpdateMetadataOpts) -> Result<()> {
+        self.track_disk_health(
+            || async { self.disk.update_metadata(volume, path, fi, opts).await },
+            get_max_timeout_duration(),
+        )
+        .await
+    }
+
+    async fn read_version(
+        &self,
+        org_volume: &str,
+        volume: &str,
+        path: &str,
+        version_id: &str,
+        opts: &ReadOptions,
+    ) -> Result<FileInfo> {
+        self.track_disk_health(
+            || async { self.disk.read_version(org_volume, volume, path, version_id, opts).await },
+            get_max_timeout_duration(),
+        )
+        .await
+    }
+
+    async fn read_xl(&self, volume: &str, path: &str, read_data: bool) -> Result<RawFileInfo> {
+        self.track_disk_health(|| async { self.disk.read_xl(volume, path, read_data).await }, get_max_timeout_duration())
+            .await
+    }
+
+    async fn rename_data(
+        &self,
+        src_volume: &str,
+        src_path: &str,
+        fi: FileInfo,
+        dst_volume: &str,
+        dst_path: &str,
+    ) -> Result<RenameDataResp> {
+        self.track_disk_health(
+            || async { self.disk.rename_data(src_volume, src_path, fi, dst_volume, dst_path).await },
+            get_max_timeout_duration(),
+        )
+        .await
+    }
+
+    async fn list_dir(&self, origvolume: &str, volume: &str, dir_path: &str, count: i32) -> Result<Vec<String>> {
+        self.track_disk_health(
+            || async { self.disk.list_dir(origvolume, volume, dir_path, count).await },
+            get_max_timeout_duration(),
+        )
+        .await
+    }
+
+    async fn read_file(&self, volume: &str, path: &str) -> Result<crate::disk::FileReader> {
+        self.track_disk_health(|| async { self.disk.read_file(volume, path).await }, get_max_timeout_duration())
+            .await
+    }
+
+    async fn read_file_stream(&self, volume: &str, path: &str, offset: usize, length: usize) -> Result<crate::disk::FileReader> {
+        self.track_disk_health(
+            || async { self.disk.read_file_stream(volume, path, offset, length).await },
+            get_max_timeout_duration(),
+        )
+        .await
+    }
+
+    async fn append_file(&self, volume: &str, path: &str) -> Result<crate::disk::FileWriter> {
+        self.track_disk_health(|| async { self.disk.append_file(volume, path).await }, Duration::ZERO)
+            .await
+    }
+
+    async fn create_file(&self, origvolume: &str, volume: &str, path: &str, file_size: i64) -> Result<crate::disk::FileWriter> {
+        self.track_disk_health(
+            || async { self.disk.create_file(origvolume, volume, path, file_size).await },
+            Duration::ZERO,
+        )
+        .await
+    }
+
+    async fn rename_file(&self, src_volume: &str, src_path: &str, dst_volume: &str, dst_path: &str) -> Result<()> {
+        self.track_disk_health(
+            || async { self.disk.rename_file(src_volume, src_path, dst_volume, dst_path).await },
+            get_max_timeout_duration(),
+        )
+        .await
+    }
+
+    async fn rename_part(&self, src_volume: &str, src_path: &str, dst_volume: &str, dst_path: &str, meta: Bytes) -> Result<()> {
+        self.track_disk_health(
+            || async { self.disk.rename_part(src_volume, src_path, dst_volume, dst_path, meta).await },
+            get_max_timeout_duration(),
+        )
+        .await
+    }
+
+    async fn delete(&self, volume: &str, path: &str, opt: DeleteOptions) -> Result<()> {
+        self.track_disk_health(|| async { self.disk.delete(volume, path, opt).await }, get_max_timeout_duration())
+            .await
+    }
+
+    async fn verify_file(&self, volume: &str, path: &str, fi: &FileInfo) -> Result<CheckPartsResp> {
+        self.track_disk_health(|| async { self.disk.verify_file(volume, path, fi).await }, Duration::ZERO)
+            .await
+    }
+
+    async fn check_parts(&self, volume: &str, path: &str, fi: &FileInfo) -> Result<CheckPartsResp> {
+        self.track_disk_health(|| async { self.disk.check_parts(volume, path, fi).await }, Duration::ZERO)
+            .await
+    }
+
+    async fn read_parts(&self, bucket: &str, paths: &[String]) -> Result<Vec<ObjectPartInfo>> {
+        self.track_disk_health(|| async { self.disk.read_parts(bucket, paths).await }, Duration::ZERO)
+            .await
+    }
+
+    async fn read_multiple(&self, req: ReadMultipleReq) -> Result<Vec<ReadMultipleResp>> {
+        self.track_disk_health(|| async { self.disk.read_multiple(req).await }, Duration::ZERO)
+            .await
+    }
+
+    async fn write_all(&self, volume: &str, path: &str, data: Bytes) -> Result<()> {
+        self.track_disk_health(|| async { self.disk.write_all(volume, path, data).await }, get_max_timeout_duration())
+            .await
+    }
+
+    async fn read_all(&self, volume: &str, path: &str) -> Result<Bytes> {
+        self.track_disk_health(|| async { self.disk.read_all(volume, path).await }, get_max_timeout_duration())
+            .await
+    }
+}
--- a/crates/ecstore/src/disk/local.rs
+++ b/crates/ecstore/src/disk/local.rs
@@ -69,7 +69,7 @@ use tokio::sync::RwLock;
 use tracing::{debug, error, info, warn};
 use uuid::Uuid;

-#[derive(Debug)]
+#[derive(Debug, Clone)]
 pub struct FormatInfo {
    pub id: Option<Uuid>,
    pub data: Bytes,
@@ -77,16 +77,6 @@ pub struct FormatInfo {
    pub last_check: Option<OffsetDateTime>,
 }

-impl FormatInfo {
-    pub fn last_check_valid(&self) -> bool {
-        let now = OffsetDateTime::now_utc();
-        self.file_info.is_some()
-            && self.id.is_some()
-            && self.last_check.is_some()
-            && (now.unix_timestamp() - self.last_check.unwrap().unix_timestamp() <= 1)
-    }
-}
-
 /// A helper enum to handle internal buffer types for writing data.
 pub enum InternalBuf<'a> {
    Ref(&'a [u8]),
@@ -185,7 +175,7 @@ impl LocalDisk {
        };
        let root_clone = root.clone();
        let update_fn: UpdateFn<DiskInfo> = Box::new(move || {
-            let disk_id = id.map_or("".to_string(), |id| id.to_string());
+            let disk_id = id;
            let root = root_clone.clone();
            Box::pin(async move {
                match get_disk_info(root.clone()).await {
@@ -200,7 +190,7 @@ impl LocalDisk {
                            minor: info.minor,
                            fs_type: info.fstype,
                            root_disk: root,
-                            id: disk_id.to_string(),
+                            id: disk_id,
                            ..Default::default()
                        };
                        // if root {
@@ -1295,7 +1285,7 @@ impl DiskAPI for LocalDisk {
    }
    #[tracing::instrument(skip(self))]
    async fn is_online(&self) -> bool {
-        self.check_format_json().await.is_ok()
+        true
    }

    #[tracing::instrument(skip(self))]
@@ -1342,24 +1332,40 @@ impl DiskAPI for LocalDisk {

    #[tracing::instrument(level = "debug", skip(self))]
    async fn get_disk_id(&self) -> Result<Option<Uuid>> {
-        let mut format_info = self.format_info.write().await;
+        let format_info = {
+            let format_info = self.format_info.read().await;
+            format_info.clone()
+        };

        let id = format_info.id;

-        if format_info.last_check_valid() {
-            return Ok(id);
+        // if format_info.last_check_valid() {
+        //     return Ok(id);
+        // }
+
+        if format_info.file_info.is_some() && id.is_some() {
+            // check last check time
+            if let Some(last_check) = format_info.last_check {
+                if last_check.unix_timestamp() + 1 < OffsetDateTime::now_utc().unix_timestamp() {
+                    return Ok(id);
+                }
+            }
        }

        let file_meta = self.check_format_json().await?;

        if let Some(file_info) = &format_info.file_info {
            if super::fs::same_file(&file_meta, file_info) {
+                let mut format_info = self.format_info.write().await;
                format_info.last_check = Some(OffsetDateTime::now_utc());
+                drop(format_info);

                return Ok(id);
            }
        }

+        debug!("get_disk_id: read format.json");
+
        let b = fs::read(&self.format_path).await.map_err(to_unformatted_disk_error)?;

        let fm = FormatV3::try_from(b.as_slice()).map_err(|e| {
@@ -1375,20 +1381,19 @@ impl DiskAPI for LocalDisk {
            return Err(DiskError::InconsistentDisk);
        }

+        let mut format_info = self.format_info.write().await;
        format_info.id = Some(disk_id);
        format_info.file_info = Some(file_meta);
        format_info.data = b.into();
        format_info.last_check = Some(OffsetDateTime::now_utc());
+        drop(format_info);

        Ok(Some(disk_id))
    }

    #[tracing::instrument(skip(self))]
-    async fn set_disk_id(&self, id: Option<Uuid>) -> Result<()> {
+    async fn set_disk_id(&self, _id: Option<Uuid>) -> Result<()> {
        // No setup is required locally
-        // TODO: add check_id_store
-        let mut format_info = self.format_info.write().await;
-        format_info.id = id;
        Ok(())
    }

@@ -2438,6 +2443,10 @@ impl DiskAPI for LocalDisk {
        info.endpoint = self.endpoint.to_string();
        info.scanning = self.scanning.load(Ordering::SeqCst) == 1;

+        if info.id.is_none() {
+            info.id = self.get_disk_id().await.unwrap_or(None);
+        }
+
        Ok(info)
    }
 }
@@ -2705,39 +2714,6 @@ mod test {
        }
    }

-    #[tokio::test]
-    async fn test_format_info_last_check_valid() {
-        let now = OffsetDateTime::now_utc();
-
-        // Valid format info
-        let valid_format_info = FormatInfo {
-            id: Some(Uuid::new_v4()),
-            data: vec![1, 2, 3].into(),
-            file_info: Some(fs::metadata("../../../..").await.unwrap()),
-            last_check: Some(now),
-        };
-        assert!(valid_format_info.last_check_valid());
-
-        // Invalid format info (missing id)
-        let invalid_format_info = FormatInfo {
-            id: None,
-            data: vec![1, 2, 3].into(),
-            file_info: Some(fs::metadata("../../../..").await.unwrap()),
-            last_check: Some(now),
-        };
-        assert!(!invalid_format_info.last_check_valid());
-
-        // Invalid format info (old timestamp)
-        let old_time = OffsetDateTime::now_utc() - time::Duration::seconds(10);
-        let old_format_info = FormatInfo {
-            id: Some(Uuid::new_v4()),
-            data: vec![1, 2, 3].into(),
-            file_info: Some(fs::metadata("../../../..").await.unwrap()),
-            last_check: Some(old_time),
-        };
-        assert!(!old_format_info.last_check_valid());
-    }
-
    #[tokio::test]
    async fn test_read_file_exists() {
        let test_file = "./test_read_exists.txt";
--- a/crates/ecstore/src/disk/mod.rs
+++ b/crates/ecstore/src/disk/mod.rs
@@ -12,6 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.

+pub mod disk_store;
 pub mod endpoint;
 pub mod error;
 pub mod error_conv;
@@ -30,6 +31,7 @@ pub const FORMAT_CONFIG_FILE: &str = "format.json";
 pub const STORAGE_FORMAT_FILE: &str = "xl.meta";
 pub const STORAGE_FORMAT_FILE_BACKUP: &str = "xl.meta.bkp";

+use crate::disk::disk_store::LocalDiskWrapper;
 use crate::rpc::RemoteDisk;
 use bytes::Bytes;
 use endpoint::Endpoint;
@@ -51,7 +53,7 @@ pub type FileWriter = Box<dyn AsyncWrite + Send + Sync + Unpin>;

 #[derive(Debug)]
 pub enum Disk {
-    Local(Box<LocalDisk>),
+    Local(Box<LocalDiskWrapper>),
    Remote(Box<RemoteDisk>),
 }

@@ -398,7 +400,7 @@ impl DiskAPI for Disk {
 pub async fn new_disk(ep: &Endpoint, opt: &DiskOption) -> Result<DiskStore> {
    if ep.is_local {
        let s = LocalDisk::new(ep, opt.cleanup).await?;
-        Ok(Arc::new(Disk::Local(Box::new(s))))
+        Ok(Arc::new(Disk::Local(Box::new(LocalDiskWrapper::new(Arc::new(s), opt.health_check)))))
    } else {
        let remote_disk = RemoteDisk::new(ep, opt).await?;
        Ok(Arc::new(Disk::Remote(Box::new(remote_disk))))
@@ -534,7 +536,7 @@ pub struct DiskInfo {
    pub scanning: bool,
    pub endpoint: String,
    pub mount_path: String,
-    pub id: String,
+    pub id: Option<Uuid>,
    pub rotational: bool,
    pub metrics: DiskMetrics,
    pub error: String,
@@ -1015,7 +1017,7 @@ mod tests {

        let endpoint = Endpoint::try_from(test_dir).unwrap();
        let local_disk = LocalDisk::new(&endpoint, false).await.unwrap();
-        let disk = Disk::Local(Box::new(local_disk));
+        let disk = Disk::Local(Box::new(LocalDiskWrapper::new(Arc::new(local_disk), false)));

        // Test basic methods
        assert!(disk.is_local());
--- a/crates/ecstore/src/erasure.rs
+++ b/crates/ecstore/src/erasure.rs
@@ -1,586 +0,0 @@
-// Copyright 2024 RustFS Team
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-use crate::bitrot::{BitrotReader, BitrotWriter};
-use crate::disk::error::{Error, Result};
-use crate::disk::error_reduce::{reduce_write_quorum_errs, OBJECT_OP_IGNORED_ERRS};
-use crate::io::Etag;
-use bytes::{Bytes, BytesMut};
-use futures::future::join_all;
-use reed_solomon_erasure::galois_8::ReedSolomon;
-use smallvec::SmallVec;
-use std::any::Any;
-use std::io::ErrorKind;
-use std::sync::{mpsc, Arc};
-use tokio::io::{AsyncRead, AsyncWrite};
-use tokio::io::{AsyncReadExt, AsyncWriteExt};
-use tokio::sync::mpsc;
-use tracing::warn;
-use tracing::{error, info};
-use uuid::Uuid;
-
-use crate::disk::error::DiskError;
-
-#[derive(Default)]
-pub struct Erasure {
-    data_shards: usize,
-    parity_shards: usize,
-    encoder: Option<ReedSolomon>,
-    pub block_size: usize,
-    _id: Uuid,
-    _buf: Vec<u8>,
-}
-
-impl Erasure {
-    pub fn new(data_shards: usize, parity_shards: usize, block_size: usize) -> Self {
-        // debug!(
-        //     "Erasure new data_shards {},parity_shards {} block_size {} ",
-        //     data_shards, parity_shards, block_size
-        // );
-        let mut encoder = None;
-        if parity_shards > 0 {
-            encoder = Some(ReedSolomon::new(data_shards, parity_shards).unwrap());
-        }
-
-        Erasure {
-            data_shards,
-            parity_shards,
-            block_size,
-            encoder,
-            _id: Uuid::new_v4(),
-            _buf: vec![0u8; block_size],
-        }
-    }
-
-    #[tracing::instrument(level = "info", skip(self, reader, writers))]
-    pub async fn encode<S>(
-        self: Arc<Self>,
-        mut reader: S,
-        writers: &mut [Option<BitrotWriter>],
-        // block_size: usize,
-        total_size: usize,
-        write_quorum: usize,
-    ) -> Result<(usize, String)>
-    where
-        S: AsyncRead + Etag + Unpin + Send + 'static,
-    {
-        let (tx, mut rx) = mpsc::channel(5);
-        let task = tokio::spawn(async move {
-            let mut buf = vec![0u8; self.block_size];
-            let mut total: usize = 0;
-            loop {
-                if total_size > 0 {
-                    let new_len = {
-                        let remain = total_size - total;
-                        if remain > self.block_size { self.block_size } else { remain }
-                    };
-
-                    if new_len == 0 && total > 0 {
-                        break;
-                    }
-
-                    buf.resize(new_len, 0u8);
-                    match reader.read_exact(&mut buf).await {
-                        Ok(res) => res,
-                        Err(e) => {
-                            if let ErrorKind::UnexpectedEof = e.kind() {
-                                break;
-                            } else {
-                                return Err(e.into());
-                            }
-                        }
-                    };
-                    total += buf.len();
-                }
-                let blocks = Arc::new(Box::pin(self.clone().encode_data(&buf)?));
-                let _ = tx.send(blocks).await;
-                if total_size == 0 {
-                    break;
-                }
-            }
-            let etag = reader.etag().await;
-            Ok((total, etag))
-        });
-
-        while let Some(blocks) = rx.recv().await {
-            let write_futures = writers.iter_mut().enumerate().map(|(i, w_op)| {
-                let i_inner = i;
-                let blocks_inner = blocks.clone();
-                async move {
-                    if let Some(w) = w_op {
-                        w.write(blocks_inner[i_inner].clone()).await.err()
-                    } else {
-                        Some(DiskError::DiskNotFound)
-                    }
-                }
-            });
-            let errs = join_all(write_futures).await;
-            let none_count = errs.iter().filter(|&x| x.is_none()).count();
-            if none_count >= write_quorum {
-                if total_size == 0 {
-                    break;
-                }
-                continue;
-            }
-
-            if let Some(err) = reduce_write_quorum_errs(&errs, OBJECT_OP_IGNORED_ERRS, write_quorum) {
-                warn!("Erasure encode errs {:?}", &errs);
-                return Err(err);
-            }
-        }
-        task.await?
-    }
-
-    pub async fn decode<W>(
-        &self,
-        writer: &mut W,
-        readers: Vec<Option<BitrotReader>>,
-        offset: usize,
-        length: usize,
-        total_length: usize,
-    ) -> (usize, Option<Error>)
-    where
-        W: AsyncWriteExt + Send + Unpin + 'static,
-    {
-        if length == 0 {
-            return (0, None);
-        }
-
-        let mut reader = ShardReader::new(readers, self, offset, total_length);
-
-        // debug!("ShardReader {:?}", &reader);
-
-        let start_block = offset / self.block_size;
-        let end_block = (offset + length) / self.block_size;
-
-        // debug!("decode block from {} to {}", start_block, end_block);
-
-        let mut bytes_written = 0;
-
-        for block_idx in start_block..=end_block {
-            let (block_offset, block_length) = if start_block == end_block {
-                (offset % self.block_size, length)
-            } else if block_idx == start_block {
-                let block_offset = offset % self.block_size;
-                (block_offset, self.block_size - block_offset)
-            } else if block_idx == end_block {
-                (0, (offset + length) % self.block_size)
-            } else {
-                (0, self.block_size)
-            };
-
-            if block_length == 0 {
-                // debug!("block_length == 0 break");
-                break;
-            }
-
-            // debug!("decode {} block_offset {},block_length {} ", block_idx, block_offset, block_length);
-
-            let mut bufs = match reader.read().await {
-                Ok(bufs) => bufs,
-                Err(err) => return (bytes_written, Some(err)),
-            };
-
-            if self.parity_shards > 0 {
-                if let Err(err) = self.decode_data(&mut bufs) {
-                    return (bytes_written, Some(err));
-                }
-            }
-
-            let written_n = match self
-                .write_data_blocks(writer, bufs, self.data_shards, block_offset, block_length)
-                .await
-            {
-                Ok(n) => n,
-                Err(err) => {
-                    error!("write_data_blocks err {:?}", &err);
-                    return (bytes_written, Some(err));
-                }
-            };
-
-            bytes_written += written_n;
-
-            // debug!("decode {} written_n {}, total_written: {} ", block_idx, written_n, bytes_written);
-        }
-
-        if bytes_written != length {
-            // debug!("bytes_written != length: {} != {} ", bytes_written, length);
-            return (bytes_written, Some(Error::other("erasure decode less data")));
-        }
-
-        (bytes_written, None)
-    }
-
-    async fn write_data_blocks<W>(
-        &self,
-        writer: &mut W,
-        bufs: Vec<Option<Vec<u8>>>,
-        data_blocks: usize,
-        offset: usize,
-        length: usize,
-    ) -> Result<usize>
-    where
-        W: AsyncWrite + Send + Unpin + 'static,
-    {
-        if bufs.len() < data_blocks {
-            return Err(Error::other("read bufs not match data_blocks"));
-        }
-
-        let data_len: usize = bufs
-            .iter()
-            .take(data_blocks)
-            .filter(|v| v.is_some())
-            .map(|v| v.as_ref().unwrap().len())
-            .sum();
-        if data_len < length {
-            return Err(Error::other(format!("write_data_blocks data_len < length {} < {}", data_len, length)));
-        }
-
-        let mut offset = offset;
-
-        // debug!("write_data_blocks offset {}, length {}", offset, length);
-
-        let mut write = length;
-        let mut total_written = 0;
-
-        for opt_buf in bufs.iter().take(data_blocks) {
-            let buf = opt_buf.as_ref().unwrap();
-
-            if offset >= buf.len() {
-                offset -= buf.len();
-                continue;
-            }
-
-            let buf = &buf[offset..];
-
-            offset = 0;
-
-            // debug!("write_data_blocks write buf len {}", buf.len());
-
-            if write < buf.len() {
-                let buf = &buf[..write];
-
-                // debug!("write_data_blocks write buf less len {}", buf.len());
-                writer.write_all(buf).await?;
-                // debug!("write_data_blocks write done len {}", buf.len());
-                total_written += buf.len();
-                break;
-            }
-
-            writer.write_all(buf).await?;
-            let n = buf.len();
-
-            // debug!("write_data_blocks write done len {}", n);
-            write -= n;
-            total_written += n;
-        }
-
-        Ok(total_written)
-    }
-
-    pub fn total_shard_count(&self) -> usize {
-        self.data_shards + self.parity_shards
-    }
-
-    #[tracing::instrument(level = "info", skip_all, fields(data_len=data.len()))]
-    pub fn encode_data(self: Arc<Self>, data: &[u8]) -> Result<Vec<Bytes>> {
-        let (shard_size, total_size) = self.need_size(data.len());
-
-        // Generate the total length required for all shards
-        let mut data_buffer = BytesMut::with_capacity(total_size);
-
-        // Copy the source data
-        data_buffer.extend_from_slice(data);
-        data_buffer.resize(total_size, 0u8);
-
-        {
-            // Perform EC encoding; the results go into data_buffer
-            let data_slices: SmallVec<[&mut [u8]; 16]> = data_buffer.chunks_exact_mut(shard_size).collect();
-
-            // Only perform EC encoding when parity shards are present
-            if self.parity_shards > 0 {
-                self.encoder.as_ref().unwrap().encode(data_slices).map_err(Error::other)?;
-            }
-        }
-
-        // Zero-copy shards: every shard references data_buffer
-        let mut data_buffer = data_buffer.freeze();
-        let mut shards = Vec::with_capacity(self.total_shard_count());
-        for _ in 0..self.total_shard_count() {
-            let shard = data_buffer.split_to(shard_size);
-            shards.push(shard);
-        }
-
-        Ok(shards)
-    }
-
-    pub fn decode_data(&self, shards: &mut [Option<Vec<u8>>]) -> Result<()> {
-        if self.parity_shards > 0 {
-            self.encoder.as_ref().unwrap().reconstruct(shards).map_err(Error::other)?;
-        }
-
-        Ok(())
-    }
-
-    // The length per shard and the total required length
-    fn need_size(&self, data_size: usize) -> (usize, usize) {
-        let shard_size = self.shard_size(data_size);
-        (shard_size, shard_size * (self.total_shard_count()))
-    }
-
-    // Compute each shard size
-    pub fn shard_size(&self, data_size: usize) -> usize {
-        data_size.div_ceil(self.data_shards)
-    }
-    // returns final erasure size from original size.
-    pub fn shard_file_size(&self, total_size: usize) -> usize {
-        if total_size == 0 {
-            return 0;
-        }
-
-        let num_shards = total_size / self.block_size;
-        let last_block_size = total_size % self.block_size;
-        let last_shard_size = last_block_size.div_ceil(self.data_shards);
-        num_shards * self.shard_size(self.block_size) + last_shard_size
-
-        // When writing, EC pads the data so the last shard length should match
-        // if last_block_size != 0 {
-        //     num_shards += 1
-        // }
-        // num_shards * self.shard_size(self.block_size)
-    }
-
-    // where erasure reading begins.
-    pub fn shard_file_offset(&self, start_offset: usize, length: usize, total_length: usize) -> usize {
-        let shard_size = self.shard_size(self.block_size);
-        let shard_file_size = self.shard_file_size(total_length);
-        let end_shard = (start_offset + length) / self.block_size;
-        let mut till_offset = end_shard * shard_size + shard_size;
-        if till_offset > shard_file_size {
-            till_offset = shard_file_size;
-        }
-
-        till_offset
-    }
-
-    pub async fn heal(
-        &self,
-        writers: &mut [Option<BitrotWriter>],
-        readers: Vec<Option<BitrotReader>>,
-        total_length: usize,
-        _prefer: &[bool],
-    ) -> Result<()> {
-        info!(
-            "Erasure heal, writers len: {}, readers len: {}, total_length: {}",
-            writers.len(),
-            readers.len(),
-            total_length
-        );
-        if writers.len() != self.parity_shards + self.data_shards {
-            return Err(Error::other("invalid argument"));
-        }
-        let mut reader = ShardReader::new(readers, self, 0, total_length);
-
-        let start_block = 0;
-        let mut end_block = total_length / self.block_size;
-        if total_length % self.block_size != 0 {
-            end_block += 1;
-        }
-
-        let mut errs = Vec::new();
-        for _ in start_block..end_block {
-            let mut bufs = reader.read().await?;
-
-            if self.parity_shards > 0 {
-                self.encoder.as_ref().unwrap().reconstruct(&mut bufs).map_err(Error::other)?;
-            }
-
-            let shards = bufs.into_iter().flatten().map(Bytes::from).collect::<Vec<_>>();
-            if shards.len() != self.parity_shards + self.data_shards {
-                return Err(Error::other("can not reconstruct data"));
-            }
-
-            for (i, w) in writers.iter_mut().enumerate() {
-                if w.is_none() {
-                    continue;
-                }
-                match w.as_mut().unwrap().write(shards[i].clone()).await {
-                    Ok(_) => {}
-                    Err(e) => {
-                        info!("write failed, err: {:?}", e);
-                        errs.push(e);
-                    }
-                }
-            }
-        }
-        if !errs.is_empty() {
-            return Err(errs[0].clone().into());
-        }
-
-        Ok(())
-    }
-}
-
-#[async_trait::async_trait]
-pub trait Writer {
-    fn as_any(&self) -> &dyn Any;
-    async fn write(&mut self, buf: Bytes) -> Result<()>;
-    async fn close(&mut self) -> Result<()> {
-        Ok(())
-    }
-}
-
-#[async_trait::async_trait]
-pub trait ReadAt {
-    async fn read_at(&mut self, offset: usize, length: usize) -> Result<(Vec<u8>, usize)>;
-}
-
-pub struct ShardReader {
-    readers: Vec<Option<BitrotReader>>, // Disk readers
-    data_block_count: usize,            // Total number of shards
-    parity_block_count: usize,
-    shard_size: usize,      // Block size per shard (read one block at a time)
-    shard_file_size: usize, // Total size of the shard file
-    offset: usize,          // Offset within the shard
-}
-
-impl ShardReader {
-    pub fn new(readers: Vec<Option<BitrotReader>>, ec: &Erasure, offset: usize, total_length: usize) -> Self {
-        Self {
-            readers,
-            data_block_count: ec.data_shards,
-            parity_block_count: ec.parity_shards,
-            shard_size: ec.shard_size(ec.block_size),
-            shard_file_size: ec.shard_file_size(total_length),
-            offset: (offset / ec.block_size) * ec.shard_size(ec.block_size),
-        }
-    }
-
-    pub async fn read(&mut self) -> Result<Vec<Option<Vec<u8>>>> {
-        // let mut disks = self.readers;
-        let reader_length = self.readers.len();
-        // Length of the block to read
-        let mut read_length = self.shard_size;
-        if self.offset + read_length > self.shard_file_size {
-            read_length = self.shard_file_size - self.offset
-        }
-
-        if read_length == 0 {
-            return Ok(vec![None; reader_length]);
-        }
-
-        // debug!("shard reader read offset {}, shard_size {}", self.offset, read_length);
-
-        let mut futures = Vec::with_capacity(reader_length);
-        let mut errors = Vec::with_capacity(reader_length);
-
-        let mut ress = Vec::with_capacity(reader_length);
-
-        for disk in self.readers.iter_mut() {
-            // if disk.is_none() {
-            //     ress.push(None);
-            //     errors.push(Some(Error::new(DiskError::DiskNotFound)));
-            //     continue;
-            // }
-
-            // let disk: &mut BitrotReader = disk.as_mut().unwrap();
-            let offset = self.offset;
-            futures.push(async move {
-                if let Some(disk) = disk {
-                    disk.read_at(offset, read_length).await
-                } else {
-                    Err(DiskError::DiskNotFound)
-                }
-            });
-        }
-
-        let results = join_all(futures).await;
-        for result in results {
-            match result {
-                Ok((res, _)) => {
-                    ress.push(Some(res));
-                    errors.push(None);
-                }
-                Err(e) => {
-                    ress.push(None);
-                    errors.push(Some(e));
-                }
-            }
-        }
-
-        if !self.can_decode(&ress) {
-            warn!("ec decode read ress {:?}", &ress);
-            warn!("ec decode read errors {:?}", &errors);
-
-            return Err(Error::other("shard reader read failed"));
-        }
-
-        self.offset += self.shard_size;
-
-        Ok(ress)
-    }
-
-    fn can_decode(&self, bufs: &[Option<Vec<u8>>]) -> bool {
-        let c = bufs.iter().filter(|v| v.is_some()).count();
-        if self.parity_block_count > 0 {
-            c >= self.data_block_count
-        } else {
-            c == self.data_block_count
-        }
-    }
-}
-
-// fn shards_to_option_shards<T: Clone>(shards: &[Vec<T>]) -> Vec<Option<Vec<T>>> {
-//     let mut result = Vec::with_capacity(shards.len());
-
-//     for v in shards.iter() {
-//         let inner: Vec<T> = v.clone();
-//         result.push(Some(inner));
-//     }
-//     result
-// }
-
-#[cfg(test)]
-mod test {
-    use super::*;
-
-    #[test]
-    fn test_erasure() {
-        let data_shards = 3;
-        let parity_shards = 2;
-        let data: &[u8] = &[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11];
-        let ec = Erasure::new(data_shards, parity_shards, 1);
-        let shards = Arc::new(ec).encode_data(data).unwrap();
-        println!("shards:{:?}", shards);
-
-        let mut s: Vec<_> = shards
-            .iter()
-            .map(|d| if d.is_empty() { None } else { Some(d.to_vec()) })
-            .collect();
-
-        // let mut s = shards_to_option_shards(&shards);
-
-        // s[0] = None;
-        s[4] = None;
-        s[3] = None;
-
-        println!("sss:{:?}", &s);
-
-        let ec = Erasure::new(data_shards, parity_shards, 1);
-        ec.decode_data(&mut s).unwrap();
-        // ec.encoder.reconstruct(&mut s).unwrap();
-
-        println!("sss:{:?}", &s);
-    }
-}
--- a/crates/ecstore/src/lib.rs
+++ b/crates/ecstore/src/lib.rs
@@ -20,7 +20,6 @@ pub mod batch_processor;
 pub mod bitrot;
 pub mod bucket;
 pub mod cache_value;
-mod chunk_stream;
 pub mod compress;
 pub mod config;
 pub mod data_usage;
--- a/crates/ecstore/src/metrics_realtime.rs
+++ b/crates/ecstore/src/metrics_realtime.rs
@@ -19,11 +19,7 @@ use crate::{
    // utils::os::get_drive_stats,
 };
 use chrono::Utc;
-use rustfs_common::{
-    globals::{GLOBAL_LOCAL_NODE_NAME, GLOBAL_RUSTFS_ADDR},
-    heal_channel::DriveState,
-    metrics::global_metrics,
-};
+use rustfs_common::{GLOBAL_LOCAL_NODE_NAME, GLOBAL_RUSTFS_ADDR, heal_channel::DriveState, metrics::global_metrics};
 use rustfs_madmin::metrics::{DiskIOStats, DiskMetric, RealtimeMetrics};
 use rustfs_utils::os::get_drive_stats;
 use serde::{Deserialize, Serialize};
--- a/crates/ecstore/src/rpc/peer_s3_client.rs
+++ b/crates/ecstore/src/rpc/peer_s3_client.rs
@@ -13,14 +13,18 @@
 // limitations under the License.

 use crate::bucket::metadata_sys;
+use crate::disk::error::DiskError;
 use crate::disk::error::{Error, Result};
 use crate::disk::error_reduce::{BUCKET_OP_IGNORED_ERRS, is_all_buckets_not_found, reduce_write_quorum_errs};
-use crate::disk::{DiskAPI, DiskStore};
+use crate::disk::{DiskAPI, DiskStore, disk_store::get_max_timeout_duration};
 use crate::global::GLOBAL_LOCAL_DISK_MAP;
 use crate::store::all_local_disk;
 use crate::store_utils::is_reserved_or_invalid_bucket;
 use crate::{
-    disk::{self, VolumeInfo},
+    disk::{
+        self, VolumeInfo,
+        disk_store::{CHECK_EVERY, CHECK_TIMEOUT_DURATION, DiskHealthTracker},
+    },
    endpoints::{EndpointServerPools, Node},
    store_api::{BucketInfo, BucketOptions, DeleteBucketOptions, MakeBucketOptions},
 };
@@ -32,10 +36,11 @@ use rustfs_protos::node_service_time_out_client;
 use rustfs_protos::proto_gen::node_service::{
    DeleteBucketRequest, GetBucketInfoRequest, HealBucketRequest, ListBucketRequest, MakeBucketRequest,
 };
-use std::{collections::HashMap, fmt::Debug, sync::Arc};
-use tokio::sync::RwLock;
+use std::{collections::HashMap, fmt::Debug, sync::Arc, time::Duration};
+use tokio::{net::TcpStream, sync::RwLock, time};
+use tokio_util::sync::CancellationToken;
 use tonic::Request;
-use tracing::info;
+use tracing::{debug, info, warn};

 type Client = Arc<Box<dyn PeerS3Client>>;

@@ -559,16 +564,160 @@ pub struct RemotePeerS3Client {
    pub node: Option<Node>,
    pub pools: Option<Vec<usize>>,
    addr: String,
+    /// Health tracker for connection monitoring
+    health: Arc<DiskHealthTracker>,
+    /// Cancellation token for monitoring tasks
+    cancel_token: CancellationToken,
 }

 impl RemotePeerS3Client {
    pub fn new(node: Option<Node>, pools: Option<Vec<usize>>) -> Self {
        let addr = node.as_ref().map(|v| v.url.to_string()).unwrap_or_default().to_string();
-        Self { node, pools, addr }
+        let client = Self {
+            node,
+            pools,
+            addr,
+            health: Arc::new(DiskHealthTracker::new()),
+            cancel_token: CancellationToken::new(),
+        };
+
+        // Start health monitoring
+        client.start_health_monitoring();
+
+        client
    }
+
    pub fn get_addr(&self) -> String {
        self.addr.clone()
    }
+
+    /// Start health monitoring for the remote peer
+    fn start_health_monitoring(&self) {
+        let health = Arc::clone(&self.health);
+        let cancel_token = self.cancel_token.clone();
+        let addr = self.addr.clone();
+
+        tokio::spawn(async move {
+            Self::monitor_remote_peer_health(addr, health, cancel_token).await;
+        });
+    }
+
+    /// Monitor remote peer health periodically
+    async fn monitor_remote_peer_health(addr: String, health: Arc<DiskHealthTracker>, cancel_token: CancellationToken) {
+        let mut interval = time::interval(CHECK_EVERY);
+
+        loop {
+            tokio::select! {
+                _ = cancel_token.cancelled() => {
+                    debug!("Health monitoring cancelled for remote peer: {}", addr);
+                    return;
+                }
+                _ = interval.tick() => {
+                    if cancel_token.is_cancelled() {
+                        return;
+                    }
+
+                    // Skip health check if peer is already marked as faulty
+                    if health.is_faulty() {
+                        continue;
+                    }
+
+                    // Perform basic connectivity check
+                    if Self::perform_connectivity_check(&addr).await.is_err() && health.swap_ok_to_faulty() {
+                        warn!("Remote peer health check failed for {}: marking as faulty", addr);
+
+                        // Start recovery monitoring
+                        let health_clone = Arc::clone(&health);
+                        let addr_clone = addr.clone();
+                        let cancel_clone = cancel_token.clone();
+
+                        tokio::spawn(async move {
+                            Self::monitor_remote_peer_recovery(addr_clone, health_clone, cancel_clone).await;
+                        });
+                    }
+                }
+            }
+        }
+    }
+
+    /// Monitor remote peer recovery and mark as healthy when recovered
+    async fn monitor_remote_peer_recovery(addr: String, health: Arc<DiskHealthTracker>, cancel_token: CancellationToken) {
+        let mut interval = time::interval(Duration::from_secs(5)); // Check every 5 seconds
+
+        loop {
+            tokio::select! {
+                _ = cancel_token.cancelled() => {
+                    return;
+                }
+                _ = interval.tick() => {
+                    if Self::perform_connectivity_check(&addr).await.is_ok() {
+                        info!("Remote peer recovered: {}", addr);
+                        health.set_ok();
+                        return;
+                    }
+                }
+            }
+        }
+    }
+
+    /// Perform basic connectivity check for remote peer
+    async fn perform_connectivity_check(addr: &str) -> Result<()> {
+        use tokio::time::timeout;
+
+        let url = url::Url::parse(addr).map_err(|e| Error::other(format!("Invalid URL: {}", e)))?;
+
+        let Some(host) = url.host_str() else {
+            return Err(Error::other("No host in URL".to_string()));
+        };
+
+        let port = url.port_or_known_default().unwrap_or(80);
+
+        // Try to establish TCP connection
+        match timeout(CHECK_TIMEOUT_DURATION, TcpStream::connect((host, port))).await {
+            Ok(Ok(_)) => Ok(()),
+            _ => Err(Error::other(format!("Cannot connect to {}:{}", host, port))),
+        }
+    }
+
+    /// Execute operation with timeout and health tracking
+    async fn execute_with_timeout<T, F, Fut>(&self, operation: F, timeout_duration: Duration) -> Result<T>
+    where
+        F: FnOnce() -> Fut,
+        Fut: std::future::Future<Output = Result<T>>,
+    {
+        // Check if peer is faulty
+        if self.health.is_faulty() {
+            return Err(DiskError::FaultyDisk);
+        }
+
+        // Record operation start
+        let now = std::time::SystemTime::now()
+            .duration_since(std::time::UNIX_EPOCH)
+            .unwrap()
+            .as_nanos() as i64;
+        self.health.last_started.store(now, std::sync::atomic::Ordering::Relaxed);
+        self.health.increment_waiting();
+
+        // Execute operation with timeout
+        let result = time::timeout(timeout_duration, operation()).await;
+
+        match result {
+            Ok(operation_result) => {
+                // Log success and decrement waiting counter
+                if operation_result.is_ok() {
+                    self.health.log_success();
+                }
+                self.health.decrement_waiting();
+                operation_result
+            }
+            Err(_) => {
+                // Timeout occurred, mark peer as potentially faulty
+                self.health.decrement_waiting();
+                warn!("Remote peer operation timeout after {:?}", timeout_duration);
+                Err(Error::other(format!("Remote peer operation timeout after {:?}", timeout_duration)))
+            }
+        }
+    }
 }

 #[async_trait]
@@ -578,115 +727,145 @@ impl PeerS3Client for RemotePeerS3Client {
    }

    async fn heal_bucket(&self, bucket: &str, opts: &HealOpts) -> Result<HealResultItem> {
-        let options: String = serde_json::to_string(opts)?;
-        let mut client = node_service_time_out_client(&self.addr)
-            .await
-            .map_err(|err| Error::other(format!("can not get client, err: {err}")))?;
-        let request = Request::new(HealBucketRequest {
-            bucket: bucket.to_string(),
-            options,
-        });
-        let response = client.heal_bucket(request).await?.into_inner();
-        if !response.success {
-            return if let Some(err) = response.error {
-                Err(err.into())
-            } else {
-                Err(Error::other(""))
-            };
-        }
+        self.execute_with_timeout(
+            || async {
+                let options: String = serde_json::to_string(opts)?;
+                let mut client = node_service_time_out_client(&self.addr)
+                    .await
+                    .map_err(|err| Error::other(format!("can not get client, err: {err}")))?;
+                let request = Request::new(HealBucketRequest {
+                    bucket: bucket.to_string(),
+                    options,
+                });
+                let response = client.heal_bucket(request).await?.into_inner();
+                if !response.success {
+                    return if let Some(err) = response.error {
+                        Err(err.into())
+                    } else {
+                        Err(Error::other(""))
+                    };
+                }

-        Ok(HealResultItem {
-            heal_item_type: HealItemType::Bucket.to_string(),
-            bucket: bucket.to_string(),
-            set_count: 0,
-            ..Default::default()
-        })
+                Ok(HealResultItem {
+                    heal_item_type: HealItemType::Bucket.to_string(),
+                    bucket: bucket.to_string(),
+                    set_count: 0,
+                    ..Default::default()
+                })
+            },
+            get_max_timeout_duration(),
+        )
+        .await
    }

    async fn list_bucket(&self, opts: &BucketOptions) -> Result<Vec<BucketInfo>> {
-        let options = serde_json::to_string(opts)?;
-        let mut client = node_service_time_out_client(&self.addr)
-            .await
-            .map_err(|err| Error::other(format!("can not get client, err: {err}")))?;
-        let request = Request::new(ListBucketRequest { options });
-        let response = client.list_bucket(request).await?.into_inner();
-        if !response.success {
-            return if let Some(err) = response.error {
-                Err(err.into())
-            } else {
-                Err(Error::other(""))
-            };
-        }
-        let bucket_infos = response
-            .bucket_infos
-            .into_iter()
-            .filter_map(|json_str| serde_json::from_str::<BucketInfo>(&json_str).ok())
-            .collect();
+        self.execute_with_timeout(
+            || async {
+                let options = serde_json::to_string(opts)?;
+                let mut client = node_service_time_out_client(&self.addr)
+                    .await
+                    .map_err(|err| Error::other(format!("can not get client, err: {err}")))?;
+                let request = Request::new(ListBucketRequest { options });
+                let response = client.list_bucket(request).await?.into_inner();
+                if !response.success {
+                    return if let Some(err) = response.error {
+                        Err(err.into())
+                    } else {
+                        Err(Error::other(""))
+                    };
+                }
+                let bucket_infos = response
+                    .bucket_infos
+                    .into_iter()
+                    .filter_map(|json_str| serde_json::from_str::<BucketInfo>(&json_str).ok())
+                    .collect();

-        Ok(bucket_infos)
+                Ok(bucket_infos)
+            },
+            get_max_timeout_duration(),
+        )
+        .await
    }
    async fn make_bucket(&self, bucket: &str, opts: &MakeBucketOptions) -> Result<()> {
-        let options = serde_json::to_string(opts)?;
-        let mut client = node_service_time_out_client(&self.addr)
-            .await
-            .map_err(|err| Error::other(format!("can not get client, err: {err}")))?;
-        let request = Request::new(MakeBucketRequest {
-            name: bucket.to_string(),
-            options,
-        });
-        let response = client.make_bucket(request).await?.into_inner();
+        self.execute_with_timeout(
+            || async {
+                let options = serde_json::to_string(opts)?;
+                let mut client = node_service_time_out_client(&self.addr)
+                    .await
+                    .map_err(|err| Error::other(format!("can not get client, err: {err}")))?;
+                let request = Request::new(MakeBucketRequest {
+                    name: bucket.to_string(),
+                    options,
+                });
+                let response = client.make_bucket(request).await?.into_inner();

-        // TODO: deal with error
-        if !response.success {
-            return if let Some(err) = response.error {
-                Err(err.into())
-            } else {
-                Err(Error::other(""))
-            };
-        }
+                // TODO: deal with error
+                if !response.success {
+                    return if let Some(err) = response.error {
+                        Err(err.into())
+                    } else {
+                        Err(Error::other(""))
+                    };
+                }

-        Ok(())
+                Ok(())
+            },
+            get_max_timeout_duration(),
+        )
+        .await
    }
    async fn get_bucket_info(&self, bucket: &str, opts: &BucketOptions) -> Result<BucketInfo> {
-        let options = serde_json::to_string(opts)?;
-        let mut client = node_service_time_out_client(&self.addr)
-            .await
-            .map_err(|err| Error::other(format!("can not get client, err: {err}")))?;
-        let request = Request::new(GetBucketInfoRequest {
-            bucket: bucket.to_string(),
-            options,
-        });
-        let response = client.get_bucket_info(request).await?.into_inner();
-        if !response.success {
-            return if let Some(err) = response.error {
-                Err(err.into())
-            } else {
-                Err(Error::other(""))
-            };
-        }
-        let bucket_info = serde_json::from_str::<BucketInfo>(&response.bucket_info)?;
+        self.execute_with_timeout(
+            || async {
+                let options = serde_json::to_string(opts)?;
+                let mut client = node_service_time_out_client(&self.addr)
+                    .await
+                    .map_err(|err| Error::other(format!("can not get client, err: {err}")))?;
+                let request = Request::new(GetBucketInfoRequest {
+                    bucket: bucket.to_string(),
+                    options,
+                });
+                let response = client.get_bucket_info(request).await?.into_inner();
+                if !response.success {
+                    return if let Some(err) = response.error {
+                        Err(err.into())
+                    } else {
+                        Err(Error::other(""))
+                    };
+                }
+                let bucket_info = serde_json::from_str::<BucketInfo>(&response.bucket_info)?;

-        Ok(bucket_info)
+                Ok(bucket_info)
+            },
+            get_max_timeout_duration(),
+        )
+        .await
    }

    async fn delete_bucket(&self, bucket: &str, _opts: &DeleteBucketOptions) -> Result<()> {
-        let mut client = node_service_time_out_client(&self.addr)
-            .await
-            .map_err(|err| Error::other(format!("can not get client, err: {err}")))?;
+        self.execute_with_timeout(
+            || async {
+                let mut client = node_service_time_out_client(&self.addr)
+                    .await
+                    .map_err(|err| Error::other(format!("can not get client, err: {err}")))?;

-        let request = Request::new(DeleteBucketRequest {
-            bucket: bucket.to_string(),
-        });
-        let response = client.delete_bucket(request).await?.into_inner();
-        if !response.success {
-            return if let Some(err) = response.error {
-                Err(err.into())
-            } else {
-                Err(Error::other(""))
-            };
-        }
+                let request = Request::new(DeleteBucketRequest {
+                    bucket: bucket.to_string(),
+                });
+                let response = client.delete_bucket(request).await?.into_inner();
+                if !response.success {
+                    return if let Some(err) = response.error {
+                        Err(err.into())
+                    } else {
+                        Err(Error::other(""))
+                    };
+                }

-        Ok(())
+                Ok(())
+            },
+            get_max_timeout_duration(),
+        )
+        .await
    }
 }

--- a/crates/ecstore/src/rpc/remote_disk.rs
+++ b/crates/ecstore/src/rpc/remote_disk.rs
--- a/crates/ecstore/src/set_disk.rs
+++ b/crates/ecstore/src/set_disk.rs
@@ -174,56 +174,56 @@ impl SetDisks {
        })
    }

-    async fn cached_disk_health(&self, index: usize) -> Option<bool> {
-        let cache = self.disk_health_cache.read().await;
-        cache
-            .get(index)
-            .and_then(|entry| entry.as_ref().and_then(|state| state.cached_value()))
-    }
+    // async fn cached_disk_health(&self, index: usize) -> Option<bool> {
+    //     let cache = self.disk_health_cache.read().await;
+    //     cache
+    //         .get(index)
+    //         .and_then(|entry| entry.as_ref().and_then(|state| state.cached_value()))
+    // }

-    async fn update_disk_health(&self, index: usize, online: bool) {
-        let mut cache = self.disk_health_cache.write().await;
-        if cache.len() <= index {
-            cache.resize(index + 1, None);
-        }
-        cache[index] = Some(DiskHealthEntry {
-            last_check: Instant::now(),
-            online,
-        });
-    }
+    // async fn update_disk_health(&self, index: usize, online: bool) {
+    //     let mut cache = self.disk_health_cache.write().await;
+    //     if cache.len() <= index {
+    //         cache.resize(index + 1, None);
+    //     }
+    //     cache[index] = Some(DiskHealthEntry {
+    //         last_check: Instant::now(),
+    //         online,
+    //     });
+    // }

-    async fn is_disk_online_cached(&self, index: usize, disk: &DiskStore) -> bool {
-        if let Some(online) = self.cached_disk_health(index).await {
-            return online;
-        }
+    // async fn is_disk_online_cached(&self, index: usize, disk: &DiskStore) -> bool {
+    //     if let Some(online) = self.cached_disk_health(index).await {
+    //         return online;
+    //     }

-        let disk_clone = disk.clone();
-        let online = timeout(DISK_ONLINE_TIMEOUT, async move { disk_clone.is_online().await })
-            .await
-            .unwrap_or(false);
-        self.update_disk_health(index, online).await;
-        online
-    }
+    //     let disk_clone = disk.clone();
+    //     let online = timeout(DISK_ONLINE_TIMEOUT, async move { disk_clone.is_online().await })
+    //         .await
+    //         .unwrap_or(false);
+    //     self.update_disk_health(index, online).await;
+    //     online
+    // }

-    async fn filter_online_disks(&self, disks: Vec<Option<DiskStore>>) -> (Vec<Option<DiskStore>>, usize) {
-        let mut filtered = Vec::with_capacity(disks.len());
-        let mut online_count = 0;
+    // async fn filter_online_disks(&self, disks: Vec<Option<DiskStore>>) -> (Vec<Option<DiskStore>>, usize) {
+    //     let mut filtered = Vec::with_capacity(disks.len());
+    //     let mut online_count = 0;

-        for (idx, disk) in disks.into_iter().enumerate() {
-            if let Some(disk_store) = disk {
-                if self.is_disk_online_cached(idx, &disk_store).await {
-                    filtered.push(Some(disk_store));
-                    online_count += 1;
-                } else {
-                    filtered.push(None);
-                }
-            } else {
-                filtered.push(None);
-            }
-        }
+    //     for (idx, disk) in disks.into_iter().enumerate() {
+    //         if let Some(disk_store) = disk {
+    //             if self.is_disk_online_cached(idx, &disk_store).await {
+    //                 filtered.push(Some(disk_store));
+    //                 online_count += 1;
+    //             } else {
+    //                 filtered.push(None);
+    //             }
+    //         } else {
+    //             filtered.push(None);
+    //         }
+    //     }

-        (filtered, online_count)
-    }
+    //     (filtered, online_count)
+    // }
    fn format_lock_error(&self, bucket: &str, object: &str, mode: &str, err: &LockResult) -> String {
        match err {
            LockResult::Timeout => {
@@ -259,9 +259,28 @@ impl SetDisks {
    }

    async fn get_online_disks(&self) -> Vec<Option<DiskStore>> {
-        let disks = self.get_disks_internal().await;
-        let (filtered, _) = self.filter_online_disks(disks).await;
-        filtered.into_iter().filter(|disk| disk.is_some()).collect()
+        let mut disks = self.get_disks_internal().await;
+
+        // TODO: diskinfo filter online
+
+        let mut new_disk = Vec::with_capacity(disks.len());
+
+        for disk in disks.iter() {
+            if let Some(d) = disk {
+                if d.is_online().await {
+                    new_disk.push(disk.clone());
+                }
+            }
+        }
+
+        let mut rng = rand::rng();
+
+        disks.shuffle(&mut rng);
+
+        new_disk
+        // let disks = self.get_disks_internal().await;
+        // let (filtered, _) = self.filter_online_disks(disks).await;
+        // filtered.into_iter().filter(|disk| disk.is_some()).collect()
    }
    async fn get_online_local_disks(&self) -> Vec<Option<DiskStore>> {
        let mut disks = self.get_online_disks().await;
@@ -1467,7 +1486,9 @@ impl SetDisks {
            let object = object.clone();
            let version_id = version_id.clone();
            tokio::spawn(async move {
-                if let Some(disk) = disk {
+                if let Some(disk) = disk
+                    && disk.is_online().await
+                {
                    if version_id.is_empty() {
                        match disk.read_xl(&bucket, &object, read_data).await {
                            Ok(info) => {
@@ -1799,14 +1820,14 @@ impl SetDisks {
    }

    pub async fn renew_disk(&self, ep: &Endpoint) {
-        debug!("renew_disk start {:?}", ep);
+        debug!("renew_disk: start {:?}", ep);

        let (new_disk, fm) = match Self::connect_endpoint(ep).await {
            Ok(res) => res,
            Err(e) => {
-                warn!("connect_endpoint err {:?}", &e);
+                warn!("renew_disk: connect_endpoint err {:?}", &e);
                if ep.is_local && e == DiskError::UnformattedDisk {
-                    info!("unformatteddisk will trigger heal_disk, {:?}", ep);
+                    info!("renew_disk unformatteddisk will trigger heal_disk, {:?}", ep);
                    let set_disk_id = format!("pool_{}_set_{}", ep.pool_idx, ep.set_idx);
                    let _ = send_heal_disk(set_disk_id, Some(HealChannelPriority::Normal)).await;
                }
@@ -1817,7 +1838,7 @@ impl SetDisks {
        let (set_idx, disk_idx) = match self.find_disk_index(&fm) {
            Ok(res) => res,
            Err(e) => {
-                warn!("find_disk_index err {:?}", e);
+                warn!("renew_disk: find_disk_index err {:?}", e);
                return;
            }
        };
@@ -1837,7 +1858,7 @@ impl SetDisks {
            }
        }

-        debug!("renew_disk update {:?}", fm.erasure.this);
+        debug!("renew_disk: update {:?}", fm.erasure.this);

        let mut disk_lock = self.disks.write().await;
        disk_lock[disk_idx] = Some(new_disk);
@@ -3051,7 +3072,7 @@ impl SetDisks {
                                for (index, disk) in latest_disks.iter().enumerate() {
                                    if let Some(outdated_disk) = &out_dated_disks[index] {
                                        info!(disk_index = index, "Creating writer for outdated disk");
-                                        let writer = create_bitrot_writer(
+                                        let writer = match create_bitrot_writer(
                                            is_inline_buffer,
                                            Some(outdated_disk),
                                            RUSTFS_META_TMP_BUCKET,
@@ -3060,7 +3081,19 @@ impl SetDisks {
                                            erasure.shard_size(),
                                            HashAlgorithm::HighwayHash256,
                                        )
-                                        .await?;
+                                        .await
+                                        {
+                                            Ok(writer) => writer,
+                                            Err(err) => {
+                                                warn!(
+                                                    "create_bitrot_writer  disk {}, err {:?}, skipping operation",
+                                                    outdated_disk.to_string(),
+                                                    err
+                                                );
+                                                writers.push(None);
+                                                continue;
+                                            }
+                                        };
                                        writers.push(Some(writer));
                                    } else {
                                        info!(disk_index = index, "Skipping writer (disk not outdated)");
@@ -3790,8 +3823,8 @@ impl ObjectIO for SetDisks {

    #[tracing::instrument(level = "debug", skip(self, data,))]
    async fn put_object(&self, bucket: &str, object: &str, data: &mut PutObjReader, opts: &ObjectOptions) -> Result<ObjectInfo> {
-        let disks_snapshot = self.get_disks_internal().await;
-        let (disks, filtered_online) = self.filter_online_disks(disks_snapshot).await;
+        let disks = self.get_disks_internal().await;
+        // let (disks, filtered_online) = self.filter_online_disks(disks_snapshot).await;

        // Acquire per-object exclusive lock via RAII guard. It auto-releases asynchronously on drop.
        let _object_lock_guard = if !opts.no_lock {
@@ -3832,13 +3865,13 @@ impl ObjectIO for SetDisks {
            write_quorum += 1
        }

-        if filtered_online < write_quorum {
-            warn!(
-                "online disk snapshot {} below write quorum {} for {}/{}; returning erasure write quorum error",
-                filtered_online, write_quorum, bucket, object
-            );
-            return Err(to_object_err(Error::ErasureWriteQuorum, vec![bucket, object]));
-        }
+        // if filtered_online < write_quorum {
+        //     warn!(
+        //         "online disk snapshot {} below write quorum {} for {}/{}; returning erasure write quorum error",
+        //         filtered_online, write_quorum, bucket, object
+        //     );
+        //     return Err(to_object_err(Error::ErasureWriteQuorum, vec![bucket, object]));
+        // }

        let mut fi = FileInfo::new([bucket, object].join("/").as_str(), data_drives, parity_drives);

@@ -3877,8 +3910,10 @@ impl ObjectIO for SetDisks {
        let mut writers = Vec::with_capacity(shuffle_disks.len());
        let mut errors = Vec::with_capacity(shuffle_disks.len());
        for disk_op in shuffle_disks.iter() {
-            if let Some(disk) = disk_op {
-                let writer = create_bitrot_writer(
+            if let Some(disk) = disk_op
+                && disk.is_online().await
+            {
+                let writer = match create_bitrot_writer(
                    is_inline_buffer,
                    Some(disk),
                    RUSTFS_META_TMP_BUCKET,
@@ -3887,29 +3922,16 @@ impl ObjectIO for SetDisks {
                    erasure.shard_size(),
                    HashAlgorithm::HighwayHash256,
                )
-                .await?;
-
-                // let writer = if is_inline_buffer {
-                //     BitrotWriter::new(
-                //         Writer::from_cursor(Cursor::new(Vec::new())),
-                //         erasure.shard_size(),
-                //         HashAlgorithm::HighwayHash256,
-                //     )
-                // } else {
-                //     let f = match disk
-                //         .create_file("", RUSTFS_META_TMP_BUCKET, &tmp_object, erasure.shard_file_size(data.content_length))
-                //         .await
-                //     {
-                //         Ok(f) => f,
-                //         Err(e) => {
-                //             errors.push(Some(e));
-                //             writers.push(None);
-                //             continue;
-                //         }
-                //     };
-
-                //     BitrotWriter::new(Writer::from_tokio_writer(f), erasure.shard_size(), HashAlgorithm::HighwayHash256)
-                // };
+                .await
+                {
+                    Ok(writer) => writer,
+                    Err(err) => {
+                        warn!("create_bitrot_writer  disk {}, err {:?}, skipping operation", disk.to_string(), err);
+                        errors.push(Some(err));
+                        writers.push(None);
+                        continue;
+                    }
+                };

                writers.push(Some(writer));
                errors.push(None);
@@ -4072,7 +4094,7 @@ impl StorageAPI for SetDisks {
    async fn local_storage_info(&self) -> rustfs_madmin::StorageInfo {
        let disks = self.get_disks_internal().await;

-        let mut local_disks: Vec<Option<Arc<disk::Disk>>> = Vec::new();
+        let mut local_disks: Vec<Option<DiskStore>> = Vec::new();
        let mut local_endpoints = Vec::new();

        for (i, ep) in self.set_endpoints.iter().enumerate() {
@@ -4908,9 +4930,7 @@ impl StorageAPI for SetDisks {

        for disk in disks.iter() {
            if let Some(disk) = disk {
-                if disk.is_online().await {
-                    continue;
-                }
+                continue;
            }
            let _ = self.add_partial(bucket, object, opts.version_id.as_ref().expect("err")).await;
            break;
@@ -5129,16 +5149,16 @@ impl StorageAPI for SetDisks {
            return Err(Error::other(format!("checksum mismatch: {checksum}")));
        }

-        let disks_snapshot = self.get_disks_internal().await;
-        let (disks, filtered_online) = self.filter_online_disks(disks_snapshot).await;
+        let disks = self.get_disks_internal().await;
+        // let (disks, filtered_online) = self.filter_online_disks(disks_snapshot).await;

-        if filtered_online < write_quorum {
-            warn!(
-                "online disk snapshot {} below write quorum {} for multipart {}/{}; returning erasure write quorum error",
-                filtered_online, write_quorum, bucket, object
-            );
-            return Err(to_object_err(Error::ErasureWriteQuorum, vec![bucket, object]));
-        }
+        // if filtered_online < write_quorum {
+        //     warn!(
+        //         "online disk snapshot {} below write quorum {} for multipart {}/{}; returning erasure write quorum error",
+        //         filtered_online, write_quorum, bucket, object
+        //     );
+        //     return Err(to_object_err(Error::ErasureWriteQuorum, vec![bucket, object]));
+        // }

        let shuffle_disks = Self::shuffle_disks(&disks, &fi.erasure.distribution);

@@ -5152,7 +5172,7 @@ impl StorageAPI for SetDisks {
        let mut errors = Vec::with_capacity(shuffle_disks.len());
        for disk_op in shuffle_disks.iter() {
            if let Some(disk) = disk_op {
-                let writer = create_bitrot_writer(
+                let writer = match create_bitrot_writer(
                    false,
                    Some(disk),
                    RUSTFS_META_TMP_BUCKET,
@@ -5161,23 +5181,16 @@ impl StorageAPI for SetDisks {
                    erasure.shard_size(),
                    HashAlgorithm::HighwayHash256,
                )
-                .await?;
-
-                // let writer = {
-                //     let f = match disk
-                //         .create_file("", RUSTFS_META_TMP_BUCKET, &tmp_part_path, erasure.shard_file_size(data.content_length))
-                //         .await
-                //     {
-                //         Ok(f) => f,
-                //         Err(e) => {
-                //             errors.push(Some(e));
-                //             writers.push(None);
-                //             continue;
-                //         }
-                //     };
-
-                //     BitrotWriter::new(Writer::from_tokio_writer(f), erasure.shard_size(), HashAlgorithm::HighwayHash256)
-                // };
+                .await
+                {
+                    Ok(writer) => writer,
+                    Err(err) => {
+                        warn!("create_bitrot_writer  disk {}, err {:?}, skipping operation", disk.to_string(), err);
+                        errors.push(Some(err));
+                        writers.push(None);
+                        continue;
+                    }
+                };

                writers.push(Some(writer));
                errors.push(None);
@@ -6769,7 +6782,7 @@ async fn get_disks_info(disks: &[Option<DiskStore>], eps: &[Endpoint]) -> Vec<ru
                    healing: res.healing,
                    scanning: res.scanning,

-                    uuid: res.id.clone(),
+                    uuid: res.id.map_or("".to_string(), |id| id.to_string()),
                    major: res.major as u32,
                    minor: res.minor as u32,
                    model: None,
--- a/crates/ecstore/src/sets.rs
+++ b/crates/ecstore/src/sets.rs
@@ -40,7 +40,7 @@ use futures::future::join_all;
 use http::HeaderMap;
 use rustfs_common::heal_channel::HealOpts;
 use rustfs_common::{
-    globals::GLOBAL_LOCAL_NODE_NAME,
+    GLOBAL_LOCAL_NODE_NAME,
    heal_channel::{DriveState, HealItemType},
 };
 use rustfs_filemeta::FileInfo;
@@ -255,7 +255,7 @@ impl Sets {
        self.connect_disks().await;

        // TODO: config interval
-        let mut interval = tokio::time::interval(Duration::from_secs(15 * 3));
+        let mut interval = tokio::time::interval(Duration::from_secs(15));
        loop {
            tokio::select! {
               _= interval.tick()=>{
--- a/crates/ecstore/src/store.rs
+++ b/crates/ecstore/src/store.rs
@@ -55,8 +55,8 @@ use futures::future::join_all;
 use http::HeaderMap;
 use lazy_static::lazy_static;
 use rand::Rng as _;
-use rustfs_common::globals::{GLOBAL_LOCAL_NODE_NAME, GLOBAL_RUSTFS_HOST, GLOBAL_RUSTFS_PORT};
 use rustfs_common::heal_channel::{HealItemType, HealOpts};
+use rustfs_common::{GLOBAL_LOCAL_NODE_NAME, GLOBAL_RUSTFS_HOST, GLOBAL_RUSTFS_PORT};
 use rustfs_filemeta::FileInfo;
 use rustfs_madmin::heal_commands::HealResultItem;
 use rustfs_utils::path::{SLASH_SEPARATOR, decode_dir_object, encode_dir_object, path_join_buf};
--- a/crates/ecstore/src/store_init.rs
+++ b/crates/ecstore/src/store_init.rs
@@ -265,7 +265,10 @@ pub async fn load_format_erasure(disk: &DiskStore, heal: bool) -> disk::error::R
        .map_err(|e| match e {
            DiskError::FileNotFound => DiskError::UnformattedDisk,
            DiskError::DiskNotFound => DiskError::UnformattedDisk,
-            _ => e,
+            _ => {
+                warn!("load_format_erasure err: {:?} {:?}", disk.to_string(), e);
+                e
+            }
        })?;

    let mut fm = FormatV3::try_from(data.as_ref())?;
@@ -312,17 +315,18 @@ async fn save_format_file_all(disks: &[Option<DiskStore>], formats: &[Option<For
 }

 pub async fn save_format_file(disk: &Option<DiskStore>, format: &Option<FormatV3>) -> disk::error::Result<()> {
-    if disk.is_none() {
+    let Some(disk) = disk else {
        return Err(DiskError::DiskNotFound);
-    }
+    };

-    let format = format.as_ref().unwrap();
+    let Some(format) = format else {
+        return Err(DiskError::other("format is none"));
+    };

    let json_data = format.to_json()?;

    let tmpfile = Uuid::new_v4().to_string();

-    let disk = disk.as_ref().unwrap();
    disk.write_all(RUSTFS_META_BUCKET, tmpfile.as_str(), json_data.into_bytes().into())
        .await?;

--- a/crates/ecstore/src/tier/warm_backend_azure2.rs
+++ b/crates/ecstore/src/tier/warm_backend_azure2.rs
@@ -1,231 +0,0 @@
-// Copyright 2024 RustFS Team
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-#![allow(unused_imports)]
-#![allow(unused_variables)]
-#![allow(unused_mut)]
-#![allow(unused_assignments)]
-#![allow(unused_must_use)]
-#![allow(clippy::all)]
-
-use std::collections::HashMap;
-use std::sync::Arc;
-
-use azure_core::http::{Body, ClientOptions, RequestContent};
-use azure_storage::StorageCredentials;
-use azure_storage_blobs::prelude::*;
-
-use crate::client::{
-    admin_handler_utils::AdminError,
-    api_put_object::PutObjectOptions,
-    transition_api::{Options, ReadCloser, ReaderImpl},
-};
-use crate::tier::{
-    tier_config::TierAzure,
-    warm_backend::{WarmBackend, WarmBackendGetOpts},
-};
-use tracing::warn;
-
-const MAX_MULTIPART_PUT_OBJECT_SIZE: i64 = 1024 * 1024 * 1024 * 1024 * 5;
-const MAX_PARTS_COUNT: i64 = 10000;
-const _MAX_PART_SIZE: i64 = 1024 * 1024 * 1024 * 5;
-const MIN_PART_SIZE: i64 = 1024 * 1024 * 128;
-
-pub struct WarmBackendAzure {
-    pub client: Arc<BlobServiceClient>,
-    pub bucket: String,
-    pub prefix: String,
-    pub storage_class: String,
-}
-
-impl WarmBackendAzure {
-    pub async fn new(conf: &TierAzure, tier: &str) -> Result<Self, std::io::Error> {
-        if conf.access_key == "" || conf.secret_key == "" {
-            return Err(std::io::Error::other("both access and secret keys are required"));
-        }
-
-        if conf.bucket == "" {
-            return Err(std::io::Error::other("no bucket name was provided"));
-        }
-
-        let creds = StorageCredentials::access_key(conf.access_key.clone(), conf.secret_key.clone());
-        let client = ClientBuilder::new(conf.access_key.clone(), creds)
-            //.endpoint(conf.endpoint)
-            .blob_service_client();
-        let client = Arc::new(client);
-        Ok(Self {
-            client,
-            bucket: conf.bucket.clone(),
-            prefix: conf.prefix.strip_suffix("/").unwrap_or(&conf.prefix).to_owned(),
-            storage_class: "".to_string(),
-        })
-    }
-
-    /*pub fn tier(&self) -> *blob.AccessTier {
-        if self.storage_class == "" {
-            return None;
-        }
-        for t in blob.PossibleAccessTierValues() {
-            if strings.EqualFold(self.storage_class, t) {
-                return &t
-            }
-        }
-        None
-    }*/
-
-    pub fn get_dest(&self, object: &str) -> String {
-        let mut dest_obj = object.to_string();
-        if self.prefix != "" {
-            dest_obj = format!("{}/{}", &self.prefix, object);
-        }
-        return dest_obj;
-    }
-}
-
-#[async_trait::async_trait]
-impl WarmBackend for WarmBackendAzure {
-    async fn put_with_meta(
-        &self,
-        object: &str,
-        r: ReaderImpl,
-        length: i64,
-        meta: HashMap<String, String>,
-    ) -> Result<String, std::io::Error> {
-        let part_size = length;
-        let client = self.client.clone();
-        let container_client = client.container_client(self.bucket.clone());
-        let blob_client = container_client.blob_client(self.get_dest(object));
-        /*let res = blob_client
-            .upload(
-                RequestContent::from(match r {
-                    ReaderImpl::Body(content_body) => content_body.to_vec(),
-                    ReaderImpl::ObjectBody(mut content_body) => content_body.read_all().await?,
-                }),
-                false,
-                length as u64,
-                None,
-            )
-            .await
-        else {
-            return Err(std::io::Error::other("upload error"));
-        };*/
-
-        let Ok(res) = blob_client
-            .put_block_blob(match r {
-                ReaderImpl::Body(content_body) => content_body.to_vec(),
-                ReaderImpl::ObjectBody(mut content_body) => content_body.read_all().await?,
-            })
-            .content_type("text/plain")
-            .into_future()
-            .await
-        else {
-            return Err(std::io::Error::other("put_block_blob error"));
-        };
-
-        //self.ToObjectError(err, object)
-        Ok(res.request_id.to_string())
-    }
-
-    async fn put(&self, object: &str, r: ReaderImpl, length: i64) -> Result<String, std::io::Error> {
-        self.put_with_meta(object, r, length, HashMap::new()).await
-    }
-
-    async fn get(&self, object: &str, rv: &str, opts: WarmBackendGetOpts) -> Result<ReadCloser, std::io::Error> {
-        let client = self.client.clone();
-        let container_client = client.container_client(self.bucket.clone());
-        let blob_client = container_client.blob_client(self.get_dest(object));
-        blob_client.get();
-        todo!();
-    }
-
-    async fn remove(&self, object: &str, rv: &str) -> Result<(), std::io::Error> {
-        let client = self.client.clone();
-        let container_client = client.container_client(self.bucket.clone());
-        let blob_client = container_client.blob_client(self.get_dest(object));
-        blob_client.delete();
-        todo!();
-    }
-
-    async fn in_use(&self) -> Result<bool, std::io::Error> {
-        /*let result = self.client
-            .list_objects_v2(&self.bucket, &self.prefix, "", "", SLASH_SEPARATOR, 1)
-            .await?;
-
-        Ok(result.common_prefixes.len() > 0 || result.contents.len() > 0)*/
-        Ok(false)
-    }
-}
-
-/*fn azure_to_object_error(err: Error, params: Vec<String>) -> Option<error> {
-    if err == nil {
-      return nil
-    }
-
-    bucket := ""
-    object := ""
-    if len(params) >= 1 {
-      bucket = params[0]
-    }
-    if len(params) == 2 {
-      object = params[1]
-    }
-
-    azureErr, ok := err.(*azcore.ResponseError)
-    if !ok {
-      // We don't interpret non Azure errors. As azure errors will
-      // have StatusCode to help to convert to object errors.
-      return err
-    }
-
-    serviceCode := azureErr.ErrorCode
-    statusCode := azureErr.StatusCode
-
-    azureCodesToObjectError(err, serviceCode, statusCode, bucket, object)
-}*/
-
-/*fn azure_codes_to_object_error(err: Error, service_code: String, status_code: i32, bucket: String, object: String) -> Option<Error> {
-  switch serviceCode {
-  case "ContainerNotFound", "ContainerBeingDeleted":
-    err = BucketNotFound{Bucket: bucket}
-  case "ContainerAlreadyExists":
-    err = BucketExists{Bucket: bucket}
-  case "InvalidResourceName":
-    err = BucketNameInvalid{Bucket: bucket}
-  case "RequestBodyTooLarge":
-    err = PartTooBig{}
-  case "InvalidMetadata":
-    err = UnsupportedMetadata{}
-  case "BlobAccessTierNotSupportedForAccountType":
-    err = NotImplemented{}
-  case "OutOfRangeInput":
-    err = ObjectNameInvalid{
-      Bucket: bucket,
-      Object: object,
-    }
-  default:
-    switch statusCode {
-    case http.StatusNotFound:
-      if object != "" {
-        err = ObjectNotFound{
-          Bucket: bucket,
-          Object: object,
-        }
-      } else {
-        err = BucketNotFound{Bucket: bucket}
-      }
-    case http.StatusBadRequest:
-      err = BucketNameInvalid{Bucket: bucket}
-    }
-  }
-  return err
-}*/
--- a/crates/filemeta/src/headers.rs
+++ b/crates/filemeta/src/headers.rs
@@ -1,52 +0,0 @@
-// Copyright 2024 RustFS Team
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-pub const AMZ_META_UNENCRYPTED_CONTENT_LENGTH: &str = "X-Amz-Meta-X-Amz-Unencrypted-Content-Length";
-pub const AMZ_META_UNENCRYPTED_CONTENT_MD5: &str = "X-Amz-Meta-X-Amz-Unencrypted-Content-Md5";
-
-pub const AMZ_STORAGE_CLASS: &str = "x-amz-storage-class";
-
-pub const RESERVED_METADATA_PREFIX: &str = "X-RustFS-Internal-";
-pub const RESERVED_METADATA_PREFIX_LOWER: &str = "x-rustfs-internal-";
-
-pub const RUSTFS_HEALING: &str = "X-Rustfs-Internal-healing";
-// pub const RUSTFS_DATA_MOVE: &str = "X-Rustfs-Internal-data-mov";
-
-// pub const X_RUSTFS_INLINE_DATA: &str = "x-rustfs-inline-data";
-
-pub const VERSION_PURGE_STATUS_KEY: &str = "X-Rustfs-Internal-purgestatus";
-
-pub const X_RUSTFS_HEALING: &str = "X-Rustfs-Internal-healing";
-pub const X_RUSTFS_DATA_MOV: &str = "X-Rustfs-Internal-data-mov";
-
-pub const AMZ_OBJECT_TAGGING: &str = "X-Amz-Tagging";
-pub const AMZ_BUCKET_REPLICATION_STATUS: &str = "X-Amz-Replication-Status";
-pub const AMZ_DECODED_CONTENT_LENGTH: &str = "X-Amz-Decoded-Content-Length";
-
-pub const RUSTFS_DATA_MOVE: &str = "X-Rustfs-Internal-data-mov";
-
-// Server-side encryption headers
-pub const AMZ_SERVER_SIDE_ENCRYPTION: &str = "x-amz-server-side-encryption";
-pub const AMZ_SERVER_SIDE_ENCRYPTION_AWS_KMS_KEY_ID: &str = "x-amz-server-side-encryption-aws-kms-key-id";
-pub const AMZ_SERVER_SIDE_ENCRYPTION_CONTEXT: &str = "x-amz-server-side-encryption-context";
-pub const AMZ_SERVER_SIDE_ENCRYPTION_CUSTOMER_ALGORITHM: &str = "x-amz-server-side-encryption-customer-algorithm";
-pub const AMZ_SERVER_SIDE_ENCRYPTION_CUSTOMER_KEY: &str = "x-amz-server-side-encryption-customer-key";
-pub const AMZ_SERVER_SIDE_ENCRYPTION_CUSTOMER_KEY_MD5: &str = "x-amz-server-side-encryption-customer-key-md5";
-
-// SSE-C copy source headers
-pub const AMZ_COPY_SOURCE_SERVER_SIDE_ENCRYPTION_CUSTOMER_ALGORITHM: &str =
-    "x-amz-copy-source-server-side-encryption-customer-algorithm";
-pub const AMZ_COPY_SOURCE_SERVER_SIDE_ENCRYPTION_CUSTOMER_KEY: &str = "x-amz-copy-source-server-side-encryption-customer-key";
-pub const AMZ_COPY_SOURCE_SERVER_SIDE_ENCRYPTION_CUSTOMER_KEY_MD5: &str =
-    "x-amz-copy-source-server-side-encryption-customer-key-md5";
--- a/crates/iam/src/error.rs
+++ b/crates/iam/src/error.rs
@@ -109,6 +109,9 @@ pub enum Error {

    #[error("io error: {0}")]
    Io(std::io::Error),
+
+    #[error("system already initialized")]
+    IamSysAlreadyInitialized,
 }

 impl PartialEq for Error {
@@ -162,6 +165,7 @@ impl Clone for Error {
            Error::PolicyTooLarge => Error::PolicyTooLarge,
            Error::ConfigNotFound => Error::ConfigNotFound,
            Error::Io(e) => Error::Io(std::io::Error::new(e.kind(), e.to_string())),
+            Error::IamSysAlreadyInitialized => Error::IamSysAlreadyInitialized,
        }
    }
 }
@@ -226,6 +230,7 @@ impl From<rustfs_policy::error::Error> for Error {
            rustfs_policy::error::Error::StringError(s) => Error::StringError(s),
            rustfs_policy::error::Error::CryptoError(e) => Error::CryptoError(e),
            rustfs_policy::error::Error::ErrCredMalformed => Error::ErrCredMalformed,
+            rustfs_policy::error::Error::IamSysAlreadyInitialized => Error::IamSysAlreadyInitialized,
        }
    }
 }
--- a/crates/iam/src/lib.rs
+++ b/crates/iam/src/lib.rs
@@ -18,30 +18,58 @@ use rustfs_ecstore::store::ECStore;
 use std::sync::{Arc, OnceLock};
 use store::object::ObjectStore;
 use sys::IamSys;
-use tracing::{debug, instrument};
+use tracing::{error, info, instrument};

 pub mod cache;
 pub mod error;
 pub mod manager;
 pub mod store;
-pub mod utils;
-
 pub mod sys;
+pub mod utils;

 static IAM_SYS: OnceLock<Arc<IamSys<ObjectStore>>> = OnceLock::new();

 #[instrument(skip(ecstore))]
 pub async fn init_iam_sys(ecstore: Arc<ECStore>) -> Result<()> {
-    debug!("init iam system");
-    let s = IamCache::new(ObjectStore::new(ecstore).await).await;
+    if IAM_SYS.get().is_some() {
+        info!("IAM system already initialized, skipping.");
+        return Ok(());
+    }

-    IAM_SYS.get_or_init(move || IamSys::new(s).into());
+    info!("Starting IAM system initialization sequence...");
+
+    // 1. Create the persistent storage adapter
+    let storage_adapter = ObjectStore::new(ecstore);
+
+    // 2. Create the cache manager.
+    // The `new` method now performs a blocking initial load from disk.
+    let cache_manager = IamCache::new(storage_adapter).await;
+
+    // 3. Construct the system interface
+    let iam_instance = Arc::new(IamSys::new(cache_manager));
+
+    // 4. Securely set the global singleton
+    if IAM_SYS.set(iam_instance).is_err() {
+        error!("Critical: Race condition detected during IAM initialization!");
+        return Err(Error::IamSysAlreadyInitialized);
+    }
+
+    info!("IAM system initialization completed successfully.");
    Ok(())
 }

 #[inline]
 pub fn get() -> Result<Arc<IamSys<ObjectStore>>> {
-    IAM_SYS.get().map(Arc::clone).ok_or(Error::IamSysNotInitialized)
+    let sys = IAM_SYS.get().map(Arc::clone).ok_or(Error::IamSysNotInitialized)?;
+
+    // Double-check the internal readiness state. The OnceLock is only set
+    // after initialization and data loading complete, so this is a defensive
+    // guard to ensure callers never operate on a partially initialized system.
+    if !sys.is_ready() {
+        return Err(Error::IamSysNotInitialized);
+    }
+
+    Ok(sys)
 }

 pub fn get_global_iam_sys() -> Option<Arc<IamSys<ObjectStore>>> {
--- a/crates/iam/src/manager.rs
+++ b/crates/iam/src/manager.rs
@@ -37,6 +37,7 @@ use rustfs_policy::{
 use rustfs_utils::path::path_join_buf;
 use serde::{Deserialize, Serialize};
 use serde_json::Value;
+use std::sync::atomic::AtomicU8;
 use std::{
    collections::{HashMap, HashSet},
    sync::{
@@ -76,9 +77,19 @@ fn get_iam_format_file_path() -> String {
    path_join_buf(&[&IAM_CONFIG_PREFIX, IAM_FORMAT_FILE])
 }

+#[repr(u8)]
+#[derive(Debug, PartialEq)]
+pub enum IamState {
+    Uninitialized = 0,
+    Loading = 1,
+    Ready = 2,
+    Error = 3,
+}
+
 pub struct IamCache<T> {
    pub cache: Cache,
    pub api: T,
+    pub state: Arc<AtomicU8>,
    pub loading: Arc<AtomicBool>,
    pub roles: HashMap<ARN, Vec<String>>,
    pub send_chan: Sender<i64>,
@@ -89,12 +100,19 @@ impl<T> IamCache<T>
 where
    T: Store,
 {
+    /// Create a new IAM system instance
+    /// # Arguments
+    /// * `api` - The storage backend implementing the Store trait
+    ///
+    /// # Returns
+    /// An Arc-wrapped instance of IamSystem
    pub(crate) async fn new(api: T) -> Arc<Self> {
        let (sender, receiver) = mpsc::channel::<i64>(100);

        let sys = Arc::new(Self {
            api,
            cache: Cache::default(),
+            state: Arc::new(AtomicU8::new(IamState::Uninitialized as u8)),
            loading: Arc::new(AtomicBool::new(false)),
            send_chan: sender,
            roles: HashMap::new(),
@@ -105,10 +123,32 @@ where
        sys
    }

+    /// Initialize the IAM system
    async fn init(self: Arc<Self>, receiver: Receiver<i64>) -> Result<()> {
+        self.state.store(IamState::Loading as u8, Ordering::SeqCst);
+        // Ensure the IAM format file is persisted first
        self.clone().save_iam_formatter().await?;
-        self.clone().load().await?;

+        // Critical: Load all existing users/policies into memory cache
+        const MAX_RETRIES: usize = 3;
+        for attempt in 0..MAX_RETRIES {
+            if let Err(e) = self.clone().load().await {
+                if attempt == MAX_RETRIES - 1 {
+                    self.state.store(IamState::Error as u8, Ordering::SeqCst);
+                    error!("IAM fail to load initial data after {} attempts: {:?}", MAX_RETRIES, e);
+                    return Err(e);
+                } else {
+                    warn!("IAM load failed, retrying... attempt {}", attempt + 1);
+                    tokio::time::sleep(Duration::from_secs(1)).await;
+                }
+            } else {
+                break;
+            }
+        }
+        self.state.store(IamState::Ready as u8, Ordering::SeqCst);
+        info!("IAM System successfully initialized and marked as READY");
+
+        // Background ticker for synchronization
        // Check if environment variable is set
        let skip_background_task = std::env::var("RUSTFS_SKIP_BACKGROUND_TASK").is_ok();

@@ -152,6 +192,11 @@ where
        Ok(())
    }

+    /// Check if IAM system is ready
+    pub fn is_ready(&self) -> bool {
+        self.state.load(Ordering::SeqCst) == IamState::Ready as u8
+    }
+
    async fn _notify(&self) {
        self.send_chan.send(OffsetDateTime::now_utc().unix_timestamp()).await.unwrap();
    }
--- a/crates/iam/src/store/object.rs
+++ b/crates/iam/src/store/object.rs
@@ -38,7 +38,7 @@ use std::sync::LazyLock;
 use std::{collections::HashMap, sync::Arc};
 use tokio::sync::mpsc::{self, Sender};
 use tokio_util::sync::CancellationToken;
-use tracing::{info, warn};
+use tracing::{debug, error, info, warn};

 pub static IAM_CONFIG_PREFIX: LazyLock<String> = LazyLock::new(|| format!("{RUSTFS_CONFIG_PREFIX}/iam"));
 pub static IAM_CONFIG_USERS_PREFIX: LazyLock<String> = LazyLock::new(|| format!("{RUSTFS_CONFIG_PREFIX}/iam/users/"));
@@ -120,52 +120,18 @@ fn split_path(s: &str, last_index: bool) -> (&str, &str) {
 #[derive(Clone)]
 pub struct ObjectStore {
    object_api: Arc<ECStore>,
-    prev_cred: Option<rustfs_policy::auth::Credentials>,
 }

 impl ObjectStore {
    const BUCKET_NAME: &'static str = ".rustfs.sys";
-    const PREV_CRED_FILE: &'static str = "config/iam/prev_cred.json";

-    /// Load previous credentials from persistent storage in .rustfs.sys bucket
-    async fn load_prev_cred(object_api: Arc<ECStore>) -> Option<rustfs_policy::auth::Credentials> {
-        match read_config(object_api, Self::PREV_CRED_FILE).await {
-            Ok(data) => serde_json::from_slice::<rustfs_policy::auth::Credentials>(&data).ok(),
-            Err(_) => None,
-        }
+    pub fn new(object_api: Arc<ECStore>) -> Self {
+        Self { object_api }
    }

-    /// Save previous credentials to persistent storage in .rustfs.sys bucket
-    async fn save_prev_cred(object_api: Arc<ECStore>, cred: &Option<rustfs_policy::auth::Credentials>) -> Result<()> {
-        match cred {
-            Some(c) => {
-                let data = serde_json::to_vec(c).map_err(|e| Error::other(format!("Failed to serialize cred: {}", e)))?;
-                save_config(object_api, Self::PREV_CRED_FILE, data)
-                    .await
-                    .map_err(|e| Error::other(format!("Failed to write cred to storage: {}", e)))
-            }
-            None => {
-                // If no credentials, remove the config
-                match delete_config(object_api, Self::PREV_CRED_FILE).await {
-                    Ok(_) => Ok(()),
-                    Err(e) => {
-                        // Ignore ConfigNotFound error when trying to delete non-existent config
-                        if matches!(e, rustfs_ecstore::error::StorageError::ConfigNotFound) {
-                            Ok(())
-                        } else {
-                            Err(Error::other(format!("Failed to delete cred from storage: {}", e)))
-                        }
-                    }
-                }
-            }
-        }
-    }
-
-    pub async fn new(object_api: Arc<ECStore>) -> Self {
-        // Load previous credentials from persistent storage in .rustfs.sys bucket
-        let prev_cred = Self::load_prev_cred(object_api.clone()).await.or_else(get_global_action_cred);
-
-        Self { object_api, prev_cred }
+    fn decrypt_data(data: &[u8]) -> Result<Vec<u8>> {
+        let de = rustfs_crypto::decrypt_data(get_global_action_cred().unwrap_or_default().secret_key.as_bytes(), data)?;
+        Ok(de)
    }

    fn encrypt_data(data: &[u8]) -> Result<Vec<u8>> {
@@ -173,65 +139,10 @@ impl ObjectStore {
        Ok(en)
    }

-    /// Decrypt data with credential fallback mechanism
-    /// First tries current credentials, then falls back to previous credentials if available
-    async fn decrypt_fallback(&self, data: &[u8], path: &str) -> Result<Vec<u8>> {
-        let current_cred = get_global_action_cred().unwrap_or_default();
-
-        // Try current credentials first
-        match rustfs_crypto::decrypt_data(current_cred.secret_key.as_bytes(), data) {
-            Ok(decrypted) => {
-                // Update persistent storage with current credentials for consistency
-                let _ = Self::save_prev_cred(self.object_api.clone(), &Some(current_cred)).await;
-                Ok(decrypted)
-            }
-            Err(_) => {
-                // Current credentials failed, try previous credentials
-                if let Some(ref prev_cred) = self.prev_cred {
-                    match rustfs_crypto::decrypt_data(prev_cred.secret_key.as_bytes(), data) {
-                        Ok(prev_decrypted) => {
-                            warn!("Decryption succeeded with previous credentials, path: {}", path);
-
-                            // Re-encrypt with current credentials
-                            match rustfs_crypto::encrypt_data(current_cred.secret_key.as_bytes(), &prev_decrypted) {
-                                Ok(re_encrypted) => {
-                                    let _ = save_config(self.object_api.clone(), path, re_encrypted).await;
-                                }
-                                Err(e) => {
-                                    warn!("Failed to re-encrypt with current credentials: {}, path: {}", e, path);
-                                }
-                            }
-
-                            // Update persistent storage with current credentials
-                            let _ = Self::save_prev_cred(self.object_api.clone(), &Some(current_cred)).await;
-                            Ok(prev_decrypted)
-                        }
-                        Err(_) => {
-                            // Both attempts failed
-                            warn!("Decryption failed with both current and previous credentials, deleting config: {}", path);
-                            let _ = self.delete_iam_config(path).await;
-                            Err(Error::ConfigNotFound)
-                        }
-                    }
-                } else {
-                    // No previous credentials available
-                    warn!(
-                        "Decryption failed with current credentials and no previous credentials available, deleting config: {}",
-                        path
-                    );
-                    let _ = self.delete_iam_config(path).await;
-                    Err(Error::ConfigNotFound)
-                }
-            }
-        }
-    }
-
    async fn load_iamconfig_bytes_with_metadata(&self, path: impl AsRef<str> + Send) -> Result<(Vec<u8>, ObjectInfo)> {
        let (data, obj) = read_config_with_metadata(self.object_api.clone(), path.as_ref(), &ObjectOptions::default()).await?;

-        let decrypted_data = self.decrypt_fallback(&data, path.as_ref()).await?;
-
-        Ok((decrypted_data, obj))
+        Ok((Self::decrypt_data(&data)?, obj))
    }

    async fn list_iam_config_items(&self, prefix: &str, ctx: CancellationToken, sender: Sender<StringOrErr>) {
@@ -430,6 +341,27 @@ impl ObjectStore {
        Ok(policies)
    }

+    /// Checks if the underlying ECStore is ready for metadata operations.
+    /// This prevents silent failures during the storage boot-up phase.
+    ///
+    /// Performs a lightweight probe by attempting to read a known configuration object.
+    /// If the object is not found, it indicates the storage metadata is not ready.
+    /// The upper-level caller should handle retries if needed.
+    async fn check_storage_readiness(&self) -> Result<()> {
+        // Probe path for a fixed object under the IAM root prefix.
+        // If it doesn't exist, the system bucket or metadata is not ready.
+        let probe_path = format!("{}/format.json", *IAM_CONFIG_PREFIX);
+
+        match read_config(self.object_api.clone(), &probe_path).await {
+            Ok(_) => Ok(()),
+            Err(rustfs_ecstore::error::StorageError::ConfigNotFound) => Err(Error::other(format!(
+                "Storage metadata not ready: probe object '{}' not found (expected IAM config to be initialized)",
+                probe_path
+            ))),
+            Err(e) => Err(e.into()),
+        }
+    }
+
    // async fn load_policy(&self, name: &str) -> Result<PolicyDoc> {
    //     let mut policy = self
    //         .load_iam_config::<PolicyDoc>(&format!("config/iam/policies/{name}/policy.json"))
@@ -475,17 +407,62 @@ impl Store for ObjectStore {
    async fn load_iam_config<Item: DeserializeOwned>(&self, path: impl AsRef<str> + Send) -> Result<Item> {
        let mut data = read_config(self.object_api.clone(), path.as_ref()).await?;

-        data = self.decrypt_fallback(&data, path.as_ref()).await?;
+        data = match Self::decrypt_data(&data) {
+            Ok(v) => v,
+            Err(err) => {
+                warn!("delete the config file when decrypt failed failed: {}, path: {}", err, path.as_ref());
+                // delete the config file when decrypt failed
+                let _ = self.delete_iam_config(path.as_ref()).await;
+                return Err(Error::ConfigNotFound);
+            }
+        };

        Ok(serde_json::from_slice(&data)?)
    }
+    /// Saves IAM configuration with a retry mechanism on failure.
+    ///
+    /// Attempts to save the IAM configuration up to 5 times if the storage layer is not ready,
+    /// using exponential backoff between attempts (starting at 200ms, doubling each retry).
+    ///
+    /// # Arguments
+    ///
+    /// * `item` - The IAM configuration item to save, must implement `Serialize` and `Send`.
+    /// * `path` - The path where the configuration will be saved.
+    ///
+    /// # Returns
+    ///
+    /// * `Result<()>` - `Ok(())` on success, or an `Error` if all attempts fail.
    #[tracing::instrument(level = "debug", skip(self, item, path))]
    async fn save_iam_config<Item: Serialize + Send>(&self, item: Item, path: impl AsRef<str> + Send) -> Result<()> {
        let mut data = serde_json::to_vec(&item)?;
        data = Self::encrypt_data(&data)?;

-        save_config(self.object_api.clone(), path.as_ref(), data).await?;
-        Ok(())
+        let mut attempts = 0;
+        let max_attempts = 5;
+        let path_ref = path.as_ref();
+
+        loop {
+            match save_config(self.object_api.clone(), path_ref, data.clone()).await {
+                Ok(_) => {
+                    debug!("Successfully saved IAM config to {}", path_ref);
+                    return Ok(());
+                }
+                Err(e) if attempts < max_attempts => {
+                    attempts += 1;
+                    // Exponential backoff: 200ms, 400ms, 800ms...
+                    let wait_ms = 200 * (1 << attempts);
+                    warn!(
+                        "Storage layer not ready for IAM write (attempt {}/{}). Retrying in {}ms. Path: {}, Error: {:?}",
+                        attempts, max_attempts, wait_ms, path_ref, e
+                    );
+                    tokio::time::sleep(std::time::Duration::from_millis(wait_ms)).await;
+                }
+                Err(e) => {
+                    error!("Final failure saving IAM config to {}: {:?}", path_ref, e);
+                    return Err(e.into());
+                }
+            }
+        }
    }
    async fn delete_iam_config(&self, path: impl AsRef<str> + Send) -> Result<()> {
        delete_config(self.object_api.clone(), path.as_ref()).await?;
@@ -499,8 +476,16 @@ impl Store for ObjectStore {
        user_identity: UserIdentity,
        _ttl: Option<usize>,
    ) -> Result<()> {
-        self.save_iam_config(user_identity, get_user_identity_path(name, user_type))
-            .await
+        // Pre-check storage health
+        self.check_storage_readiness().await?;
+
+        let path = get_user_identity_path(name, user_type);
+        debug!("Saving IAM identity to path: {}", path);
+
+        self.save_iam_config(user_identity, path).await.map_err(|e| {
+            error!("ObjectStore save failure for {}: {:?}", name, e);
+            e
+        })
    }
    async fn delete_user_identity(&self, name: &str, user_type: UserType) -> Result<()> {
        self.delete_iam_config(get_user_identity_path(name, user_type))
--- a/crates/iam/src/sys.rs
+++ b/crates/iam/src/sys.rs
@@ -67,6 +67,13 @@ pub struct IamSys<T> {
 }

 impl<T: Store> IamSys<T> {
+    /// Create a new IamSys instance with the given IamCache store
+    ///
+    /// # Arguments
+    /// * `store` - An Arc to the IamCache instance
+    ///
+    /// # Returns
+    /// A new instance of IamSys
    pub fn new(store: Arc<IamCache<T>>) -> Self {
        tokio::spawn(async move {
            match opa::lookup_config().await {
@@ -87,6 +94,11 @@ impl<T: Store> IamSys<T> {
            roles_map: HashMap::new(),
        }
    }
+
+    /// Check if the IamSys has a watcher configured
+    ///
+    /// # Returns
+    /// `true` if a watcher is configured, `false` otherwise
    pub fn has_watcher(&self) -> bool {
        self.store.api.has_watcher()
    }
@@ -859,6 +871,11 @@ impl<T: Store> IamSys<T> {

        self.get_combined_policy(&policies).await.is_allowed(args).await
    }
+
+    /// Check if the underlying store is ready
+    pub fn is_ready(&self) -> bool {
+        self.store.is_ready()
+    }
 }

 fn is_allowed_by_session_policy(args: &Args<'_>) -> (bool, bool) {
--- a/crates/lock/src/fast_lock/benchmarks.rs
+++ b/crates/lock/src/fast_lock/benchmarks.rs
@@ -1,325 +0,0 @@
-// Copyright 2024 RustFS Team
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-// Benchmarks comparing fast lock vs old lock performance
-
-#[cfg(test)]
-#[allow(dead_code)] // Temporarily disable benchmark tests
-mod benchmarks {
-    use super::super::*;
-    use std::sync::Arc;
-    use std::time::{Duration, Instant};
-    use tokio::task;
-
-    /// Benchmark single-threaded lock operations
-    #[tokio::test]
-    async fn bench_single_threaded_fast_locks() {
-        let manager = Arc::new(FastObjectLockManager::new());
-        let iterations = 10000;
-        
-        // Warm up
-        for i in 0..100 {
-            let _guard = manager
-                .acquire_write_lock("bucket", &format!("warm_{}", i), "owner")
-                .await
-                .unwrap();
-        }
-
-        // Benchmark write locks
-        let start = Instant::now();
-        for i in 0..iterations {
-            let _guard = manager
-                .acquire_write_lock("bucket", &format!("object_{}", i), "owner")
-                .await
-                .unwrap();
-        }
-        let duration = start.elapsed();
-        
-        println!("Fast locks: {} write locks in {:?}", iterations, duration);
-        println!("Average: {:?} per lock", duration / iterations);
-        
-        let metrics = manager.get_metrics();
-        println!("Fast path rate: {:.2}%", metrics.shard_metrics.fast_path_rate() * 100.0);
-        
-        // Should be much faster than old implementation
-        assert!(duration.as_millis() < 1000, "Should complete 10k locks in <1s");
-        assert!(metrics.shard_metrics.fast_path_rate() > 0.95, "Should have >95% fast path rate");
-    }
-
-    /// Benchmark concurrent lock operations
-    #[tokio::test]
-    async fn bench_concurrent_fast_locks() {
-        let manager = Arc::new(FastObjectLockManager::new());
-        let concurrent_tasks = 100;
-        let iterations_per_task = 100;
-        
-        let start = Instant::now();
-        
-        let mut handles = Vec::new();
-        for task_id in 0..concurrent_tasks {
-            let manager_clone = manager.clone();
-            let handle = task::spawn(async move {
-                for i in 0..iterations_per_task {
-                    let object_name = format!("obj_{}_{}", task_id, i);
-                    let _guard = manager_clone
-                        .acquire_write_lock("bucket", &object_name, &format!("owner_{}", task_id))
-                        .await
-                        .unwrap();
-                    
-                    // Simulate some work
-                    tokio::task::yield_now().await;
-                }
-            });
-            handles.push(handle);
-        }
-        
-        // Wait for all tasks
-        for handle in handles {
-            handle.await.unwrap();
-        }
-        
-        let duration = start.elapsed();
-        let total_ops = concurrent_tasks * iterations_per_task;
-        
-        println!("Concurrent fast locks: {} operations across {} tasks in {:?}", 
-                 total_ops, concurrent_tasks, duration);
-        println!("Throughput: {:.2} ops/sec", total_ops as f64 / duration.as_secs_f64());
-        
-        let metrics = manager.get_metrics();
-        println!("Fast path rate: {:.2}%", metrics.shard_metrics.fast_path_rate() * 100.0);
-        println!("Contention events: {}", metrics.shard_metrics.contention_events);
-        
-        // Should maintain high throughput even with concurrency
-        assert!(duration.as_millis() < 5000, "Should complete concurrent ops in <5s");
-    }
-
-    /// Benchmark contended lock operations
-    #[tokio::test]
-    async fn bench_contended_locks() {
-        let manager = Arc::new(FastObjectLockManager::new());
-        let concurrent_tasks = 50;
-        let shared_objects = 10; // High contention on few objects
-        let iterations_per_task = 50;
-        
-        let start = Instant::now();
-        
-        let mut handles = Vec::new();
-        for task_id in 0..concurrent_tasks {
-            let manager_clone = manager.clone();
-            let handle = task::spawn(async move {
-                for i in 0..iterations_per_task {
-                    let object_name = format!("shared_{}", i % shared_objects);
-                    
-                    // Mix of read and write operations
-                    if i % 3 == 0 {
-                        // Write operation
-                        if let Ok(_guard) = manager_clone
-                            .acquire_write_lock("bucket", &object_name, &format!("owner_{}", task_id))
-                            .await
-                        {
-                            tokio::task::yield_now().await;
-                        }
-                    } else {
-                        // Read operation
-                        if let Ok(_guard) = manager_clone
-                            .acquire_read_lock("bucket", &object_name, &format!("owner_{}", task_id))
-                            .await
-                        {
-                            tokio::task::yield_now().await;
-                        }
-                    }
-                }
-            });
-            handles.push(handle);
-        }
-        
-        // Wait for all tasks
-        for handle in handles {
-            handle.await.unwrap();
-        }
-        
-        let duration = start.elapsed();
-        
-        println!("Contended locks: {} tasks on {} objects in {:?}", 
-                 concurrent_tasks, shared_objects, duration);
-        
-        let metrics = manager.get_metrics();
-        println!("Total acquisitions: {}", metrics.shard_metrics.total_acquisitions());
-        println!("Fast path rate: {:.2}%", metrics.shard_metrics.fast_path_rate() * 100.0);
-        println!("Average wait time: {:?}", metrics.shard_metrics.avg_wait_time());
-        println!("Timeout rate: {:.2}%", metrics.shard_metrics.timeout_rate() * 100.0);
-        
-        // Even with contention, should maintain reasonable performance
-        assert!(metrics.shard_metrics.timeout_rate() < 0.1, "Should have <10% timeout rate");
-        assert!(metrics.shard_metrics.avg_wait_time() < Duration::from_millis(100), "Avg wait should be <100ms");
-    }
-
-    /// Benchmark batch operations
-    #[tokio::test]
-    async fn bench_batch_operations() {
-        let manager = FastObjectLockManager::new();
-        let batch_sizes = vec![10, 50, 100, 500];
-        
-        for batch_size in batch_sizes {
-            // Create batch request
-            let mut batch = BatchLockRequest::new("batch_owner");
-            for i in 0..batch_size {
-                batch = batch.add_write_lock("bucket", &format!("batch_obj_{}", i));
-            }
-            
-            let start = Instant::now();
-            let result = manager.acquire_locks_batch(batch).await;
-            let duration = start.elapsed();
-            
-            assert!(result.all_acquired, "Batch should succeed");
-            println!("Batch size {}: {:?} ({:.2} μs per lock)", 
-                     batch_size, 
-                     duration,
-                     duration.as_micros() as f64 / batch_size as f64);
-            
-            // Batch should be much faster than individual acquisitions
-            assert!(duration.as_millis() < batch_size as u128 / 10, 
-                    "Batch should be 10x+ faster than individual locks");
-        }
-    }
-
-    /// Benchmark version-specific locks
-    #[tokio::test]
-    async fn bench_versioned_locks() {
-        let manager = Arc::new(FastObjectLockManager::new());
-        let objects = 100;
-        let versions_per_object = 10;
-        
-        let start = Instant::now();
-        
-        let mut handles = Vec::new();
-        for obj_id in 0..objects {
-            let manager_clone = manager.clone();
-            let handle = task::spawn(async move {
-                for version in 0..versions_per_object {
-                    let _guard = manager_clone
-                        .acquire_write_lock_versioned(
-                            "bucket",
-                            &format!("obj_{}", obj_id),
-                            &format!("v{}", version),
-                            "version_owner"
-                        )
-                        .await
-                        .unwrap();
-                }
-            });
-            handles.push(handle);
-        }
-        
-        for handle in handles {
-            handle.await.unwrap();
-        }
-        
-        let duration = start.elapsed();
-        let total_ops = objects * versions_per_object;
-        
-        println!("Versioned locks: {} version locks in {:?}", total_ops, duration);
-        println!("Throughput: {:.2} locks/sec", total_ops as f64 / duration.as_secs_f64());
-        
-        let metrics = manager.get_metrics();
-        println!("Fast path rate: {:.2}%", metrics.shard_metrics.fast_path_rate() * 100.0);
-        
-        // Versioned locks should not interfere with each other
-        assert!(metrics.shard_metrics.fast_path_rate() > 0.9, "Should maintain high fast path rate");
-    }
-
-    /// Compare with theoretical maximum performance
-    #[tokio::test]
-    async fn bench_theoretical_maximum() {
-        let manager = Arc::new(FastObjectLockManager::new());
-        let iterations = 100000;
-        
-        // Measure pure fast path performance (no contention)
-        let start = Instant::now();
-        for i in 0..iterations {
-            let _guard = manager
-                .acquire_write_lock("bucket", &format!("unique_{}", i), "owner")
-                .await
-                .unwrap();
-        }
-        let duration = start.elapsed();
-        
-        println!("Theoretical maximum: {} unique locks in {:?}", iterations, duration);
-        println!("Rate: {:.2} locks/sec", iterations as f64 / duration.as_secs_f64());
-        println!("Latency: {:?} per lock", duration / iterations);
-        
-        let metrics = manager.get_metrics();
-        println!("Fast path rate: {:.2}%", metrics.shard_metrics.fast_path_rate() * 100.0);
-        
-        // Should achieve very high performance with no contention
-        assert!(metrics.shard_metrics.fast_path_rate() > 0.99, "Should be nearly 100% fast path");
-        assert!(duration.as_secs_f64() / (iterations as f64) < 0.0001, "Should be <100μs per lock");
-    }
-
-    /// Performance regression test
-    #[tokio::test]
-    async fn performance_regression_test() {
-        let manager = Arc::new(FastObjectLockManager::new());
-        
-        // This test ensures we maintain performance targets
-        let test_cases = vec![
-            ("single_thread", 1, 10000),
-            ("low_contention", 10, 1000),
-            ("high_contention", 100, 100),
-        ];
-        
-        for (test_name, threads, ops_per_thread) in test_cases {
-            let start = Instant::now();
-            
-            let mut handles = Vec::new();
-            for thread_id in 0..threads {
-                let manager_clone = manager.clone();
-                let handle = task::spawn(async move {
-                    for op_id in 0..ops_per_thread {
-                        let object = if threads == 1 {
-                            format!("obj_{}_{}", thread_id, op_id)
-                        } else {
-                            format!("obj_{}", op_id % 100) // Create contention
-                        };
-                        
-                        let owner = format!("owner_{}", thread_id);
-                        let _guard = manager_clone
-                            .acquire_write_lock("bucket", object, owner)
-                            .await
-                            .unwrap();
-                    }
-                });
-                handles.push(handle);
-            }
-            
-            for handle in handles {
-                handle.await.unwrap();
-            }
-            
-            let duration = start.elapsed();
-            let total_ops = threads * ops_per_thread;
-            let ops_per_sec = total_ops as f64 / duration.as_secs_f64();
-            
-            println!("{}: {:.2} ops/sec", test_name, ops_per_sec);
-            
-            // Performance targets (adjust based on requirements)
-            match test_name {
-                "single_thread" => assert!(ops_per_sec > 50000.0, "Single thread should exceed 50k ops/sec"),
-                "low_contention" => assert!(ops_per_sec > 20000.0, "Low contention should exceed 20k ops/sec"),
-                "high_contention" => assert!(ops_per_sec > 5000.0, "High contention should exceed 5k ops/sec"),
-                _ => {}
-            }
-        }
-    }
-}
--- a/crates/lock/src/fast_lock/mod.rs
+++ b/crates/lock/src/fast_lock/mod.rs
@@ -37,9 +37,6 @@ pub mod shard;
 pub mod state;
 pub mod types;

-// #[cfg(test)]
-// pub mod benchmarks; // Temporarily disabled due to compilation issues
-
 // Re-export main types
 pub use disabled_manager::DisabledLockManager;
 pub use guard::FastLockGuard;
--- a/crates/mcp/Dockerfile
+++ b/crates/mcp/Dockerfile
@@ -12,4 +12,6 @@ WORKDIR /app

 COPY --from=builder /build/target/release/rustfs-mcp /app/

-ENTRYPOINT ["/app/rustfs-mcp"]
+RUN apt-get update && apt-get install -y ca-certificates && update-ca-certificates
+
+ENTRYPOINT ["/app/rustfs-mcp"]
--- a/crates/notify/Cargo.toml
+++ b/crates/notify/Cargo.toml
@@ -30,6 +30,7 @@ rustfs-config = { workspace = true, features = ["notify", "constants"] }
 rustfs-ecstore = { workspace = true }
 rustfs-targets = { workspace = true }
 rustfs-utils = { workspace = true }
+arc-swap = { workspace = true }
 async-trait = { workspace = true }
 chrono = { workspace = true, features = ["serde"] }
 futures = { workspace = true }
--- a/crates/notify/src/factory.rs
+++ b/crates/notify/src/factory.rs
@@ -60,8 +60,9 @@ impl TargetFactory for WebhookTargetFactory {
        let endpoint = config
            .lookup(WEBHOOK_ENDPOINT)
            .ok_or_else(|| TargetError::Configuration("Missing webhook endpoint".to_string()))?;
-        let endpoint_url = Url::parse(&endpoint)
-            .map_err(|e| TargetError::Configuration(format!("Invalid endpoint URL: {e} (value: '{endpoint}')")))?;
+        let parsed_endpoint = endpoint.trim();
+        let endpoint_url = Url::parse(parsed_endpoint)
+            .map_err(|e| TargetError::Configuration(format!("Invalid endpoint URL: {e} (value: '{parsed_endpoint}')")))?;

        let args = WebhookArgs {
            enable: true, // If we are here, it's already enabled.
--- a/crates/notify/src/integration.rs
+++ b/crates/notify/src/integration.rs
@@ -12,6 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.

+use crate::notification_system_subscriber::NotificationSystemSubscriberView;
 use crate::{
    Event, error::NotificationError, notifier::EventNotifier, registry::TargetRegistry, rules::BucketNotificationConfig, stream,
 };
@@ -104,6 +105,8 @@ pub struct NotificationSystem {
    concurrency_limiter: Arc<Semaphore>,
    /// Monitoring indicators
    metrics: Arc<NotificationMetrics>,
+    /// Subscriber view
+    subscriber_view: NotificationSystemSubscriberView,
 }

 impl NotificationSystem {
@@ -112,6 +115,7 @@ impl NotificationSystem {
        let concurrency_limiter =
            rustfs_utils::get_env_usize(ENV_NOTIFY_TARGET_STREAM_CONCURRENCY, DEFAULT_NOTIFY_TARGET_STREAM_CONCURRENCY);
        NotificationSystem {
+            subscriber_view: NotificationSystemSubscriberView::new(),
            notifier: Arc::new(EventNotifier::new()),
            registry: Arc::new(TargetRegistry::new()),
            config: Arc::new(RwLock::new(config)),
@@ -188,8 +192,11 @@ impl NotificationSystem {
    }

    /// Checks if there are active subscribers for the given bucket and event name.
-    pub async fn has_subscriber(&self, bucket: &str, event_name: &EventName) -> bool {
-        self.notifier.has_subscriber(bucket, event_name).await
+    pub async fn has_subscriber(&self, bucket: &str, event: &EventName) -> bool {
+        if !self.subscriber_view.has_subscriber(bucket, event) {
+            return false;
+        }
+        self.notifier.has_subscriber(bucket, event).await
    }

    async fn update_config_and_reload<F>(&self, mut modifier: F) -> Result<(), NotificationError>
@@ -236,15 +243,18 @@ impl NotificationSystem {
    pub async fn remove_target(&self, target_id: &TargetID, target_type: &str) -> Result<(), NotificationError> {
        info!("Attempting to remove target: {}", target_id);

+        let ttype = target_type.to_lowercase();
+        let tname = target_id.name.to_lowercase();
+
        self.update_config_and_reload(|config| {
            let mut changed = false;
-            if let Some(targets_of_type) = config.0.get_mut(target_type) {
-                if targets_of_type.remove(&target_id.name).is_some() {
+            if let Some(targets_of_type) = config.0.get_mut(&ttype) {
+                if targets_of_type.remove(&tname).is_some() {
                    info!("Removed target {} from configuration", target_id);
                    changed = true;
                }
                if targets_of_type.is_empty() {
-                    config.0.remove(target_type);
+                    config.0.remove(&ttype);
                }
            }
            if !changed {
@@ -269,20 +279,24 @@ impl NotificationSystem {
    /// If the target configuration is invalid, it returns Err(NotificationError::Configuration).
    pub async fn set_target_config(&self, target_type: &str, target_name: &str, kvs: KVS) -> Result<(), NotificationError> {
        info!("Setting config for target {} of type {}", target_name, target_type);
+        let ttype = target_type.to_lowercase();
+        let tname = target_name.to_lowercase();
        self.update_config_and_reload(|config| {
-            config
-                .0
-                .entry(target_type.to_lowercase())
-                .or_default()
-                .insert(target_name.to_lowercase(), kvs.clone());
+            config.0.entry(ttype.clone()).or_default().insert(tname.clone(), kvs.clone());
            true // The configuration is always modified
        })
        .await
    }

    /// Removes all notification configurations for a bucket.
-    pub async fn remove_bucket_notification_config(&self, bucket_name: &str) {
-        self.notifier.remove_rules_map(bucket_name).await;
+    /// If the configuration is successfully removed, the entire notification system will be automatically reloaded.
+    ///
+    /// # Arguments
+    /// * `bucket` - The name of the bucket whose notification configuration is to be removed.
+    ///
+    pub async fn remove_bucket_notification_config(&self, bucket: &str) {
+        self.subscriber_view.clear_bucket(bucket);
+        self.notifier.remove_rules_map(bucket).await;
    }

    /// Removes a Target configuration.
@@ -299,11 +313,28 @@ impl NotificationSystem {
    /// If the target configuration does not exist, it returns Ok(()) without making any changes.
    pub async fn remove_target_config(&self, target_type: &str, target_name: &str) -> Result<(), NotificationError> {
        info!("Removing config for target {} of type {}", target_name, target_type);
+
+        let ttype = target_type.to_lowercase();
+        let tname = target_name.to_lowercase();
+
+        let target_id = TargetID {
+            id: tname.clone(),
+            name: ttype.clone(),
+        };
+
+        // Deletion is prohibited if bucket rules refer to it
+        if self.notifier.is_target_bound_to_any_bucket(&target_id).await {
+            return Err(NotificationError::Configuration(format!(
+                "Target is still bound to bucket rules and deletion is prohibited: type={} name={}",
+                ttype, tname
+            )));
+        }
+
        let config_result = self
            .update_config_and_reload(|config| {
                let mut changed = false;
-                if let Some(targets) = config.0.get_mut(&target_type.to_lowercase()) {
-                    if targets.remove(&target_name.to_lowercase()).is_some() {
+                if let Some(targets) = config.0.get_mut(&ttype) {
+                    if targets.remove(&tname).is_some() {
                        changed = true;
                    }
                    if targets.is_empty() {
@@ -319,8 +350,6 @@ impl NotificationSystem {
            .await;

        if config_result.is_ok() {
-            let target_id = TargetID::new(target_name.to_string(), target_type.to_string());
-
            // Remove from target list
            let target_list = self.notifier.target_list();
            let mut target_list_guard = target_list.write().await;
@@ -358,6 +387,9 @@ impl NotificationSystem {
            let _ = cancel_tx.send(()).await;
        }

+        // Clear the target_list and ensure that reload is a replacement reconstruction (solve the target_list len unchanged/residual problem)
+        self.notifier.remove_all_bucket_targets().await;
+
        // Update the config
        self.update_config(new_config.clone()).await;

@@ -388,15 +420,16 @@ impl NotificationSystem {

                    // The storage of the cloned target and the target itself
                    let store_clone = store.boxed_clone();
-                    let target_box = target.clone_dyn();
-                    let target_arc = Arc::from(target_box);
-
-                    // Add a reference to the monitoring metrics
-                    let metrics = self.metrics.clone();
-                    let semaphore = self.concurrency_limiter.clone();
+                    // let target_box = target.clone_dyn();
+                    let target_arc = Arc::from(target.clone_dyn());

                    // Encapsulated enhanced version of start_event_stream
-                    let cancel_tx = self.enhanced_start_event_stream(store_clone, target_arc, metrics, semaphore);
+                    let cancel_tx = self.enhanced_start_event_stream(
+                        store_clone,
+                        target_arc,
+                        self.metrics.clone(),
+                        self.concurrency_limiter.clone(),
+                    );

                    // Start event stream processing and save cancel sender
                    // let cancel_tx = start_event_stream(store_clone, target_clone);
@@ -423,17 +456,18 @@ impl NotificationSystem {
    /// Loads the bucket notification configuration
    pub async fn load_bucket_notification_config(
        &self,
-        bucket_name: &str,
-        config: &BucketNotificationConfig,
+        bucket: &str,
+        cfg: &BucketNotificationConfig,
    ) -> Result<(), NotificationError> {
-        let arn_list = self.notifier.get_arn_list(&config.region).await;
+        self.subscriber_view.apply_bucket_config(bucket, cfg);
+        let arn_list = self.notifier.get_arn_list(&cfg.region).await;
        if arn_list.is_empty() {
            return Err(NotificationError::Configuration("No targets configured".to_string()));
        }
        info!("Available ARNs: {:?}", arn_list);
        // Validate the configuration against the available ARNs
-        if let Err(e) = config.validate(&config.region, &arn_list) {
-            debug!("Bucket notification config validation region:{} failed: {}", &config.region, e);
+        if let Err(e) = cfg.validate(&cfg.region, &arn_list) {
+            debug!("Bucket notification config validation region:{} failed: {}", &cfg.region, e);
            if !e.to_string().contains("ARN not found") {
                return Err(NotificationError::BucketNotification(e.to_string()));
            } else {
@@ -441,9 +475,9 @@ impl NotificationSystem {
            }
        }

-        let rules_map = config.get_rules_map();
-        self.notifier.add_rules_map(bucket_name, rules_map.clone()).await;
-        info!("Loaded notification config for bucket: {}", bucket_name);
+        let rules_map = cfg.get_rules_map();
+        self.notifier.add_rules_map(bucket, rules_map.clone()).await;
+        info!("Loaded notification config for bucket: {}", bucket);
        Ok(())
    }

--- a/crates/notify/src/lib.rs
+++ b/crates/notify/src/lib.rs
@@ -23,6 +23,7 @@ mod event;
 pub mod factory;
 mod global;
 pub mod integration;
+mod notification_system_subscriber;
 pub mod notifier;
 pub mod registry;
 pub mod rules;
--- a/crates/notify/src/notification_system_subscriber.rs
+++ b/crates/notify/src/notification_system_subscriber.rs
@@ -0,0 +1,74 @@
+// Copyright 2024 RustFS Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use crate::BucketNotificationConfig;
+use crate::rules::{BucketRulesSnapshot, DynRulesContainer, SubscriberIndex};
+use rustfs_targets::EventName;
+
+/// NotificationSystemSubscriberView - Provides an interface to manage and query
+/// the subscription status of buckets in the notification system.
+#[derive(Debug)]
+pub struct NotificationSystemSubscriberView {
+    index: SubscriberIndex,
+}
+
+impl NotificationSystemSubscriberView {
+    /// Creates a new NotificationSystemSubscriberView with an empty SubscriberIndex.
+    ///
+    /// Returns a new instance of NotificationSystemSubscriberView.
+    pub fn new() -> Self {
+        Self {
+            index: SubscriberIndex::default(),
+        }
+    }
+
+    /// Checks if a bucket has any subscribers for a specific event.
+    /// This is a quick check using the event mask in the snapshot.
+    ///
+    /// # Arguments
+    /// * `bucket` - The name of the bucket to check.
+    /// * `event` - The event name to check for subscriptions.
+    ///
+    /// Returns `true` if there are subscribers for the event, `false` otherwise.
+    #[inline]
+    pub fn has_subscriber(&self, bucket: &str, event: &EventName) -> bool {
+        self.index.has_subscriber(bucket, event)
+    }
+
+    /// Builds and atomically replaces a bucket's subscription snapshot from the configuration.
+    ///
+    /// Core principle: masks and rules are calculated and stored together in the same update.
+    ///
+    /// # Arguments
+    /// * `bucket` - The name of the bucket to update.
+    /// * `cfg` - The bucket notification configuration to compile into a snapshot.
+    pub fn apply_bucket_config(&self, bucket: &str, cfg: &BucketNotificationConfig) {
+        // *It is recommended to merge compile into one function to ensure the same origin.
+        let snapshot: BucketRulesSnapshot<DynRulesContainer> = cfg.compile_snapshot();
+
+        // *debug to prevent inconsistencies from being introduced when modifying the compile logic in the future.
+        snapshot.debug_assert_mask_consistent();
+
+        self.index.store_snapshot(bucket, snapshot);
+    }
+
+    /// Clears a bucket's subscription snapshot.
+    ///
+    /// #Arguments
+    /// * `bucket` - The name of the bucket to clear.
+    #[inline]
+    pub fn clear_bucket(&self, bucket: &str) {
+        self.index.clear_bucket(bucket);
+    }
+}
--- a/crates/notify/src/notifier.rs
+++ b/crates/notify/src/notifier.rs
@@ -14,19 +14,21 @@

 use crate::{error::NotificationError, event::Event, rules::RulesMap};
 use hashbrown::HashMap;
+use rustfs_config::notify::{DEFAULT_NOTIFY_SEND_CONCURRENCY, ENV_NOTIFY_SEND_CONCURRENCY};
 use rustfs_targets::EventName;
 use rustfs_targets::Target;
 use rustfs_targets::arn::TargetID;
 use rustfs_targets::target::EntityTarget;
 use starshard::AsyncShardedHashMap;
 use std::sync::Arc;
-use tokio::sync::RwLock;
+use tokio::sync::{RwLock, Semaphore};
 use tracing::{debug, error, info, instrument, warn};

 /// Manages event notification to targets based on rules
 pub struct EventNotifier {
    target_list: Arc<RwLock<TargetList>>,
    bucket_rules_map: Arc<AsyncShardedHashMap<String, RulesMap, rustc_hash::FxBuildHasher>>,
+    send_limiter: Arc<Semaphore>,
 }

 impl Default for EventNotifier {
@@ -37,16 +39,41 @@ impl Default for EventNotifier {

 impl EventNotifier {
    /// Creates a new EventNotifier
+    ///
+    /// # Returns
+    /// Returns a new instance of EventNotifier.
    pub fn new() -> Self {
+        let max_inflight = rustfs_utils::get_env_usize(ENV_NOTIFY_SEND_CONCURRENCY, DEFAULT_NOTIFY_SEND_CONCURRENCY);
        EventNotifier {
            target_list: Arc::new(RwLock::new(TargetList::new())),
            bucket_rules_map: Arc::new(AsyncShardedHashMap::new(0)),
+            send_limiter: Arc::new(Semaphore::new(max_inflight)),
        }
    }

+    /// Checks whether a TargetID is still referenced by any bucket's rules.
+    ///
+    /// # Arguments
+    /// * `target_id` - The TargetID to check.
+    ///
+    /// # Returns
+    /// Returns `true` if the TargetID is bound to any bucket, otherwise `false`.
+    pub async fn is_target_bound_to_any_bucket(&self, target_id: &TargetID) -> bool {
+        // `AsyncShardedHashMap::iter()`: Traverse (bucket_name, rules_map)
+        let items = self.bucket_rules_map.iter().await;
+        for (_bucket, rules_map) in items {
+            if rules_map.contains_target_id(target_id) {
+                return true;
+            }
+        }
+        false
+    }
+
    /// Returns a reference to the target list
    /// This method provides access to the target list for external use.
    ///
+    /// # Returns
+    /// Returns an `Arc<RwLock<TargetList>>` representing the target list.
    pub fn target_list(&self) -> Arc<RwLock<TargetList>> {
        Arc::clone(&self.target_list)
    }
@@ -54,17 +81,23 @@ impl EventNotifier {
    /// Removes all notification rules for a bucket
    ///
    /// # Arguments
-    /// * `bucket_name` - The name of the bucket for which to remove rules
+    /// * `bucket` - The name of the bucket for which to remove rules
    ///
    /// This method removes all rules associated with the specified bucket name.
    /// It will log a message indicating the removal of rules.
-    pub async fn remove_rules_map(&self, bucket_name: &str) {
-        if self.bucket_rules_map.remove(&bucket_name.to_string()).await.is_some() {
-            info!("Removed all notification rules for bucket: {}", bucket_name);
+    pub async fn remove_rules_map(&self, bucket: &str) {
+        if self.bucket_rules_map.remove(&bucket.to_string()).await.is_some() {
+            info!("Removed all notification rules for bucket: {}", bucket);
        }
    }

    /// Returns a list of ARNs for the registered targets
+    ///
+    /// # Arguments
+    /// * `region` - The region to use for generating the ARNs
+    ///
+    /// # Returns
+    /// Returns a vector of strings representing the ARNs of the registered targets
    pub async fn get_arn_list(&self, region: &str) -> Vec<String> {
        let target_list_guard = self.target_list.read().await;
        target_list_guard
@@ -75,24 +108,37 @@ impl EventNotifier {
    }

    /// Adds a rules map for a bucket
-    pub async fn add_rules_map(&self, bucket_name: &str, rules_map: RulesMap) {
+    ///
+    /// # Arguments
+    /// * `bucket` - The name of the bucket for which to add the rules map
+    /// * `rules_map` - The rules map to add for the bucket
+    pub async fn add_rules_map(&self, bucket: &str, rules_map: RulesMap) {
        if rules_map.is_empty() {
-            self.bucket_rules_map.remove(&bucket_name.to_string()).await;
+            self.bucket_rules_map.remove(&bucket.to_string()).await;
        } else {
-            self.bucket_rules_map.insert(bucket_name.to_string(), rules_map).await;
+            self.bucket_rules_map.insert(bucket.to_string(), rules_map).await;
        }
-        info!("Added rules for bucket: {}", bucket_name);
+        info!("Added rules for bucket: {}", bucket);
    }

    /// Gets the rules map for a specific bucket.
-    pub async fn get_rules_map(&self, bucket_name: &str) -> Option<RulesMap> {
-        self.bucket_rules_map.get(&bucket_name.to_string()).await
+    ///
+    /// # Arguments
+    /// * `bucket` - The name of the bucket for which to get the rules map
+    ///
+    /// # Returns
+    /// Returns `Some(RulesMap)` if rules exist for the bucket, otherwise returns `None`.
+    pub async fn get_rules_map(&self, bucket: &str) -> Option<RulesMap> {
+        self.bucket_rules_map.get(&bucket.to_string()).await
    }

    /// Removes notification rules for a bucket
-    pub async fn remove_notification(&self, bucket_name: &str) {
-        self.bucket_rules_map.remove(&bucket_name.to_string()).await;
-        info!("Removed notification rules for bucket: {}", bucket_name);
+    ///
+    /// # Arguments
+    /// * `bucket` - The name of the bucket for which to remove notification rules
+    pub async fn remove_notification(&self, bucket: &str) {
+        self.bucket_rules_map.remove(&bucket.to_string()).await;
+        info!("Removed notification rules for bucket: {}", bucket);
    }

    /// Removes all targets
@@ -125,69 +171,87 @@ impl EventNotifier {
    }

    /// Sends an event to the appropriate targets based on the bucket rules
+    ///
+    /// # Arguments
+    /// * `event` - The event to send
    #[instrument(skip_all)]
    pub async fn send(&self, event: Arc<Event>) {
        let bucket_name = &event.s3.bucket.name;
        let object_key = &event.s3.object.key;
        let event_name = event.event_name;
-        if let Some(rules) = self.bucket_rules_map.get(bucket_name).await {
-            let target_ids = rules.match_rules(event_name, object_key);
-            if target_ids.is_empty() {
-                debug!("No matching targets for event in bucket: {}", bucket_name);
-                return;
-            }
-            let target_ids_len = target_ids.len();
-            let mut handles = vec![];

-            // Use scope to limit the borrow scope of target_list
-            {
-                let target_list_guard = self.target_list.read().await;
-                info!("Sending event to targets: {:?}", target_ids);
-                for target_id in target_ids {
-                    // `get` now returns Option<Arc<dyn Target + Send + Sync>>
-                    if let Some(target_arc) = target_list_guard.get(&target_id) {
-                        // Clone an Arc<Box<dyn Target>> (which is where target_list is stored) to move into an asynchronous task
-                        // target_arc is already Arc, clone it for the async task
-                        let cloned_target_for_task = target_arc.clone();
-                        let event_clone = event.clone();
-                        let target_name_for_task = cloned_target_for_task.name(); // Get the name before generating the task
-                        debug!("Preparing to send event to target: {}", target_name_for_task);
-                        // Use cloned data in closures to avoid borrowing conflicts
-                        // Create an EntityTarget from the event
-                        let entity_target: Arc<EntityTarget<Event>> = Arc::new(EntityTarget {
-                            object_name: object_key.to_string(),
-                            bucket_name: bucket_name.to_string(),
-                            event_name,
-                            data: event_clone.clone().as_ref().clone(),
-                        });
-                        let handle = tokio::spawn(async move {
-                            if let Err(e) = cloned_target_for_task.save(entity_target.clone()).await {
-                                error!("Failed to send event to target {}: {}", target_name_for_task, e);
-                            } else {
-                                debug!("Successfully saved event to target {}", target_name_for_task);
-                            }
-                        });
-                        handles.push(handle);
-                    } else {
-                        warn!("Target ID {:?} found in rules but not in target list.", target_id);
-                    }
-                }
-                // target_list is automatically released here
-            }
-
-            // Wait for all tasks to be completed
-            for handle in handles {
-                if let Err(e) = handle.await {
-                    error!("Task for sending/saving event failed: {}", e);
-                }
-            }
-            info!("Event processing initiated for {} targets for bucket: {}", target_ids_len, bucket_name);
-        } else {
+        let Some(rules) = self.bucket_rules_map.get(bucket_name).await else {
            debug!("No rules found for bucket: {}", bucket_name);
+            return;
+        };
+
+        let target_ids = rules.match_rules(event_name, object_key);
+        if target_ids.is_empty() {
+            debug!("No matching targets for event in bucket: {}", bucket_name);
+            return;
        }
+        let target_ids_len = target_ids.len();
+        let mut handles = vec![];
+
+        // Use scope to limit the borrow scope of target_list
+        let target_list_guard = self.target_list.read().await;
+        info!("Sending event to targets: {:?}", target_ids);
+        for target_id in target_ids {
+            // `get` now returns Option<Arc<dyn Target + Send + Sync>>
+            if let Some(target_arc) = target_list_guard.get(&target_id) {
+                // Clone an Arc<Box<dyn Target>> (which is where target_list is stored) to move into an asynchronous task
+                // target_arc is already Arc, clone it for the async task
+                let target_for_task = target_arc.clone();
+                let limiter = self.send_limiter.clone();
+                let event_clone = event.clone();
+                let target_name_for_task = target_for_task.name(); // Get the name before generating the task
+                debug!("Preparing to send event to target: {}", target_name_for_task);
+                // Use cloned data in closures to avoid borrowing conflicts
+                // Create an EntityTarget from the event
+                let entity_target: Arc<EntityTarget<Event>> = Arc::new(EntityTarget {
+                    object_name: object_key.to_string(),
+                    bucket_name: bucket_name.to_string(),
+                    event_name,
+                    data: event_clone.as_ref().clone(),
+                });
+                let handle = tokio::spawn(async move {
+                    let _permit = match limiter.acquire_owned().await {
+                        Ok(p) => p,
+                        Err(e) => {
+                            error!("Failed to acquire send permit for target {}: {}", target_name_for_task, e);
+                            return;
+                        }
+                    };
+                    if let Err(e) = target_for_task.save(entity_target.clone()).await {
+                        error!("Failed to send event to target {}: {}", target_name_for_task, e);
+                    } else {
+                        debug!("Successfully saved event to target {}", target_name_for_task);
+                    }
+                });
+                handles.push(handle);
+            } else {
+                warn!("Target ID {:?} found in rules but not in target list.", target_id);
+            }
+        }
+        // target_list is automatically released here
+        drop(target_list_guard);
+
+        // Wait for all tasks to be completed
+        for handle in handles {
+            if let Err(e) = handle.await {
+                error!("Task for sending/saving event failed: {}", e);
+            }
+        }
+        info!("Event processing initiated for {} targets for bucket: {}", target_ids_len, bucket_name);
    }

    /// Initializes the targets for buckets
+    ///
+    /// # Arguments
+    /// * `targets_to_init` - A vector of boxed targets to initialize
+    ///
+    /// # Returns
+    /// Returns `Ok(())` if initialization is successful, otherwise returns a `NotificationError`.
    #[instrument(skip(self, targets_to_init))]
    pub async fn init_bucket_targets(
        &self,
@@ -218,6 +282,7 @@ impl EventNotifier {

 /// A thread-safe list of targets
 pub struct TargetList {
+    /// Map of TargetID to Target
    targets: HashMap<TargetID, Arc<dyn Target<Event> + Send + Sync>>,
 }

@@ -234,6 +299,12 @@ impl TargetList {
    }

    /// Adds a target to the list
+    ///
+    /// # Arguments
+    /// * `target` - The target to add
+    ///
+    /// # Returns
+    /// Returns `Ok(())` if the target was added successfully, or a `NotificationError` if an error occurred.
    pub fn add(&mut self, target: Arc<dyn Target<Event> + Send + Sync>) -> Result<(), NotificationError> {
        let id = target.id();
        if self.targets.contains_key(&id) {
@@ -251,6 +322,12 @@ impl TargetList {

    /// Removes a target by ID. Note: This does not stop its associated event stream.
    /// Stream cancellation should be handled by EventNotifier.
+    ///
+    /// # Arguments
+    /// * `id` - The ID of the target to remove
+    ///
+    /// # Returns
+    /// Returns the removed target if it existed, otherwise `None`.
    pub async fn remove_target_only(&mut self, id: &TargetID) -> Option<Arc<dyn Target<Event> + Send + Sync>> {
        if let Some(target_arc) = self.targets.remove(id) {
            if let Err(e) = target_arc.close().await {
@@ -278,6 +355,12 @@ impl TargetList {
    }

    /// Returns a target by ID
+    ///
+    /// # Arguments
+    /// * `id` - The ID of the target to retrieve
+    ///
+    /// # Returns
+    /// Returns the target if it exists, otherwise `None`.
    pub fn get(&self, id: &TargetID) -> Option<Arc<dyn Target<Event> + Send + Sync>> {
        self.targets.get(id).cloned()
    }
@@ -292,7 +375,7 @@ impl TargetList {
        self.targets.len()
    }

-    // is_empty can be derived from len()
+    /// is_empty can be derived from len()
    pub fn is_empty(&self) -> bool {
        self.targets.is_empty()
    }
--- a/crates/notify/src/rules/config.rs
+++ b/crates/notify/src/rules/config.rs
@@ -15,13 +15,60 @@
 use super::rules_map::RulesMap;
 use super::xml_config::ParseConfigError as BucketNotificationConfigError;
 use crate::rules::NotificationConfiguration;
-use crate::rules::pattern_rules;
-use crate::rules::target_id_set;
-use hashbrown::HashMap;
+use crate::rules::subscriber_snapshot::{BucketRulesSnapshot, DynRulesContainer, RuleEvents, RulesContainer};
 use rustfs_targets::EventName;
 use rustfs_targets::arn::TargetID;
 use serde::{Deserialize, Serialize};
 use std::io::Read;
+use std::sync::Arc;
+
+/// A "rule view", only used for snapshot mask/consistency verification.
+/// Here we choose to generate the view by "single event" to ensure that event_mask calculation is reliable and simple.
+#[derive(Debug)]
+struct RuleView {
+    events: Vec<EventName>,
+}
+
+impl RuleEvents for RuleView {
+    fn subscribed_events(&self) -> &[EventName] {
+        &self.events
+    }
+}
+
+/// Adapt RulesMap to RulesContainer.
+/// Key point: The items returned by iter_rules are &dyn RuleEvents, so a RuleView list is cached in the container.
+#[derive(Debug)]
+struct CompiledRules {
+    // Keep RulesMap (can be used later if you want to make more complex judgments during the snapshot reading phase)
+    #[allow(dead_code)]
+    rules_map: RulesMap,
+    // for RulesContainer::iter_rules
+    rule_views: Vec<RuleView>,
+}
+
+impl CompiledRules {
+    fn from_rules_map(rules_map: &RulesMap) -> Self {
+        let mut rule_views = Vec::new();
+
+        for ev in rules_map.iter_events() {
+            rule_views.push(RuleView { events: vec![ev] });
+        }
+
+        Self {
+            rules_map: rules_map.clone(),
+            rule_views,
+        }
+    }
+}
+
+impl RulesContainer for CompiledRules {
+    type Rule = dyn RuleEvents;
+
+    fn iter_rules<'a>(&'a self) -> Box<dyn Iterator<Item = &'a Self::Rule> + 'a> {
+        // Key: Convert &RuleView into &dyn RuleEvents
+        Box::new(self.rule_views.iter().map(|v| v as &dyn RuleEvents))
+    }
+}

 /// Configuration for bucket notifications.
 /// This struct now holds the parsed and validated rules in the new RulesMap format.
@@ -119,11 +166,26 @@ impl BucketNotificationConfig {
    pub fn set_region(&mut self, region: &str) {
        self.region = region.to_string();
    }
-}

-// Add a helper to PatternRules if not already present
-impl pattern_rules::PatternRules {
-    pub fn inner(&self) -> &HashMap<String, target_id_set::TargetIdSet> {
-        &self.rules
+    /// Compiles the current BucketNotificationConfig into a BucketRulesSnapshot.
+    /// This involves transforming the rules into a format suitable for runtime use,
+    /// and calculating the event mask based on the subscribed events of the rules.
+    ///
+    /// # Returns
+    /// A BucketRulesSnapshot containing the compiled rules and event mask.
+    pub fn compile_snapshot(&self) -> BucketRulesSnapshot<DynRulesContainer> {
+        // 1) Generate container from RulesMap
+        let compiled = CompiledRules::from_rules_map(self.get_rules_map());
+        let rules: Arc<DynRulesContainer> = Arc::new(compiled) as Arc<DynRulesContainer>;
+
+        // 2) Calculate event_mask
+        let mut mask = 0u64;
+        for rule in rules.iter_rules() {
+            for ev in rule.subscribed_events() {
+                mask |= ev.mask();
+            }
+        }
+
+        BucketRulesSnapshot { event_mask: mask, rules }
    }
 }
--- a/crates/notify/src/rules/mod.rs
+++ b/crates/notify/src/rules/mod.rs
@@ -12,22 +12,24 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.

+mod config;
 pub mod pattern;
-pub mod pattern_rules;
-pub mod rules_map;
-pub mod target_id_set;
+mod pattern_rules;
+mod rules_map;
+mod subscriber_index;
+mod subscriber_snapshot;
+mod target_id_set;
 pub mod xml_config; // For XML structure definition and parsing
-
-pub mod config; // Definition and parsing for BucketNotificationConfig
+// Definition and parsing for BucketNotificationConfig

 // Re-export key types from submodules for easy access to `crate::rules::TypeName`
 // Re-export key types from submodules for external use
 pub use config::BucketNotificationConfig;
 // Assume that BucketNotificationConfigError is also defined in config.rs
 // Or if it is still an alias for xml_config::ParseConfigError , adjust accordingly
-pub use xml_config::ParseConfigError as BucketNotificationConfigError;
-
 pub use pattern_rules::PatternRules;
 pub use rules_map::RulesMap;
+pub use subscriber_index::*;
+pub use subscriber_snapshot::*;
 pub use target_id_set::TargetIdSet;
-pub use xml_config::{NotificationConfiguration, ParseConfigError};
+pub use xml_config::{NotificationConfiguration, ParseConfigError, ParseConfigError as BucketNotificationConfigError};
--- a/crates/notify/src/rules/pattern_rules.rs
+++ b/crates/notify/src/rules/pattern_rules.rs
@@ -12,8 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.

-use super::pattern;
-use super::target_id_set::TargetIdSet;
+use crate::rules::TargetIdSet;
+use crate::rules::pattern;
 use hashbrown::HashMap;
 use rayon::prelude::*;
 use rustfs_targets::arn::TargetID;
@@ -27,31 +27,69 @@ pub struct PatternRules {
 }

 impl PatternRules {
+    /// Create a new, empty PatternRules.
    pub fn new() -> Self {
        Default::default()
    }

    /// Add rules: Pattern and Target ID.
    /// If the schema already exists, add target_id to the existing TargetIdSet.
+    ///
+    /// # Arguments
+    /// * `pattern` - The object name pattern.
+    /// * `target_id` - The TargetID to associate with the pattern.
    pub fn add(&mut self, pattern: String, target_id: TargetID) {
        self.rules.entry(pattern).or_default().insert(target_id);
    }

    /// Checks if there are any rules that match the given object name.
+    ///
+    /// # Arguments
+    /// * `object_name` - The object name to match against the patterns.
+    ///
+    /// # Returns
+    /// `true` if any pattern matches the object name, otherwise `false`.
    pub fn match_simple(&self, object_name: &str) -> bool {
        self.rules.keys().any(|p| pattern::match_simple(p, object_name))
    }

    /// Returns all TargetIDs that match the object name.
+    ///
+    /// Performance optimization points:
+    /// 1) Small collections are serialized directly to avoid rayon scheduling/merging overhead
+    /// 2) When hitting, no longer temporarily allocate TargetIdSet for each rule, but directly extend
+    ///
+    /// # Arguments
+    /// * `object_name` - The object name to match against the patterns.
+    ///
+    /// # Returns
+    /// A TargetIdSet containing all TargetIDs that match the object name.
    pub fn match_targets(&self, object_name: &str) -> TargetIdSet {
+        let n = self.rules.len();
+        if n == 0 {
+            return TargetIdSet::new();
+        }
+
+        // Experience Threshold: Serial is usually faster below this value (can be adjusted after benchmarking)
+        const PAR_THRESHOLD: usize = 128;
+
+        if n < PAR_THRESHOLD {
+            let mut out = TargetIdSet::new();
+            for (pattern_str, target_set) in self.rules.iter() {
+                if pattern::match_simple(pattern_str, object_name) {
+                    out.extend(target_set.iter().cloned());
+                }
+            }
+            return out;
+        }
+        // Parallel path: Each thread accumulates a local set and finally merges it to reduce frequent allocations
        self.rules
            .par_iter()
-            .filter_map(|(pattern_str, target_set)| {
+            .fold(TargetIdSet::new, |mut local, (pattern_str, target_set)| {
                if pattern::match_simple(pattern_str, object_name) {
-                    Some(target_set.iter().cloned().collect::<TargetIdSet>())
-                } else {
-                    None
+                    local.extend(target_set.iter().cloned());
                }
+                local
            })
            .reduce(TargetIdSet::new, |mut acc, set| {
                acc.extend(set);
@@ -65,6 +103,11 @@ impl PatternRules {

    /// Merge another PatternRules.
    /// Corresponding to Go's `Rules.Union`.
+    /// # Arguments
+    /// * `other` - The PatternRules to merge with.
+    ///
+    /// # Returns
+    /// A new PatternRules containing the union of both.
    pub fn union(&self, other: &Self) -> Self {
        let mut new_rules = self.clone();
        for (pattern, their_targets) in &other.rules {
@@ -76,6 +119,13 @@ impl PatternRules {

    /// Calculate the difference from another PatternRules.
    /// Corresponding to Go's `Rules.Difference`.
+    /// The result contains only the patterns and TargetIDs that are in `self` but not in `other`.
+    ///
+    /// # Arguments
+    /// * `other` - The PatternRules to compare against.
+    ///
+    /// # Returns
+    /// A new PatternRules containing the difference.
    pub fn difference(&self, other: &Self) -> Self {
        let mut result_rules = HashMap::new();
        for (pattern, self_targets) in &self.rules {
@@ -94,4 +144,59 @@ impl PatternRules {
        }
        PatternRules { rules: result_rules }
    }
+
+    /// Merge another PatternRules into self in place.
+    /// Corresponding to Go's `Rules.UnionInPlace`.
+    /// # Arguments
+    /// * `other` - The PatternRules to merge with.
+    pub fn union_in_place(&mut self, other: &Self) {
+        for (pattern, their_targets) in &other.rules {
+            self.rules
+                .entry(pattern.clone())
+                .or_default()
+                .extend(their_targets.iter().cloned());
+        }
+    }
+
+    /// Calculate the difference from another PatternRules in place.
+    /// Corresponding to Go's `Rules.DifferenceInPlace`.
+    /// The result contains only the patterns and TargetIDs that are in `self` but not in `other`.
+    /// # Arguments
+    /// * `other` - The PatternRules to compare against.
+    pub fn difference_in_place(&mut self, other: &Self) {
+        self.rules.retain(|pattern, self_targets| {
+            if let Some(other_targets) = other.rules.get(pattern) {
+                // Remove other_targets from self_targets
+                self_targets.retain(|tid| !other_targets.contains(tid));
+            }
+            !self_targets.is_empty()
+        });
+    }
+
+    /// Remove a pattern and its associated TargetID set from the PatternRules.
+    ///
+    /// # Arguments
+    /// * `pattern` - The pattern to remove.
+    pub fn remove_pattern(&mut self, pattern: &str) -> bool {
+        self.rules.remove(pattern).is_some()
+    }
+
+    /// Determine whether the current PatternRules contains the specified TargetID (referenced by any pattern).
+    ///
+    /// # Parameters
+    /// * `target_id` - The TargetID to check for existence within the PatternRules
+    ///
+    /// # Returns
+    /// * `true` if the TargetID exists in any of the patterns; `false` otherwise.
+    pub fn contains_target_id(&self, target_id: &TargetID) -> bool {
+        self.rules.values().any(|set| set.contains(target_id))
+    }
+
+    /// Expose the internal rules for use in scenarios such as BucketNotificationConfig::validate.
+    ///
+    /// # Returns
+    /// A reference to the internal HashMap of patterns to TargetIdSets.
+    pub fn inner(&self) -> &HashMap<String, TargetIdSet> {
+        &self.rules
+    }
 }
--- a/crates/notify/src/rules/rules_map.rs
+++ b/crates/notify/src/rules/rules_map.rs
@@ -12,8 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.

-use super::pattern_rules::PatternRules;
-use super::target_id_set::TargetIdSet;
+use crate::rules::{PatternRules, TargetIdSet};
 use hashbrown::HashMap;
 use rustfs_targets::EventName;
 use rustfs_targets::arn::TargetID;
@@ -31,6 +30,9 @@ pub struct RulesMap {

 impl RulesMap {
    /// Create a new, empty RulesMap.
+    ///
+    /// # Returns
+    /// A new instance of RulesMap with an empty map and a total_events_mask set to 0.
    pub fn new() -> Self {
        Default::default()
    }
@@ -67,12 +69,12 @@ impl RulesMap {

    /// Merge another RulesMap.
    /// `RulesMap.Add(rulesMap2 RulesMap) corresponding to Go
+    ///
+    /// # Parameters
+    /// * `other_map` - The other RulesMap to be merged into the current one.
    pub fn add_map(&mut self, other_map: &Self) {
        for (event_name, other_pattern_rules) in &other_map.map {
-            let self_pattern_rules = self.map.entry(*event_name).or_default();
-            // PatternRules::union Returns the new PatternRules, we need to modify the existing ones
-            let merged_rules = self_pattern_rules.union(other_pattern_rules);
-            *self_pattern_rules = merged_rules;
+            self.map.entry(*event_name).or_default().union_in_place(other_pattern_rules);
        }
        // Directly merge two masks.
        self.total_events_mask |= other_map.total_events_mask;
@@ -81,11 +83,14 @@ impl RulesMap {
    /// Remove another rule defined in the RulesMap from the current RulesMap.
    ///
    /// After the rule is removed, `total_events_mask` is recalculated to ensure its accuracy.
+    ///
+    /// # Parameters
+    /// * `other_map` - The other RulesMap containing rules to be removed from the current one.
    pub fn remove_map(&mut self, other_map: &Self) {
        let mut events_to_remove = Vec::new();
        for (event_name, self_pattern_rules) in &mut self.map {
            if let Some(other_pattern_rules) = other_map.map.get(event_name) {
-                *self_pattern_rules = self_pattern_rules.difference(other_pattern_rules);
+                self_pattern_rules.difference_in_place(other_pattern_rules);
                if self_pattern_rules.is_empty() {
                    events_to_remove.push(*event_name);
                }
@@ -102,6 +107,9 @@ impl RulesMap {
    ///
    /// This method uses a bitmask for a quick check of O(1) complexity.
    /// `event_name` can be a compound type, such as `ObjectCreatedAll`.
+    ///
+    /// # Parameters
+    /// * `event_name` - The event name to check for subscribers.
    pub fn has_subscriber(&self, event_name: &EventName) -> bool {
        // event_name.mask() will handle compound events correctly
        (self.total_events_mask & event_name.mask()) != 0
@@ -112,39 +120,54 @@ impl RulesMap {
    /// # Notice
    /// The `event_name` parameter should be a specific, non-compound event type.
    /// Because this is taken from the `Event` object that actually occurs.
+    ///
+    /// # Parameters
+    /// * `event_name` - The specific event name to match against.
+    /// * `object_key` - The object key to match against the patterns in the rules.
+    ///
+    /// # Returns
+    /// * A set of TargetIDs that match the given event and object key.
    pub fn match_rules(&self, event_name: EventName, object_key: &str) -> TargetIdSet {
        // Use bitmask to quickly determine whether there is a matching rule
        if (self.total_events_mask & event_name.mask()) == 0 {
            return TargetIdSet::new(); // No matching rules
        }

-        // First try to directly match the event name
-        if let Some(pattern_rules) = self.map.get(&event_name) {
-            let targets = pattern_rules.match_targets(object_key);
-            if !targets.is_empty() {
-                return targets;
-            }
-        }
-        // Go's RulesMap[eventName] is directly retrieved, and if it does not exist, it is empty Rules.
-        // Rust's HashMap::get returns Option. If the event name does not exist, there is no rule.
-        // Compound events (such as ObjectCreatedAll) have been expanded as a single event when add_rule_config.
-        // Therefore, a single event name should be used when querying.
-        // If event_name itself is a single type, look it up directly.
-        // If event_name is a compound type, Go's logic is expanded when added.
-        // Here match_rules should receive events that may already be single.
-        // If the caller passes in a compound event, it should expand itself or handle this function first.
-        // Assume that event_name is already a specific event that can be used for searching.
+        // In Go, RulesMap[eventName] returns empty rules if the key doesn't exist.
+        // Rust's HashMap::get returns Option, so missing key means no rules.
+        // Compound events like ObjectCreatedAll are expanded into specific events during add_rule_config.
+        // Thus, queries should use specific event names.
+        // If event_name is compound, expansion happens at addition time.
+        // match_rules assumes event_name is already a specific event for lookup.
+        // Callers should expand compound events before calling this method.
        self.map
            .get(&event_name)
            .map_or_else(TargetIdSet::new, |pr| pr.match_targets(object_key))
    }

    /// Check if RulesMap is empty.
+    ///
+    /// # Returns
+    /// * `true` if there are no rules in the map; `false` otherwise
    pub fn is_empty(&self) -> bool {
        self.map.is_empty()
    }

+    /// Determine whether the current RulesMap contains the specified TargetID (referenced by any event / pattern).
+    ///
+    /// # Parameters
+    /// * `target_id` - The TargetID to check for existence within the RulesMap
+    ///
+    /// # Returns
+    /// * `true` if the TargetID exists in any of the PatternRules; `false` otherwise.
+    pub fn contains_target_id(&self, target_id: &TargetID) -> bool {
+        self.map.values().any(|pr| pr.contains_target_id(target_id))
+    }
+
    /// Returns a clone of internal rules for use in scenarios such as BucketNotificationConfig::validate.
+    ///
+    /// # Returns
+    /// A reference to the internal HashMap of EventName to PatternRules.
    pub fn inner(&self) -> &HashMap<EventName, PatternRules> {
        &self.map
    }
@@ -160,18 +183,32 @@ impl RulesMap {
    }

    /// Remove rules and optimize performance
+    ///
+    /// # Parameters
+    /// * `event_name` - The EventName from which to remove the rule.
+    /// * `pattern` - The pattern of the rule to be removed.
    #[allow(dead_code)]
    pub fn remove_rule(&mut self, event_name: &EventName, pattern: &str) {
+        let mut remove_event = false;
+
        if let Some(pattern_rules) = self.map.get_mut(event_name) {
-            pattern_rules.rules.remove(pattern);
+            pattern_rules.remove_pattern(pattern);
            if pattern_rules.is_empty() {
-                self.map.remove(event_name);
+                remove_event = true;
            }
        }
+
+        if remove_event {
+            self.map.remove(event_name);
+        }
+
        self.recalculate_mask(); // Delay calculation mask
    }

-    /// Batch Delete Rules
+    /// Batch Delete Rules and Optimize Performance
+    ///
+    /// # Parameters
+    /// * `event_names` - A slice of EventNames to be removed.
    #[allow(dead_code)]
    pub fn remove_rules(&mut self, event_names: &[EventName]) {
        for event_name in event_names {
@@ -181,9 +218,27 @@ impl RulesMap {
    }

    /// Update rules and optimize performance
+    ///
+    /// # Parameters
+    /// * `event_name` - The EventName to update.
+    /// * `pattern` - The pattern of the rule to be updated.
+    /// * `target_id` - The TargetID to be added.
    #[allow(dead_code)]
    pub fn update_rule(&mut self, event_name: EventName, pattern: String, target_id: TargetID) {
        self.map.entry(event_name).or_default().add(pattern, target_id);
        self.total_events_mask |= event_name.mask(); // Update only the relevant bitmask
    }
+
+    /// Iterate all EventName keys contained in this RulesMap.
+    ///
+    /// Used by snapshot compilation to compute bucket event_mask.
+    ///
+    /// # Returns
+    /// An iterator over all EventName keys in the RulesMap.
+    #[inline]
+    pub fn iter_events(&self) -> impl Iterator<Item = EventName> + '_ {
+        // `inner()` is already used by config.rs, so we reuse it here.
+        // If the key type is `EventName`, `.copied()` is the cheapest way to return values.
+        self.inner().keys().copied()
+    }
 }
--- a/crates/notify/src/rules/subscriber_index.rs
+++ b/crates/notify/src/rules/subscriber_index.rs
@@ -0,0 +1,131 @@
+// Copyright 2024 RustFS Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use crate::rules::{BucketRulesSnapshot, BucketSnapshotRef, DynRulesContainer};
+use arc_swap::ArcSwap;
+use rustfs_targets::EventName;
+use starshard::ShardedHashMap;
+use std::fmt;
+use std::sync::Arc;
+
+/// A global bucket -> snapshot index.
+///
+/// Read path: lock-free load (ArcSwap)
+/// Write path: atomic replacement after building a new snapshot
+pub struct SubscriberIndex {
+    // Use starshard for sharding to reduce lock competition when the number of buckets is large
+    inner: ShardedHashMap<String, Arc<ArcSwap<BucketRulesSnapshot<DynRulesContainer>>>>,
+    // Cache an "empty rule container" for empty snapshots (avoids building every time)
+    empty_rules: Arc<DynRulesContainer>,
+}
+
+/// Avoid deriving fields that do not support Debug
+impl fmt::Debug for SubscriberIndex {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        f.debug_struct("SubscriberIndex").finish_non_exhaustive()
+    }
+}
+
+impl SubscriberIndex {
+    /// Create a new SubscriberIndex.
+    ///
+    /// # Arguments
+    /// * `empty_rules` - An Arc to an empty rules container used for empty snapshots
+    ///
+    /// Returns a new instance of SubscriberIndex.
+    pub fn new(empty_rules: Arc<DynRulesContainer>) -> Self {
+        Self {
+            inner: ShardedHashMap::new(64),
+            empty_rules,
+        }
+    }
+
+    /// Get the current snapshot of a bucket.
+    /// If it does not exist, return empty snapshot.
+    ///
+    /// # Arguments
+    /// * `bucket` - The name of the bucket to load.
+    ///
+    /// Returns the snapshot reference for the specified bucket.
+    pub fn load_snapshot(&self, bucket: &str) -> BucketSnapshotRef {
+        match self.inner.get(&bucket.to_string()) {
+            Some(cell) => cell.load_full(),
+            None => Arc::new(BucketRulesSnapshot::empty(self.empty_rules.clone())),
+        }
+    }
+
+    /// Quickly determine whether the bucket has a subscription to an event.
+    /// This judgment can be consistent with subsequent rule matching when reading the same snapshot.
+    ///
+    /// # Arguments
+    /// * `bucket` - The name of the bucket to check.
+    /// * `event` - The event name to check for subscriptions.
+    ///
+    /// Returns `true` if there are subscribers for the event, `false` otherwise.
+    #[inline]
+    pub fn has_subscriber(&self, bucket: &str, event: &EventName) -> bool {
+        let snap = self.load_snapshot(bucket);
+        if snap.event_mask == 0 {
+            return false;
+        }
+        snap.has_event(event)
+    }
+
+    /// Atomically update a bucket's snapshot (whole package replacement).
+    ///
+    /// - The caller first builds the complete `BucketRulesSnapshot` (including event\_mask and rules).
+    /// - This method ensures that the read path will not observe intermediate states.
+    ///
+    /// # Arguments
+    /// * `bucket` - The name of the bucket to update.
+    /// * `new_snapshot` - The new snapshot to store for the bucket.
+    pub fn store_snapshot(&self, bucket: &str, new_snapshot: BucketRulesSnapshot<DynRulesContainer>) {
+        let key = bucket.to_string();
+
+        let cell = self.inner.get(&key).unwrap_or_else(|| {
+            // Insert a default cell (empty snapshot)
+            let init = Arc::new(ArcSwap::from_pointee(BucketRulesSnapshot::empty(self.empty_rules.clone())));
+            self.inner.insert(key.clone(), init.clone());
+            init
+        });
+
+        cell.store(Arc::new(new_snapshot));
+    }
+
+    /// Delete the bucket's subscription view (make it empty).
+    ///
+    /// # Arguments
+    /// * `bucket` - The name of the bucket to clear.
+    pub fn clear_bucket(&self, bucket: &str) {
+        if let Some(cell) = self.inner.get(&bucket.to_string()) {
+            cell.store(Arc::new(BucketRulesSnapshot::empty(self.empty_rules.clone())));
+        }
+    }
+}
+
+impl Default for SubscriberIndex {
+    fn default() -> Self {
+        // An available empty rule container is required; here it is implemented using minimal empty
+        #[derive(Debug)]
+        struct EmptyRules;
+        impl crate::rules::subscriber_snapshot::RulesContainer for EmptyRules {
+            type Rule = dyn crate::rules::subscriber_snapshot::RuleEvents;
+            fn iter_rules<'a>(&'a self) -> Box<dyn Iterator<Item = &'a Self::Rule> + 'a> {
+                Box::new(std::iter::empty())
+            }
+        }
+
+        Self::new(Arc::new(EmptyRules) as Arc<DynRulesContainer>)
+    }
+}
--- a/crates/notify/src/rules/subscriber_snapshot.rs
+++ b/crates/notify/src/rules/subscriber_snapshot.rs
@@ -0,0 +1,117 @@
+// Copyright 2024 RustFS Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use rustfs_targets::EventName;
+use std::sync::Arc;
+
+/// Let the rules structure provide "what events it is subscribed to".
+/// This way BucketRulesSnapshot does not need to know the internal shape of rules.
+pub trait RuleEvents {
+    fn subscribed_events(&self) -> &[EventName];
+}
+
+/// Let the rules container provide the ability to iterate over all rules (abstracting only to the minimum necessary).
+pub trait RulesContainer {
+    type Rule: RuleEvents + ?Sized;
+    fn iter_rules<'a>(&'a self) -> Box<dyn Iterator<Item = &'a Self::Rule> + 'a>;
+
+    /// Fast empty judgment for snapshots (fix missing `rules.is_empty()`)
+    fn is_empty(&self) -> bool {
+        self.iter_rules().next().is_none()
+    }
+}
+
+/// Represents a bucket's notification subscription view snapshot (immutable).
+///
+/// - `event_mask`: Quickly determine whether there is a subscription to a certain type of event (bitset/flags).
+/// - `rules`: precise rule mapping (prefix/suffix/pattern -> targets).
+///
+/// The read path only reads this snapshot to ensure consistency.
+#[derive(Debug, Clone)]
+pub struct BucketRulesSnapshot<R>
+where
+    R: RulesContainer + ?Sized,
+{
+    pub event_mask: u64,
+    pub rules: Arc<R>,
+}
+
+impl<R> BucketRulesSnapshot<R>
+where
+    R: RulesContainer + ?Sized,
+{
+    /// Create an empty snapshot with no subscribed events and no rules.
+    ///
+    /// # Arguments
+    /// * `rules` - An Arc to a rules container (can be an empty container).
+    ///
+    /// # Returns
+    /// An instance of `BucketRulesSnapshot` with an empty event mask.
+    #[inline]
+    pub fn empty(rules: Arc<R>) -> Self {
+        Self { event_mask: 0, rules }
+    }
+
+    /// Check if the snapshot has any subscribers for the specified event.
+    ///
+    /// # Arguments
+    /// * `event` - The event name to check for subscriptions.
+    ///
+    /// # Returns
+    /// `true` if there are subscribers for the event, `false` otherwise.
+    #[inline]
+    pub fn has_event(&self, event: &EventName) -> bool {
+        (self.event_mask & event.mask()) != 0
+    }
+
+    /// Check if the snapshot is empty (no subscribed events or rules).
+    ///
+    /// # Returns
+    /// `true` if the snapshot is empty, `false` otherwise.
+    #[inline]
+    pub fn is_empty(&self) -> bool {
+        self.event_mask == 0 || self.rules.is_empty()
+    }
+
+    /// [debug] Assert that `event_mask` is consistent with the event declared in `rules`.
+    ///
+    /// Constraints:
+    /// - only runs in debug builds (release incurs no cost).
+    /// - If the rule contains compound events (\*All / Everything), rely on `EventName::mask()` to automatically expand.
+    #[inline]
+    pub fn debug_assert_mask_consistent(&self) {
+        #[cfg(debug_assertions)]
+        {
+            let mut recomputed = 0u64;
+            for rule in self.rules.iter_rules() {
+                for ev in rule.subscribed_events() {
+                    recomputed |= ev.mask();
+                }
+            }
+
+            debug_assert!(
+                recomputed == self.event_mask,
+                "BucketRulesSnapshot.event_mask inconsistent: stored={:#x}, recomputed={:#x}",
+                self.event_mask,
+                recomputed
+            );
+        }
+    }
+}
+
+/// Unify trait-object snapshot types (fix Sized / missing generic arguments)
+pub type DynRulesContainer = dyn RulesContainer<Rule = dyn RuleEvents> + Send + Sync;
+
+/// Expose Arc form to facilitate sharing.
+pub type BucketSnapshotRef = Arc<BucketRulesSnapshot<DynRulesContainer>>;
--- a/crates/notify/src/rules/xml_config.rs
+++ b/crates/notify/src/rules/xml_config.rs
@@ -12,7 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.

-use super::pattern;
+use crate::rules::pattern;
 use hashbrown::HashSet;
 use rustfs_targets::EventName;
 use rustfs_targets::arn::{ARN, ArnError, TargetIDError};
--- a/crates/notify/src/stream.rs
+++ b/crates/notify/src/stream.rs
@@ -13,18 +13,23 @@
 // limitations under the License.

 use crate::{Event, integration::NotificationMetrics};
-use rustfs_targets::StoreError;
-use rustfs_targets::Target;
-use rustfs_targets::TargetError;
-use rustfs_targets::store::{Key, Store};
-use rustfs_targets::target::EntityTarget;
+use rustfs_targets::{
+    StoreError, Target, TargetError,
+    store::{Key, Store},
+    target::EntityTarget,
+};
 use std::sync::Arc;
 use std::time::{Duration, Instant};
 use tokio::sync::{Semaphore, mpsc};
 use tokio::time::sleep;
 use tracing::{debug, error, info, warn};

-/// Streams events from the store to the target
+/// Streams events from the store to the target with retry logic
+///
+/// # Arguments
+/// - `store`: The event store
+/// - `target`: The target to send events to
+/// - `cancel_rx`: Receiver to listen for cancellation signals
 pub async fn stream_events(
    store: &mut (dyn Store<Event, Error = StoreError, Key = Key> + Send),
    target: &dyn Target<Event>,
@@ -67,6 +72,7 @@ pub async fn stream_events(
                match target.send_from_store(key.clone()).await {
                    Ok(_) => {
                        info!("Successfully sent event for target: {}", target.name());
+                        // send_from_store deletes the event from store on success
                        success = true;
                    }
                    Err(e) => {
@@ -104,6 +110,13 @@ pub async fn stream_events(
 }

 /// Starts the event streaming process for a target
+///
+/// # Arguments
+/// - `store`: The event store
+/// - `target`: The target to send events to
+///
+/// # Returns
+/// A sender to signal cancellation of the event stream
 pub fn start_event_stream(
    mut store: Box<dyn Store<Event, Error = StoreError, Key = Key> + Send>,
    target: Arc<dyn Target<Event> + Send + Sync>,
@@ -119,6 +132,15 @@ pub fn start_event_stream(
 }

 /// Start event stream with batch processing
+///
+/// # Arguments
+/// - `store`: The event store
+/// - `target`: The target to send events to clients
+/// - `metrics`: Metrics for monitoring
+/// - `semaphore`: Semaphore to limit concurrency
+///
+/// # Returns
+/// A sender to signal cancellation of the event stream
 pub fn start_event_stream_with_batching(
    mut store: Box<dyn Store<EntityTarget<Event>, Error = StoreError, Key = Key> + Send>,
    target: Arc<dyn Target<Event> + Send + Sync>,
@@ -136,6 +158,16 @@ pub fn start_event_stream_with_batching(
 }

 /// Event stream processing with batch processing
+///
+/// # Arguments
+/// - `store`: The event store
+/// - `target`: The target to send events to clients
+/// - `cancel_rx`: Receiver to listen for cancellation signals
+/// - `metrics`: Metrics for monitoring
+/// - `semaphore`: Semaphore to limit concurrency
+///
+/// # Notes
+/// This function processes events in batches to improve efficiency.
 pub async fn stream_events_with_batching(
    store: &mut (dyn Store<EntityTarget<Event>, Error = StoreError, Key = Key> + Send),
    target: &dyn Target<Event>,
@@ -231,7 +263,17 @@ pub async fn stream_events_with_batching(
    }
 }

-/// Processing event batches
+/// Processing event batches for targets
+/// # Arguments
+/// - `batch`: The batch of events to process
+/// - `batch_keys`: The corresponding keys of the events in the batch
+/// - `target`: The target to send events to clients
+/// - `max_retries`: Maximum number of retries for sending an event
+/// - `base_delay`: Base delay duration for retries
+/// - `metrics`: Metrics for monitoring
+/// - `semaphore`: Semaphore to limit concurrency
+/// # Notes
+/// This function processes a batch of events, sending each event to the target with retry
 async fn process_batch(
    batch: &mut Vec<EntityTarget<Event>>,
    batch_keys: &mut Vec<Key>,
@@ -262,6 +304,7 @@ async fn process_batch(

        // Retry logic
        while retry_count < max_retries && !success {
+            // After sending successfully, the event in the storage is deleted synchronously.
            match target.send_from_store(key.clone()).await {
                Ok(_) => {
                    info!("Successfully sent event for target: {}, Key: {}", target.name(), key.to_string());
--- a/crates/obs/src/telemetry.rs
+++ b/crates/obs/src/telemetry.rs
@@ -39,9 +39,9 @@ use rustfs_config::{
        ENV_OBS_LOG_DIRECTORY, ENV_OBS_LOG_FLUSH_MS, ENV_OBS_LOG_MESSAGE_CAPA, ENV_OBS_LOG_POOL_CAPA,
    },
 };
-use rustfs_utils::{get_env_u64, get_env_usize, get_local_ip_with_default};
+use rustfs_utils::{get_env_opt_str, get_env_u64, get_env_usize, get_local_ip_with_default};
 use smallvec::SmallVec;
-use std::{borrow::Cow, env, fs, io::IsTerminal, time::Duration};
+use std::{borrow::Cow, fs, io::IsTerminal, time::Duration};
 use tracing::info;
 use tracing_error::ErrorLayer;
 use tracing_opentelemetry::{MetricsLayer, OpenTelemetryLayer};
@@ -574,8 +574,8 @@ pub(crate) fn init_telemetry(config: &OtelConfig) -> Result<OtelGuard, Telemetry
    }

    // Rule 2: The user has explicitly customized the log directory (determined by whether ENV_OBS_LOG_DIRECTORY is set)
-    let user_set_log_dir = env::var(ENV_OBS_LOG_DIRECTORY).is_ok();
-    if user_set_log_dir {
+    let user_set_log_dir = get_env_opt_str(ENV_OBS_LOG_DIRECTORY);
+    if user_set_log_dir.filter(|d| !d.is_empty()).is_some() {
        return init_file_logging(config, logger_level, is_production);
    }

--- a/crates/policy/src/error.rs
+++ b/crates/policy/src/error.rs
@@ -89,6 +89,7 @@ pub enum Error {

    #[error("invalid access_key")]
    InvalidAccessKey,
+
    #[error("action not allowed")]
    IAMActionNotAllowed,

@@ -106,6 +107,9 @@ pub enum Error {

    #[error("io error: {0}")]
    Io(std::io::Error),
+
+    #[error("system already initialized")]
+    IamSysAlreadyInitialized,
 }

 impl Error {
--- a/crates/protos/src/lib.rs
+++ b/crates/protos/src/lib.rs
@@ -16,7 +16,7 @@
 mod generated;

 use proto_gen::node_service::node_service_client::NodeServiceClient;
-use rustfs_common::globals::{GLOBAL_CONN_MAP, GLOBAL_ROOT_CERT, evict_connection};
+use rustfs_common::{GLOBAL_CONN_MAP, GLOBAL_ROOT_CERT, evict_connection};
 use std::{error::Error, time::Duration};
 use tonic::{
    Request, Status,
@@ -26,6 +26,11 @@ use tonic::{
 };
 use tracing::{debug, warn};

+// Type alias for the complex client type
+pub type NodeServiceClientType = NodeServiceClient<
+    InterceptedService<Channel, Box<dyn Fn(Request<()>) -> Result<Request<()>, Status> + Send + Sync + 'static>>,
+>;
+
 pub use generated::*;

 // Default 100 MB
--- a/crates/s3select-api/src/lib.rs
+++ b/crates/s3select-api/src/lib.rs
@@ -12,10 +12,9 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.

-use std::fmt::Display;
-
 use datafusion::{common::DataFusionError, sql::sqlparser::parser::ParserError};
 use snafu::{Backtrace, Location, Snafu};
+use std::fmt::Display;

 pub mod object_store;
 pub mod query;
--- a/crates/s3select-api/src/query/datasource/mod.rs
+++ b/crates/s3select-api/src/query/datasource/mod.rs
@@ -1,13 +0,0 @@
-// Copyright 2024 RustFS Team
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
--- a/crates/s3select-api/src/query/mod.rs
+++ b/crates/s3select-api/src/query/mod.rs
@@ -12,13 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.

-use std::sync::Arc;
-
 use s3s::dto::SelectObjectContentInput;
+use std::sync::Arc;

 pub mod analyzer;
 pub mod ast;
-pub mod datasource;
 pub mod dispatcher;
 pub mod execution;
 pub mod function;
--- a/crates/s3select-api/src/query/session.rs
+++ b/crates/s3select-api/src/query/session.rs
@@ -12,20 +12,17 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.

-use std::sync::Arc;
-
+use crate::query::Context;
+use crate::{QueryError, QueryResult, object_store::EcObjectStore};
 use datafusion::{
    execution::{SessionStateBuilder, context::SessionState, runtime_env::RuntimeEnvBuilder},
    parquet::data_type::AsBytes,
    prelude::SessionContext,
 };
 use object_store::{ObjectStore, memory::InMemory, path::Path};
+use std::sync::Arc;
 use tracing::error;

-use crate::{QueryError, QueryResult, object_store::EcObjectStore};
-
-use super::Context;
-
 #[derive(Clone)]
 pub struct SessionCtx {
    _desc: Arc<SessionCtxDesc>,
--- a/crates/utils/Cargo.toml
+++ b/crates/utils/Cargo.toml
@@ -65,6 +65,7 @@ tracing = { workspace = true }
 transform-stream = { workspace = true, optional = true }
 url = { workspace = true, optional = true }
 zstd = { workspace = true, optional = true }
+ipnet = { workspace = true, optional = true }

 [dev-dependencies]
 tempfile = { workspace = true }
@@ -92,5 +93,5 @@ hash = ["dep:highway", "dep:md-5", "dep:sha2", "dep:blake3", "dep:serde", "dep:s
 os = ["dep:nix", "dep:tempfile", "winapi"]  # operating system utilities
 integration = []  # integration test features
 sys = ["dep:sysinfo"]  # system information features
-http = ["dep:convert_case", "dep:http", "dep:regex"]
+http = ["dep:convert_case", "dep:http", "dep:regex", "dep:ipnet"]  # http utilities
 full = ["ip", "tls", "net", "io", "hash", "os", "integration", "path", "crypto", "string", "compress", "sys", "notify", "http"]  # all features
--- a/crates/utils/src/http/ip.rs
+++ b/crates/utils/src/http/ip.rs
@@ -13,9 +13,10 @@
 // limitations under the License.

 use http::HeaderMap;
+use ipnet::{IpNet, Ipv4Net, Ipv6Net};
 use regex::Regex;
 use std::env;
-use std::net::SocketAddr;
+use std::net::{IpAddr, Ipv4Addr, Ipv6Addr, SocketAddr};
 use std::str::FromStr;
 use std::sync::LazyLock;

@@ -45,6 +46,100 @@ fn is_xff_header_enabled() -> bool {
        == "on"
 }

+/// TrustedProxies holds configuration for validating proxy sources
+#[derive(Debug, Clone)]
+pub struct TrustedProxies {
+    /// List of trusted proxy IP networks (CIDR format)
+    pub cidrs: Vec<IpNet>,
+    /// Whether to enable proxy validation
+    pub enable_validation: bool,
+    /// Maximum allowed proxy chain length
+    pub max_chain_length: usize,
+}
+
+impl TrustedProxies {
+    /// Create a new TrustedProxies configuration
+    pub fn new(cidrs: Vec<String>, enable_validation: bool, max_chain_length: usize) -> Self {
+        let cidrs = cidrs.into_iter().filter_map(|s| s.parse::<IpNet>().ok()).collect();
+        Self {
+            cidrs,
+            enable_validation,
+            max_chain_length,
+        }
+    }
+
+    /// Check if an IP address is within the trusted proxy ranges
+    pub fn is_trusted_proxy(&self, ip: IpAddr) -> bool {
+        if !self.enable_validation {
+            return true; // Backward compatibility: trust all when disabled
+        }
+        self.cidrs.iter().any(|net| net.contains(&ip))
+    }
+}
+
+impl Default for TrustedProxies {
+    fn default() -> Self {
+        Self {
+            cidrs: vec![],
+            enable_validation: true,
+            max_chain_length: 10,
+        }
+    }
+}
+
+/// Validate if an IP string represents a valid client IP
+/// Returns false for private/loopback addresses and invalid formats
+fn is_valid_client_ip(ip_str: &str, max_chain_length: usize) -> bool {
+    // Handle X-Forwarded-For chains
+    if ip_str.contains(',') {
+        let parts: Vec<&str> = ip_str.split(',').map(|s| s.trim()).collect();
+        if parts.len() > max_chain_length {
+            return false;
+        }
+        // Validate each IP in the chain
+        for part in parts {
+            if !is_valid_single_ip(part) {
+                return false;
+            }
+        }
+        return true;
+    }
+
+    is_valid_single_ip(ip_str)
+}
+
+/// Validate a single IP address string
+fn is_valid_single_ip(ip_str: &str) -> bool {
+    match ip_str.parse::<IpAddr>() {
+        Ok(ip) => {
+            // Reject private and loopback addresses as client IPs
+            // (they should come from trusted proxies only)
+            !is_private(ip) && !ip.is_loopback()
+        }
+        Err(_) => false,
+    }
+}
+
+/// Check if an IP address is private
+///
+/// # Arguments
+/// * `ip` - The IP address to check
+///
+/// # Returns
+/// A `bool` indicating whether the IP is private
+///
+
+fn is_private(ip: IpAddr) -> bool {
+    match ip {
+        IpAddr::V4(ipv4) => ipv4.is_private(),
+        IpAddr::V6(ipv6) => {
+            // Check if it's in fc00::/7 (Unique Local Address)
+            let octets = ipv6.octets();
+            (octets[0] & 0xfe) == 0xfc
+        }
+    }
+}
+
 /// GetSourceScheme retrieves the scheme from the X-Forwarded-Proto and RFC7239
 /// Forwarded headers (in that order).
 ///
@@ -147,18 +242,43 @@ pub fn get_source_ip_from_headers(headers: &HeaderMap) -> Option<String> {
    addr
 }

-/// GetSourceIPRaw retrieves the IP from the request headers
-/// and falls back to remote_addr when necessary.
-/// however returns without bracketing.
+/// GetSourceIPRaw retrieves the IP from the request headers with trusted proxy validation
+/// and falls back to peer_addr when necessary.
 ///
 /// # Arguments
 /// * `headers` - HTTP headers from the request
-/// * `remote_addr` - Remote address as a string
+/// * `peer_addr` - Peer IP address from the connection
+/// * `trusted_proxies` - Trusted proxy configuration
 ///
 /// # Returns
-/// A `String` containing the source IP address
+/// A `String` containing the validated source IP address
 ///
-pub fn get_source_ip_raw(headers: &HeaderMap, remote_addr: &str) -> String {
+pub fn get_source_ip_raw(headers: &HeaderMap, peer_addr: IpAddr, trusted_proxies: &TrustedProxies) -> String {
+    // If validation is disabled, use legacy behavior for backward compatibility
+    if !trusted_proxies.enable_validation {
+        let remote_addr_str = peer_addr.to_string();
+        return get_source_ip_raw_legacy(headers, &remote_addr_str);
+    }
+
+    // Check if the direct connection is from a trusted proxy
+    if trusted_proxies.is_trusted_proxy(peer_addr) {
+        // Trusted proxy: try to get real client IP from headers
+        if let Some(header_ip) = get_source_ip_from_headers(headers) {
+            // Validate the IP from headers
+            if is_valid_client_ip(&header_ip, trusted_proxies.max_chain_length) {
+                return header_ip;
+            }
+            // If header IP is invalid, log warning and fall back to peer
+            tracing::warn!("Invalid client IP in headers from trusted proxy {}: {}", peer_addr, header_ip);
+        }
+    }
+
+    // Untrusted source or no valid header IP: use connection peer address
+    peer_addr.to_string()
+}
+
+/// Legacy GetSourceIPRaw for backward compatibility when validation is disabled
+fn get_source_ip_raw_legacy(headers: &HeaderMap, remote_addr: &str) -> String {
    let addr = get_source_ip_from_headers(headers).unwrap_or_else(|| remote_addr.to_string());

    // Default to remote address if headers not set.
@@ -169,19 +289,20 @@ pub fn get_source_ip_raw(headers: &HeaderMap, remote_addr: &str) -> String {
    }
 }

-/// GetSourceIP retrieves the IP from the request headers
-/// and falls back to remote_addr when necessary.
+/// GetSourceIP retrieves the IP from the request headers with trusted proxy validation
+/// and falls back to peer_addr when necessary.
 /// It brackets IPv6 addresses.
 ///
 /// # Arguments
 /// * `headers` - HTTP headers from the request
-/// * `remote_addr` - Remote address as a string
+/// * `peer_addr` - Peer IP address from the connection
+/// * `trusted_proxies` - Trusted proxy configuration
 ///
 /// # Returns
 /// A `String` containing the source IP address, with IPv6 addresses bracketed
 ///
-pub fn get_source_ip(headers: &HeaderMap, remote_addr: &str) -> String {
-    let addr = get_source_ip_raw(headers, remote_addr);
+pub fn get_source_ip(headers: &HeaderMap, peer_addr: IpAddr, trusted_proxies: &TrustedProxies) -> String {
+    let addr = get_source_ip_raw(headers, peer_addr, trusted_proxies);
    if addr.contains(':') { format!("[{addr}]") } else { addr }
 }

@@ -210,18 +331,58 @@ mod tests {
    }

    #[test]
-    fn test_get_source_ip_raw() {
+    fn test_trusted_proxies_validation() {
+        let trusted_proxies = TrustedProxies::new(vec!["192.168.1.0/24".to_string(), "10.0.0.0/8".to_string()], true, 5);
+
+        // Trusted IPs
+        assert!(trusted_proxies.is_trusted_proxy("192.168.1.1".parse().unwrap()));
+        assert!(trusted_proxies.is_trusted_proxy("10.1.1.1".parse().unwrap()));
+
+        // Untrusted IPs
+        assert!(!trusted_proxies.is_trusted_proxy("203.0.113.1".parse().unwrap()));
+    }
+
+    #[test]
+    fn test_get_source_ip_raw_with_trusted_proxy() {
+        let mut headers = HeaderMap::new();
+        headers.insert("x-forwarded-for", HeaderValue::from_static("203.0.113.1"));
+
+        let trusted_proxies = TrustedProxies::new(vec!["192.168.1.1/32".to_string()], true, 5);
+        let peer_addr: IpAddr = "192.168.1.1".parse().unwrap();
+
+        let result = get_source_ip_raw(&headers, peer_addr, &trusted_proxies);
+        assert_eq!(result, "203.0.113.1");
+    }
+
+    #[test]
+    fn test_get_source_ip_raw_with_untrusted_proxy() {
+        let mut headers = HeaderMap::new();
+        headers.insert("x-forwarded-for", HeaderValue::from_static("203.0.113.1"));
+
+        let trusted_proxies = TrustedProxies::new(vec![], true, 5);
+        let peer_addr: IpAddr = "203.0.113.2".parse().unwrap();
+
+        let result = get_source_ip_raw(&headers, peer_addr, &trusted_proxies);
+        assert_eq!(result, "203.0.113.2"); // Should use peer_addr
+    }
+
+    #[test]
+    fn test_get_source_ip_raw_legacy_mode() {
        let headers = create_test_headers();
-        let remote_addr = "127.0.0.1:8080";
-        let result = get_source_ip_raw(&headers, remote_addr);
-        assert_eq!(result, "192.168.1.1");
+        let trusted_proxies = TrustedProxies::new(vec![], false, 5); // Disabled validation
+        let peer_addr: IpAddr = "127.0.0.1".parse().unwrap();
+
+        let result = get_source_ip_raw(&headers, peer_addr, &trusted_proxies);
+        assert_eq!(result, "192.168.1.1"); // Should use header IP
    }

    #[test]
    fn test_get_source_ip() {
        let headers = create_test_headers();
-        let remote_addr = "127.0.0.1:8080";
-        let result = get_source_ip(&headers, remote_addr);
+        let trusted_proxies = TrustedProxies::new(vec!["192.168.1.1/32".to_string()], true, 5);
+        let peer_addr: IpAddr = "192.168.1.1".parse().unwrap();
+
+        let result = get_source_ip(&headers, peer_addr, &trusted_proxies);
        assert_eq!(result, "192.168.1.1");
    }

@@ -229,8 +390,32 @@ mod tests {
    fn test_get_source_ip_ipv6() {
        let mut headers = HeaderMap::new();
        headers.insert("x-forwarded-for", HeaderValue::from_static("2001:db8::1"));
-        let remote_addr = "127.0.0.1:8080";
-        let result = get_source_ip(&headers, remote_addr);
+
+        let trusted_proxies = TrustedProxies::new(vec!["192.168.1.1/32".to_string()], true, 5);
+        let peer_addr: IpAddr = "192.168.1.1".parse().unwrap();
+
+        let result = get_source_ip(&headers, peer_addr, &trusted_proxies);
        assert_eq!(result, "[2001:db8::1]");
    }
+
+    #[test]
+    fn test_is_valid_client_ip() {
+        // Valid public IPs
+        assert!(is_valid_client_ip("203.0.113.1", 5));
+        assert!(is_valid_client_ip("2001:db8::1", 5));
+
+        // Invalid private IPs
+        assert!(!is_valid_client_ip("192.168.1.1", 5));
+        assert!(!is_valid_client_ip("10.0.0.1", 5));
+        assert!(!is_valid_client_ip("127.0.0.1", 5));
+
+        // Valid chain
+        assert!(is_valid_client_ip("203.0.113.1, 198.51.100.1", 5));
+
+        // Invalid chain (too long)
+        assert!(!is_valid_client_ip(
+            "203.0.113.1, 198.51.100.1, 192.0.2.1, 192.0.2.2, 192.0.2.3, 192.0.2.4",
+            5
+        ));
+    }
 }
--- a/crates/utils/src/string.rs
+++ b/crates/utils/src/string.rs
@@ -48,6 +48,14 @@ pub fn parse_bool(str: &str) -> Result<bool> {
    }
 }

+pub fn parse_bool_with_default(str: &str, default: bool) -> bool {
+    match str {
+        "1" | "t" | "T" | "true" | "TRUE" | "True" | "on" | "ON" | "On" | "enabled" => true,
+        "0" | "f" | "F" | "false" | "FALSE" | "False" | "off" | "OFF" | "Off" | "disabled" => false,
+        _ => default,
+    }
+}
+
 /// Matches a simple pattern against a name using wildcards.
 ///
 /// # Arguments
--- a/flake.nix
+++ b/flake.nix
@@ -37,7 +37,10 @@

            src = ./.;

-            cargoLock.lockFile = ./Cargo.lock;
+            cargoLock = {
+              lockFile = ./Cargo.lock;
+              allowBuiltinFetchGit = true;
+            };

            nativeBuildInputs = with pkgs; [
              pkg-config
--- a/helm/README.md
+++ b/helm/README.md
@@ -44,7 +44,7 @@ RustFS helm chart supports **standalone and distributed mode**. For standalone m
 | imageRegistryCredentials.username | string | `""` | The username to pull rustfs image from private registry. |
 | ingress.className | string | `"traefik"` | Specify the ingress class, traefik or nginx. |
 | ingress.enabled | bool | `true` |  |
-| ingress.hosts[0].host | string | `"your.rustfs.com"` |  |
+| ingress.hosts[0].host | string | `"example.rustfs.com"` |  |
 | ingress.hosts[0].paths[0].path | string | `"/"` |  |
 | ingress.hosts[0].paths[0].pathType | string | `"ImplementationSpecific"` |  |
 | ingress.nginxAnnotations."nginx.ingress.kubernetes.io/affinity" | string | `"cookie"` |  |
@@ -52,6 +52,7 @@ RustFS helm chart supports **standalone and distributed mode**. For standalone m
 | ingress.nginxAnnotations."nginx.ingress.kubernetes.io/session-cookie-hash" | string | `"sha1"` |  |
 | ingress.nginxAnnotations."nginx.ingress.kubernetes.io/session-cookie-max-age" | string | `"3600"` |  |
 | ingress.nginxAnnotations."nginx.ingress.kubernetes.io/session-cookie-name" | string | `"rustfs"` |  |
+| ingress.customAnnotations | dict | `{}` |Customize annotations.  |
 | ingress.traefikAnnotations."traefik.ingress.kubernetes.io/service.sticky.cookie" | string | `"true"` |  |
 | ingress.traefikAnnotations."traefik.ingress.kubernetes.io/service.sticky.cookie.httponly" | string | `"true"` |  |
 | ingress.traefikAnnotations."traefik.ingress.kubernetes.io/service.sticky.cookie.name" | string | `"rustfs"` |  |
@@ -59,8 +60,6 @@ RustFS helm chart supports **standalone and distributed mode**. For standalone m
 | ingress.traefikAnnotations."traefik.ingress.kubernetes.io/service.sticky.cookie.secure" | string | `"true"` |  |
 | ingress.tls.enabled | bool | `false` | Enable tls and access rustfs via https. |
 | ingress.tls.certManager.enabled | string | `false` | Enable cert manager support to generate certificate automatically. |
-| ingress.tls.certManager.issuer.name | string | `false` | The name of cert manager issuer. |
-| ingress.tls.certManager.issuer.kind | string | `false` | The kind of cert manager issuer, issuer or cluster-issuer. |
 | ingress.tls.crt | string | "" | The content of certificate file. |
 | ingress.tls.key | string | "" | The content of key file. |
 | livenessProbe.failureThreshold | int | `3` |  |
@@ -94,9 +93,11 @@ RustFS helm chart supports **standalone and distributed mode**. For standalone m
 | secret.existingSecret | string | `""` | Use existing secret with a credentials. |
 | secret.rustfs.access_key | string | `"rustfsadmin"` | RustFS Access Key ID |
 | secret.rustfs.secret_key | string | `"rustfsadmin"` | RustFS Secret Key ID |
-| service.console_port | int | `9001` |  |
-| service.ep_port | int | `9000` |  |
 | service.type | string | `"NodePort"` |  |
+| service.console.nodePort | int | `32001` |  |
+| service.console.port | int | `9001` |  |
+| service.endpoint.nodePort | int | `32000` |  |
+| service.endpoint.port | int | `9000` |  |
 | serviceAccount.annotations | object | `{}` |  |
 | serviceAccount.automount | bool | `true` |  |
 | serviceAccount.create | bool | `true` |  |
@@ -179,12 +180,12 @@ Check the ingress status
 ```
 kubectl -n rustfs get ing
 NAME     CLASS   HOSTS            ADDRESS         PORTS     AGE
-rustfs   nginx   your.rustfs.com   10.43.237.152   80, 443   29m
+rustfs   nginx   example.rustfs.com   10.43.237.152   80, 443   29m
 ```

-Access the rustfs cluster via `https://your.rustfs.com` with the default username and password `rustfsadmin`.
+Access the rustfs cluster via `https://example.rustfs.com` with the default username and password `rustfsadmin`.

-> Replace the `your.rustfs.com` with your own domain as well as the certificates.
+> Replace the `example.rustfs.com` with your own domain as well as the certificates.

 # TLS configuration

--- a/helm/rustfs/templates/_helpers.tpl
+++ b/helm/rustfs/templates/_helpers.tpl
@@ -104,10 +104,10 @@ Render RUSTFS_VOLUMES
 */}}
 {{- define "rustfs.volumes" -}}
 {{- if eq (int .Values.replicaCount) 4 }}
-{{- printf "http://%s-{0...%d}.%s-headless:%d/data/rustfs{0...%d}" (include "rustfs.fullname" .) (sub (.Values.replicaCount | int) 1) (include "rustfs.fullname" . ) (.Values.service.ep_port | int) (sub (.Values.replicaCount | int) 1) }}
+{{- printf "http://%s-{0...%d}.%s-headless:%d/data/rustfs{0...%d}" (include "rustfs.fullname" .) (sub (.Values.replicaCount | int) 1) (include "rustfs.fullname" . ) (.Values.service.endpoint.port | int) (sub (.Values.replicaCount | int) 1) }}
 {{- end }}
 {{- if eq (int .Values.replicaCount) 16 }}
-{{- printf "http://%s-{0...%d}.%s-headless:%d/data" (include "rustfs.fullname" .) (sub (.Values.replicaCount | int) 1) (include "rustfs.fullname" .) (.Values.service.ep_port | int) }}
+{{- printf "http://%s-{0...%d}.%s-headless:%d/data" (include "rustfs.fullname" .) (sub (.Values.replicaCount | int) 1) (include "rustfs.fullname" .) (.Values.service.endpoint.port | int) }}
 {{- end }}
 {{- end }}

--- a/helm/rustfs/templates/certificate.yml
+++ b/helm/rustfs/templates/certificate.yml
@@ -1,15 +0,0 @@
-{{- if and .Values.ingress.tls.enabled .Values.ingress.tls.certManager.enabled }}
-{{- $host := index .Values.ingress.hosts 0 }}
-apiVersion: cert-manager.io/v1
-kind: Certificate
-metadata:
-  name: {{ include "rustfs.fullname" . }}-tls
-  namespace: {{ .Release.Namespace }}
-spec:
-  secretName: {{ .Values.ingress.tls.secretName }}
-  issuerRef:
-    name: {{ .Values.ingress.tls.certManager.issuer.name }}
-    kind: {{ .Values.ingress.tls.certManager.issuer.kind }}
-  dnsNames:
-    - {{ $host.host }}
-{{- end }}
--- a/helm/rustfs/templates/deployment.yaml
+++ b/helm/rustfs/templates/deployment.yaml
@@ -57,11 +57,10 @@ spec:
      {{- end }}
      initContainers:
        - name: init-step
-          image: busybox
-          imagePullPolicy: {{ .Values.image.pullPolicy }}
+          image: "{{ .Values.initStep.image.repository }}:{{ .Values.initStep.image.tag }}"
+          imagePullPolicy: {{ .Values.initStep.image.pullPolicy }}
          securityContext:
-            runAsUser: 0
-            runAsGroup: 0
+            {{- toYaml .Values.initStep.containerSecurityContext | nindent 12 }}
          command:
            - sh
            - -c
@@ -83,10 +82,10 @@ spec:
          {{- toYaml .Values.containerSecurityContext | nindent 12 }}
          {{- end }}
          ports:
-            - containerPort: {{ .Values.service.ep_port }}
-              name: endpoint
-            - containerPort: {{ .Values.service.console_port }}
-              name: console
+            - name: endpoint
+              containerPort: {{ .Values.service.endpoint.port }}
+            - name: console
+              containerPort: {{ .Values.service.console.port }}
          envFrom:
            - configMapRef:
                name: {{ include "rustfs.fullname" . }}-config
--- a/helm/rustfs/templates/ingress.yaml
+++ b/helm/rustfs/templates/ingress.yaml
@@ -2,12 +2,9 @@
 {{- $secretName := .Values.ingress.tls.secretName }}
 {{- $ingressAnnotations := dict }}
 {{- if eq .Values.ingress.className "nginx" }}
-  {{- $ingressAnnotations = merge $ingressAnnotations (.Values.ingress.nginxAnnotations | default dict) }}
+{{- $ingressAnnotations = .Values.ingress.nginxAnnotations }}
 {{- else if eq .Values.ingress.className "" }}
-  {{- $ingressAnnotations = merge $ingressAnnotations (.Values.ingress.customAnnoations | default dict) }}
-{{- end }}
-{{- if .Values.ingress.tls.certManager.enabled }}
-  {{- $ingressAnnotations = merge $ingressAnnotations (.Values.ingress.certManagerAnnotations | default dict) }}
+{{- $ingressAnnotations = .Values.ingress.customAnnotations }}
 {{- end }}
 apiVersion: networking.k8s.io/v1
 kind: Ingress
--- a/helm/rustfs/templates/statefulset.yaml
+++ b/helm/rustfs/templates/statefulset.yaml
@@ -42,6 +42,7 @@ spec:
        {{- else }}
          {}
        {{- if .Values.affinity.podAntiAffinity.enabled }}
+        {{- end }}
        podAntiAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
            - labelSelector:
@@ -52,7 +53,6 @@ spec:
                      - {{ include "rustfs.name" . }}
              topologyKey: {{ .Values.affinity.podAntiAffinity.topologyKey }}
        {{- end }}
-        {{- end }}
      {{- end }}
      {{- if .Values.tolerations }}
      tolerations:
@@ -68,11 +68,10 @@ spec:
      {{- end }}
      initContainers:
        - name: init-step
-          image: busybox
-          imagePullPolicy: {{ .Values.image.pullPolicy }}
+          image: "{{ .Values.initStep.image.repository }}:{{ .Values.initStep.image.tag }}"
+          imagePullPolicy: {{ .Values.initStep.image.pullPolicy }}
          securityContext:
-            runAsUser: 0
-            runAsGroup: 0
+            {{- toYaml .Values.initStep.containerSecurityContext | nindent 12 }}
          env:
            - name: REPLICA_COUNT
              value: {{ .Values.replicaCount | quote }}
@@ -111,13 +110,10 @@ spec:
            {{- toYaml .Values.containerSecurityContext | nindent 12 }}
          {{- end }}
          ports:
-            - containerPort: {{ .Values.service.ep_port }}
-              name: endpoint
-            - containerPort: {{ .Values.service.console_port }}
-              name: console
-          env:
-            - name: REPLICA_COUNT
-              value: {{ .Values.replicaCount | quote }}
+            - name: endpoint
+              containerPort: {{ .Values.service.endpoint.port }}
+            - name: console
+              containerPort: {{ .Values.service.console.port }}
          envFrom:
            - configMapRef:
                name: {{ include "rustfs.fullname" . }}-config
--- a/helm/rustfs/templates/tests/test-connection.yaml
+++ b/helm/rustfs/templates/tests/test-connection.yaml
@@ -11,5 +11,5 @@ spec:
    - name: wget
      image: busybox
      command: ['wget']
-      args: ['-O', '/dev/null', '{{ include "rustfs.fullname" . }}-svc:{{ .Values.service.ep_port }}/health']
+      args: ['-O', '/dev/null', '{{ include "rustfs.fullname" . }}-svc:{{ .Values.service.endpoint.port }}/health']
  restartPolicy: Never
--- a/helm/rustfs/values.yaml
+++ b/helm/rustfs/values.yaml
@@ -114,23 +114,17 @@ ingress:
    nginx.ingress.kubernetes.io/session-cookie-hash: sha1
    nginx.ingress.kubernetes.io/session-cookie-max-age: "3600"
    nginx.ingress.kubernetes.io/session-cookie-name: rustfs
-  certManagerAnnotations:
-    {} # Specify cert manager issuer annotations,cert-manager.io/issuer or cert-manager.io/cluster-issuer.
-    # cert-manager.io/issuer: "letsencrypt-staging"
  customAnnotations: # Specify custom annotations
    {} # Customize annotations
  hosts:
-    - host: xmg.rustfs.com
+    - host: example.rustfs.com
      paths:
        - path: /
          pathType: Prefix
-  tls: 
+  tls:
    enabled: false # Enable tls and access rustfs via https.
    certManager:
      enabled: false # Enable certmanager to generate certificate for rustfs, default false.
-      issuer:
-        name: letsencrypt-staging # Specify cert manager issuer name
-        kind: Issuer # Specify cert manager issuer kind, Issuer or ClusterIssuer.
    secretName: secret-tls
    crt: tls.crt
    key: tls.key
@@ -183,4 +177,16 @@ storageclass:
  dataStorageSize: 256Mi
  logStorageSize: 256Mi

+# Init container parameters.
+initStep:
+  image:
+    repository: busybox
+    pullPolicy: IfNotPresent
+    tag: "latest"
+  containerSecurityContext:
+    runAsUser: 0
+    runAsGroup: 0
+
+
+
 extraManifests: []
--- a/rustfs/Cargo.toml
+++ b/rustfs/Cargo.toml
@@ -72,6 +72,7 @@ hyper.workspace = true
 hyper-util.workspace = true
 http.workspace = true
 http-body.workspace = true
+http-body-util.workspace = true
 reqwest = { workspace = true }
 socket2 = { workspace = true }
 tokio = { workspace = true, features = ["rt-multi-thread", "macros", "net", "signal", "process", "io-util"] }
--- a/rustfs/src/admin/auth.rs
+++ b/rustfs/src/admin/auth.rs
@@ -1,6 +1,18 @@
-use std::collections::HashMap;
-use std::sync::Arc;
+// Copyright 2024 RustFS Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.

+use crate::auth::get_condition_values;
 use http::HeaderMap;
 use rustfs_iam::store::object::ObjectStore;
 use rustfs_iam::sys::IamSys;
@@ -9,8 +21,8 @@ use rustfs_policy::policy::Args;
 use rustfs_policy::policy::action::Action;
 use s3s::S3Result;
 use s3s::s3_error;
-
-use crate::auth::get_condition_values;
+use std::collections::HashMap;
+use std::sync::Arc;

 pub async fn validate_admin_request(
    headers: &HeaderMap,
--- a/rustfs/src/admin/console.rs
+++ b/rustfs/src/admin/console.rs
@@ -14,6 +14,7 @@

 use crate::config::build;
 use crate::license::get_license;
+use crate::server::{CONSOLE_PREFIX, FAVICON_PATH, HEALTH_PREFIX, RUSTFS_ADMIN_PREFIX};
 use axum::{
    Router,
    body::Body,
@@ -45,9 +46,6 @@ use tower_http::timeout::TimeoutLayer;
 use tower_http::trace::TraceLayer;
 use tracing::{debug, error, info, instrument, warn};

-pub(crate) const CONSOLE_PREFIX: &str = "/rustfs/console";
-const RUSTFS_ADMIN_PREFIX: &str = "/rustfs/admin/v3";
-
 #[derive(RustEmbed)]
 #[folder = "$CARGO_MANIFEST_DIR/static"]
 struct StaticFiles;
@@ -457,7 +455,7 @@ fn get_console_config_from_env() -> (bool, u32, u64, String) {
 /// # Returns:
 /// - `true` if the path is for console access, `false` otherwise.
 pub fn is_console_path(path: &str) -> bool {
-    path == "/favicon.ico" || path.starts_with(CONSOLE_PREFIX)
+    path == FAVICON_PATH || path.starts_with(CONSOLE_PREFIX)
 }

 /// Setup comprehensive middleware stack with tower-http features
@@ -477,11 +475,11 @@ fn setup_console_middleware_stack(
    auth_timeout: u64,
 ) -> Router {
    let mut app = Router::new()
-        .route("/favicon.ico", get(static_handler))
+        .route(FAVICON_PATH, get(static_handler))
        .route(&format!("{CONSOLE_PREFIX}/license"), get(license_handler))
        .route(&format!("{CONSOLE_PREFIX}/config.json"), get(config_handler))
        .route(&format!("{CONSOLE_PREFIX}/version"), get(version_handler))
-        .route(&format!("{CONSOLE_PREFIX}/health"), get(health_check).head(health_check))
+        .route(&format!("{CONSOLE_PREFIX}{HEALTH_PREFIX}"), get(health_check).head(health_check))
        .nest(CONSOLE_PREFIX, Router::new().fallback_service(get(static_handler)))
        .fallback_service(get(static_handler));

--- a/rustfs/src/admin/handlers.rs
+++ b/rustfs/src/admin/handlers.rs
@@ -72,7 +72,6 @@ use tokio_stream::wrappers::ReceiverStream;
 use tracing::debug;
 use tracing::{error, info, warn};
 use url::Host;
-// use url::UrlQuery;

 pub mod bucket_meta;
 pub mod event;
--- a/rustfs/src/admin/handlers/rebalance.rs
+++ b/rustfs/src/admin/handlers/rebalance.rs
@@ -12,8 +12,13 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.

+use crate::{
+    admin::{auth::validate_admin_request, router::Operation},
+    auth::{check_key_valid, get_session_token},
+};
 use http::{HeaderMap, StatusCode};
 use matchit::Params;
+use rustfs_ecstore::rebalance::RebalanceMeta;
 use rustfs_ecstore::{
    StorageAPI,
    error::StorageError,
@@ -33,12 +38,6 @@ use std::time::Duration;
 use time::OffsetDateTime;
 use tracing::warn;

-use crate::{
-    admin::{auth::validate_admin_request, router::Operation},
-    auth::{check_key_valid, get_session_token},
-};
-use rustfs_ecstore::rebalance::RebalanceMeta;
-
 #[derive(Debug, Clone, Deserialize, Serialize)]
 pub struct RebalanceResp {
    pub id: String,
--- a/rustfs/src/admin/handlers/trace.rs
+++ b/rustfs/src/admin/handlers/trace.rs
@@ -12,6 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.

+use crate::admin::router::Operation;
 use http::StatusCode;
 use hyper::Uri;
 use matchit::Params;
@@ -20,8 +21,6 @@ use rustfs_madmin::service_commands::ServiceTraceOpts;
 use s3s::{Body, S3Request, S3Response, S3Result, s3_error};
 use tracing::warn;

-use crate::admin::router::Operation;
-
 #[allow(dead_code)]
 fn extract_trace_options(uri: &Uri) -> S3Result<ServiceTraceOpts> {
    let mut st_opts = ServiceTraceOpts::default();
--- a/rustfs/src/admin/mod.rs
+++ b/rustfs/src/admin/mod.rs
@@ -22,6 +22,7 @@ pub mod utils;
 #[cfg(test)]
 mod console_test;

+use crate::server::{ADMIN_PREFIX, HEALTH_PREFIX, PROFILE_CPU_PATH, PROFILE_MEMORY_PATH};
 use handlers::{
    GetReplicationMetricsHandler, HealthCheckHandler, IsAdminHandler, ListRemoteTargetHandler, RemoveRemoteTargetHandler,
    SetRemoteTargetHandler, bucket_meta,
@@ -37,17 +38,21 @@ use router::{AdminOperation, S3Router};
 use rpc::register_rpc_route;
 use s3s::route::S3Route;

-const ADMIN_PREFIX: &str = "/rustfs/admin";
-// const ADMIN_PREFIX: &str = "/minio/admin";
-
+/// Create admin router
+///
+/// # Arguments
+/// * `console_enabled` - Whether the console is enabled
+///
+/// # Returns
+/// An instance of S3Route for admin operations
 pub fn make_admin_route(console_enabled: bool) -> std::io::Result<impl S3Route> {
    let mut r: S3Router<AdminOperation> = S3Router::new(console_enabled);

    // Health check endpoint for monitoring and orchestration
-    r.insert(Method::GET, "/health", AdminOperation(&HealthCheckHandler {}))?;
-    r.insert(Method::HEAD, "/health", AdminOperation(&HealthCheckHandler {}))?;
-    r.insert(Method::GET, "/profile/cpu", AdminOperation(&TriggerProfileCPU {}))?;
-    r.insert(Method::GET, "/profile/memory", AdminOperation(&TriggerProfileMemory {}))?;
+    r.insert(Method::GET, HEALTH_PREFIX, AdminOperation(&HealthCheckHandler {}))?;
+    r.insert(Method::HEAD, HEALTH_PREFIX, AdminOperation(&HealthCheckHandler {}))?;
+    r.insert(Method::GET, PROFILE_CPU_PATH, AdminOperation(&TriggerProfileCPU {}))?;
+    r.insert(Method::GET, PROFILE_MEMORY_PATH, AdminOperation(&TriggerProfileMemory {}))?;

    // 1
    r.insert(Method::POST, "/", AdminOperation(&sts::AssumeRoleHandle {}))?;
--- a/rustfs/src/admin/router.rs
+++ b/rustfs/src/admin/router.rs
@@ -12,10 +12,9 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.

-use crate::admin::ADMIN_PREFIX;
 use crate::admin::console::is_console_path;
 use crate::admin::console::make_console_server;
-use crate::admin::rpc::RPC_PREFIX;
+use crate::server::{ADMIN_PREFIX, HEALTH_PREFIX, PROFILE_CPU_PATH, PROFILE_MEMORY_PATH, RPC_PREFIX};
 use hyper::HeaderMap;
 use hyper::Method;
 use hyper::StatusCode;
@@ -86,12 +85,12 @@ where
    fn is_match(&self, method: &Method, uri: &Uri, headers: &HeaderMap, _: &mut Extensions) -> bool {
        let path = uri.path();
        // Profiling endpoints
-        if method == Method::GET && (path == "/profile/cpu" || path == "/profile/memory") {
+        if method == Method::GET && (path == PROFILE_CPU_PATH || path == PROFILE_MEMORY_PATH) {
            return true;
        }

        // Health check
-        if (method == Method::HEAD || method == Method::GET) && path == "/health" {
+        if (method == Method::HEAD || method == Method::GET) && path == HEALTH_PREFIX {
            return true;
        }

@@ -117,12 +116,12 @@ where
        let path = req.uri.path();

        // Profiling endpoints
-        if req.method == Method::GET && (path == "/profile/cpu" || path == "/profile/memory") {
+        if req.method == Method::GET && (path == PROFILE_CPU_PATH || path == PROFILE_MEMORY_PATH) {
            return Ok(());
        }

        // Health check
-        if (req.method == Method::HEAD || req.method == Method::GET) && path == "/health" {
+        if (req.method == Method::HEAD || req.method == Method::GET) && path == HEALTH_PREFIX {
            return Ok(());
        }

--- a/rustfs/src/admin/rpc.rs
+++ b/rustfs/src/admin/rpc.rs
@@ -15,6 +15,7 @@
 use super::router::AdminOperation;
 use super::router::Operation;
 use super::router::S3Router;
+use crate::server::RPC_PREFIX;
 use futures::StreamExt;
 use http::StatusCode;
 use hyper::Method;
@@ -36,8 +37,6 @@ use tokio::io::AsyncWriteExt;
 use tokio_util::io::ReaderStream;
 use tracing::warn;

-pub const RPC_PREFIX: &str = "/rustfs/rpc";
-
 pub fn register_rpc_route(r: &mut S3Router<AdminOperation>) -> std::io::Result<()> {
    r.insert(
        Method::GET,
--- a/rustfs/src/auth.rs
+++ b/rustfs/src/auth.rs
@@ -66,7 +66,7 @@ const SIGN_V2_ALGORITHM: &str = "AWS ";
 const SIGN_V4_ALGORITHM: &str = "AWS4-HMAC-SHA256";
 const STREAMING_CONTENT_SHA256: &str = "STREAMING-AWS4-HMAC-SHA256-PAYLOAD";
 const STREAMING_CONTENT_SHA256_TRAILER: &str = "STREAMING-AWS4-HMAC-SHA256-PAYLOAD-TRAILER";
-pub const UNSIGNED_PAYLOAD_TRAILER: &str = "STREAMING-UNSIGNED-PAYLOAD-TRAILER";
+pub(crate) const UNSIGNED_PAYLOAD_TRAILER: &str = "STREAMING-UNSIGNED-PAYLOAD-TRAILER";
 const ACTION_HEADER: &str = "Action";
 const AMZ_CREDENTIAL: &str = "X-Amz-Credential";
 const AMZ_ACCESS_KEY_ID: &str = "AWSAccessKeyId";
--- a/rustfs/src/init.rs
+++ b/rustfs/src/init.rs
@@ -13,7 +13,8 @@
 // limitations under the License.

 use crate::storage::ecfs::{process_lambda_configurations, process_queue_configurations, process_topic_configurations};
-use crate::{admin, config};
+use crate::{admin, config, version};
+use chrono::Datelike;
 use rustfs_config::{DEFAULT_UPDATE_CHECK, ENV_UPDATE_CHECK};
 use rustfs_ecstore::bucket::metadata_sys;
 use rustfs_notify::notifier_global;
@@ -23,6 +24,21 @@ use std::env;
 use std::io::Error;
 use tracing::{debug, error, info, instrument, warn};

+#[instrument]
+pub(crate) fn print_server_info() {
+    let current_year = chrono::Utc::now().year();
+    // Use custom macros to print server information
+    info!("RustFS Object Storage Server");
+    info!("Copyright: 2024-{} RustFS, Inc", current_year);
+    info!("License: Apache-2.0 https://www.apache.org/licenses/LICENSE-2.0");
+    info!("Version: {}", version::get_version());
+    info!("Docs: https://rustfs.com/docs/");
+}
+
+/// Initialize the asynchronous update check system.
+/// This function checks if update checking is enabled via
+/// environment variable or default configuration. If enabled,
+/// it spawns an asynchronous task to check for updates with a timeout.
 pub(crate) fn init_update_check() {
    let update_check_enable = env::var(ENV_UPDATE_CHECK)
        .unwrap_or_else(|_| DEFAULT_UPDATE_CHECK.to_string())
@@ -70,6 +86,12 @@ pub(crate) fn init_update_check() {
    });
 }

+/// Add existing bucket notification configurations to the global notifier system.
+/// This function retrieves notification configurations for each bucket
+/// and registers the corresponding event rules with the notifier system.
+///  It processes queue, topic, and lambda configurations and maps them to event rules.
+///  # Arguments
+/// * `buckets` - A vector of bucket names to process
 #[instrument(skip_all)]
 pub(crate) async fn add_bucket_notification_configuration(buckets: Vec<String>) {
    let region_opt = rustfs_ecstore::global::get_global_region();
@@ -128,6 +150,15 @@ pub(crate) async fn add_bucket_notification_configuration(buckets: Vec<String>)
 }

 /// Initialize KMS system and configure if enabled
+///
+/// This function initializes the global KMS service manager. If KMS is enabled
+/// via command line options, it configures and starts the service accordingly.
+/// If not enabled, it attempts to load any persisted KMS configuration from
+/// cluster storage and starts the service if found.
+/// # Arguments
+/// * `opt` - The application configuration options
+///
+/// Returns `std::io::Result<()>` indicating success or failure
 #[instrument(skip(opt))]
 pub(crate) async fn init_kms_system(opt: &config::Opt) -> std::io::Result<()> {
    // Initialize global KMS service manager (starts in NotConfigured state)
--- a/rustfs/src/main.rs
+++ b/rustfs/src/main.rs
@@ -25,19 +25,20 @@ mod update;
 mod version;

 // Ensure the correct path for parse_license is imported
-use crate::init::{add_bucket_notification_configuration, init_buffer_profile_system, init_kms_system, init_update_check};
+use crate::init::{
+    add_bucket_notification_configuration, init_buffer_profile_system, init_kms_system, init_update_check, print_server_info,
+};
 use crate::server::{
    SHUTDOWN_TIMEOUT, ServiceState, ServiceStateManager, ShutdownSignal, init_cert, init_event_notifier, shutdown_event_notifier,
    start_audit_system, start_http_server, stop_audit_system, wait_for_shutdown,
 };
-use chrono::Datelike;
 use clap::Parser;
 use license::init_license;
 use rustfs_ahm::{
    Scanner, create_ahm_services_cancel_token, heal::storage::ECStoreHealStorage, init_heal_manager,
    scanner::data_scanner::ScannerConfig, shutdown_ahm_services,
 };
-use rustfs_common::globals::set_global_addr;
+use rustfs_common::{GlobalReadiness, SystemStage, set_global_addr};
 use rustfs_ecstore::{
    StorageAPI,
    bucket::metadata_sys::init_bucket_metadata_sys,
@@ -69,25 +70,6 @@ static GLOBAL: tikv_jemallocator::Jemalloc = tikv_jemallocator::Jemalloc;
 #[global_allocator]
 static GLOBAL: mimalloc::MiMalloc = mimalloc::MiMalloc;

-const LOGO: &str = r#"
-
-░█▀▄░█░█░█▀▀░▀█▀░█▀▀░█▀▀
-░█▀▄░█░█░▀▀█░░█░░█▀▀░▀▀█
-░▀░▀░▀▀▀░▀▀▀░░▀░░▀░░░▀▀▀
-
-"#;
-
-#[instrument]
-fn print_server_info() {
-    let current_year = chrono::Utc::now().year();
-    // Use custom macros to print server information
-    info!("RustFS Object Storage Server");
-    info!("Copyright: 2024-{} RustFS, Inc", current_year);
-    info!("License: Apache-2.0 https://www.apache.org/licenses/LICENSE-2.0");
-    info!("Version: {}", version::get_version());
-    info!("Docs: https://rustfs.com/docs/");
-}
-
 fn main() -> Result<()> {
    let runtime = server::get_tokio_runtime_builder()
        .build()
@@ -120,7 +102,7 @@ async fn async_main() -> Result<()> {
    }

    // print startup logo
-    info!("{}", LOGO);
+    info!("{}", server::LOGO);

    // Initialize performance profiling if enabled
    profiling::init_from_env().await;
@@ -143,6 +125,8 @@ async fn async_main() -> Result<()> {
 #[instrument(skip(opt))]
 async fn run(opt: config::Opt) -> Result<()> {
    debug!("opt: {:?}", &opt);
+    // 1. Initialize global readiness tracker
+    let readiness = Arc::new(GlobalReadiness::new());

    if let Some(region) = &opt.region {
        rustfs_ecstore::global::set_global_region(region.clone());
@@ -214,14 +198,14 @@ async fn run(opt: config::Opt) -> Result<()> {
    let s3_shutdown_tx = {
        let mut s3_opt = opt.clone();
        s3_opt.console_enable = false;
-        let s3_shutdown_tx = start_http_server(&s3_opt, state_manager.clone()).await?;
+        let s3_shutdown_tx = start_http_server(&s3_opt, state_manager.clone(), readiness.clone()).await?;
        Some(s3_shutdown_tx)
    };

    let console_shutdown_tx = if opt.console_enable && !opt.console_address.is_empty() {
        let mut console_opt = opt.clone();
        console_opt.address = console_opt.console_address.clone();
-        let console_shutdown_tx = start_http_server(&console_opt, state_manager.clone()).await?;
+        let console_shutdown_tx = start_http_server(&console_opt, state_manager.clone(), readiness.clone()).await?;
        Some(console_shutdown_tx)
    } else {
        None
@@ -236,6 +220,7 @@ async fn run(opt: config::Opt) -> Result<()> {
    let ctx = CancellationToken::new();

    // init store
+    // 2. Start Storage Engine (ECStore)
    let store = ECStore::new(server_addr, endpoint_pools.clone(), ctx.clone())
        .await
        .inspect_err(|err| {
@@ -243,10 +228,20 @@ async fn run(opt: config::Opt) -> Result<()> {
        })?;

    ecconfig::init();
-    // config system configuration
-    GLOBAL_CONFIG_SYS.init(store.clone()).await?;

-    // init  replication_pool
+    // // Initialize global configuration system
+    let mut retry_count = 0;
+    while let Err(e) = GLOBAL_CONFIG_SYS.init(store.clone()).await {
+        error!("GLOBAL_CONFIG_SYS.init failed {:?}", e);
+        // TODO: check error type
+        retry_count += 1;
+        if retry_count > 15 {
+            return Err(Error::other("GLOBAL_CONFIG_SYS.init failed"));
+        }
+        tokio::time::sleep(tokio::time::Duration::from_secs(1)).await;
+    }
+    readiness.mark_stage(SystemStage::StorageReady);
+    // init replication_pool
    init_background_replication(store.clone()).await;
    // Initialize KMS system if enabled
    init_kms_system(&opt).await?;
@@ -279,7 +274,10 @@ async fn run(opt: config::Opt) -> Result<()> {

    init_bucket_metadata_sys(store.clone(), buckets.clone()).await;

+    // 3. Initialize IAM System (Blocking load)
+    // This ensures data is in memory before moving forward
    init_iam_sys(store.clone()).await.map_err(Error::other)?;
+    readiness.mark_stage(SystemStage::IamReady);

    add_bucket_notification_configuration(buckets.clone()).await;

@@ -331,6 +329,15 @@ async fn run(opt: config::Opt) -> Result<()> {

    init_update_check();

+    info!(target: "rustfs::main::run","server started successfully at {}", &server_address);
+    // 4. Mark as Full Ready now that critical components are warm
+    readiness.mark_stage(SystemStage::FullReady);
+    println!(
+        "RustFS server started successfully at {}, current time: {}",
+        &server_address,
+        chrono::offset::Utc::now().to_string()
+    );
+
    // Perform hibernation for 1 second
    tokio::time::sleep(SHUTDOWN_TIMEOUT).await;
    // listen to the shutdown signal
--- a/rustfs/src/server/cert.rs
+++ b/rustfs/src/server/cert.rs
@@ -12,7 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.

-use rustfs_common::globals::set_global_root_cert;
+use rustfs_common::set_global_root_cert;
 use rustfs_config::{RUSTFS_CA_CERT, RUSTFS_PUBLIC_CERT, RUSTFS_TLS_CERT};
 use tracing::{debug, info};

--- a/rustfs/src/server/http.rs
+++ b/rustfs/src/server/http.rs
@@ -17,7 +17,7 @@ use super::compress::{CompressionConfig, CompressionPredicate};
 use crate::admin;
 use crate::auth::IAMAuth;
 use crate::config;
-use crate::server::{ServiceState, ServiceStateManager, hybrid::hybrid, layer::RedirectLayer};
+use crate::server::{ReadinessGateLayer, ServiceState, ServiceStateManager, hybrid::hybrid, layer::RedirectLayer};
 use crate::storage;
 use crate::storage::tonic_service::make_server;
 use bytes::Bytes;
@@ -29,6 +29,7 @@ use hyper_util::{
    service::TowerToHyperService,
 };
 use metrics::{counter, histogram};
+use rustfs_common::GlobalReadiness;
 use rustfs_config::{DEFAULT_ACCESS_KEY, DEFAULT_SECRET_KEY, MI_B, RUSTFS_TLS_CERT, RUSTFS_TLS_KEY};
 use rustfs_protos::proto_gen::node_service::node_service_server::NodeServiceServer;
 use rustfs_utils::net::parse_and_resolve_address;
@@ -112,6 +113,7 @@ fn get_cors_allowed_origins() -> String {
 pub async fn start_http_server(
    opt: &config::Opt,
    worker_state_manager: ServiceStateManager,
+    readiness: Arc<GlobalReadiness>,
 ) -> Result<tokio::sync::broadcast::Sender<()>> {
    let server_addr = parse_and_resolve_address(opt.address.as_str()).map_err(Error::other)?;
    let server_port = server_addr.port();
@@ -119,16 +121,26 @@ pub async fn start_http_server(
    // The listening address and port are obtained from the parameters
    let listener = {
        let mut server_addr = server_addr;
-        let mut socket = socket2::Socket::new(
+
+        // Try to create a socket for the address family; if that fails, fallback to IPv4.
+        let mut socket = match socket2::Socket::new(
            socket2::Domain::for_address(server_addr),
            socket2::Type::STREAM,
            Some(socket2::Protocol::TCP),
-        )?;
+        ) {
+            Ok(s) => s,
+            Err(e) => {
+                warn!("Failed to create socket for {:?}: {}, falling back to IPv4", server_addr, e);
+                let ipv4_addr = SocketAddr::new(std::net::Ipv4Addr::UNSPECIFIED.into(), server_addr.port());
+                server_addr = ipv4_addr;
+                socket2::Socket::new(socket2::Domain::IPV4, socket2::Type::STREAM, Some(socket2::Protocol::TCP))?
+            }
+        };

+        // If address is IPv6 try to enable dual-stack; on failure, switch to IPv4 socket.
        if server_addr.is_ipv6() {
            if let Err(e) = socket.set_only_v6(false) {
-                warn!("Failed to set IPV6_V6ONLY=false, falling back to IPv4-only: {}", e);
-                // Fallback to a new IPv4 socket if setting dual-stack fails.
+                warn!("Failed to set IPV6_V6ONLY=false, attempting IPv4 fallback: {}", e);
                let ipv4_addr = SocketAddr::new(std::net::Ipv4Addr::UNSPECIFIED.into(), server_addr.port());
                server_addr = ipv4_addr;
                socket = socket2::Socket::new(socket2::Domain::IPV4, socket2::Type::STREAM, Some(socket2::Protocol::TCP))?;
@@ -140,8 +152,27 @@ pub async fn start_http_server(
        socket.set_reuse_address(true)?;
        // Set the socket to non-blocking before passing it to Tokio.
        socket.set_nonblocking(true)?;
-        socket.bind(&server_addr.into())?;
-        socket.listen(backlog)?;
+
+        // Attempt bind; if bind fails for IPv6, try IPv4 fallback once more.
+        if let Err(bind_err) = socket.bind(&server_addr.into()) {
+            warn!("Failed to bind to {}: {}.", server_addr, bind_err);
+            if server_addr.is_ipv6() {
+                // Try IPv4 fallback
+                let ipv4_addr = SocketAddr::new(std::net::Ipv4Addr::UNSPECIFIED.into(), server_addr.port());
+                server_addr = ipv4_addr;
+                socket = socket2::Socket::new(socket2::Domain::IPV4, socket2::Type::STREAM, Some(socket2::Protocol::TCP))?;
+                socket.set_reuse_address(true)?;
+                socket.set_nonblocking(true)?;
+                socket.bind(&server_addr.into())?;
+                // [FIX] Ensure fallback socket is moved to listening state as well.
+                socket.listen(backlog)?;
+            } else {
+                return Err(bind_err);
+            }
+        } else {
+            // Listen on the socket when initial bind succeeded
+            socket.listen(backlog)?;
+        }
        TcpListener::from_std(socket.into())?
    };

@@ -179,7 +210,7 @@ pub async fn start_http_server(
        println!("Console WebUI (localhost): {protocol}://127.0.0.1:{server_port}/rustfs/console/index.html",);
    } else {
        info!(target: "rustfs::main::startup","RustFS API: {api_endpoints}  {localhost_endpoint}");
-        println!("RustFS API: {api_endpoints}  {localhost_endpoint}");
+        println!("RustFS Http API: {api_endpoints}  {localhost_endpoint}");
        println!("RustFS Start Time: {now_time}");
        if DEFAULT_ACCESS_KEY.eq(&opt.access_key) && DEFAULT_SECRET_KEY.eq(&opt.secret_key) {
            warn!(
@@ -359,6 +390,7 @@ pub async fn start_http_server(
                cors_layer: cors_layer.clone(),
                compression_config: compression_config.clone(),
                is_console,
+                readiness: readiness.clone(),
            };

            process_connection(socket, tls_acceptor.clone(), connection_ctx, graceful.clone());
@@ -461,6 +493,7 @@ struct ConnectionContext {
    cors_layer: CorsLayer,
    compression_config: CompressionConfig,
    is_console: bool,
+    readiness: Arc<GlobalReadiness>,
 }

 /// Process a single incoming TCP connection.
@@ -484,6 +517,7 @@ fn process_connection(
            cors_layer,
            compression_config,
            is_console,
+            readiness,
        } = context;

        // Build services inside each connected task to avoid passing complex service types across tasks,
@@ -494,6 +528,9 @@ fn process_connection(
        let hybrid_service = ServiceBuilder::new()
            .layer(SetRequestIdLayer::x_request_id(MakeRequestUuid))
            .layer(CatchPanicLayer::new())
+            // CRITICAL: Insert ReadinessGateLayer before business logic
+            // This stops requests from hitting IAMAuth or Storage if they are not ready.
+            .layer(ReadinessGateLayer::new(readiness))
            .layer(
                TraceLayer::new_for_http()
                    .make_span_with(|request: &HttpRequest<_>| {
--- a/rustfs/src/server/mod.rs
+++ b/rustfs/src/server/mod.rs
@@ -19,6 +19,8 @@ mod event;
 mod http;
 mod hybrid;
 mod layer;
+mod prefix;
+mod readiness;
 mod runtime;
 mod service_state;

@@ -26,6 +28,8 @@ pub(crate) use audit::{start_audit_system, stop_audit_system};
 pub(crate) use cert::init_cert;
 pub(crate) use event::{init_event_notifier, shutdown_event_notifier};
 pub(crate) use http::start_http_server;
+pub(crate) use prefix::*;
+pub(crate) use readiness::ReadinessGateLayer;
 pub(crate) use runtime::get_tokio_runtime_builder;
 pub(crate) use service_state::SHUTDOWN_TIMEOUT;
 pub(crate) use service_state::ServiceState;
--- a/rustfs/src/server/prefix.rs
+++ b/rustfs/src/server/prefix.rs
@@ -0,0 +1,55 @@
+// Copyright 2024 RustFS Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+/// Predefined CPU profiling path for RustFS server.
+/// This path is used to access CPU profiling data.
+pub(crate) const PROFILE_CPU_PATH: &str = "/profile/cpu";
+
+/// This path is used to access memory profiling data.
+pub(crate) const PROFILE_MEMORY_PATH: &str = "/profile/memory";
+
+/// Favicon path to handle browser requests for the favicon.
+/// This path serves the favicon.ico file.
+pub(crate) const FAVICON_PATH: &str = "/favicon.ico";
+
+/// Predefined health check path for RustFS server.
+/// This path is used to check the health status of the server.
+pub(crate) const HEALTH_PREFIX: &str = "/health";
+
+/// Predefined administrative prefix for RustFS server routes.
+/// This prefix is used for endpoints that handle administrative tasks
+/// such as configuration, monitoring, and management.
+pub(crate) const ADMIN_PREFIX: &str = "/rustfs/admin";
+
+/// Environment variable name for overriding the default
+/// administrative prefix path.
+pub(crate) const RUSTFS_ADMIN_PREFIX: &str = "/rustfs/admin/v3";
+
+/// Predefined console prefix for RustFS server routes.
+/// This prefix is used for endpoints that handle console-related tasks
+/// such as user interface and management.
+pub(crate) const CONSOLE_PREFIX: &str = "/rustfs/console";
+
+/// Predefined RPC prefix for RustFS server routes.
+/// This prefix is used for endpoints that handle remote procedure calls (RPC).
+pub(crate) const RPC_PREFIX: &str = "/rustfs/rpc";
+
+/// LOGO art for RustFS server.
+pub(crate) const LOGO: &str = r#"
+
+░█▀▄░█░█░█▀▀░▀█▀░█▀▀░█▀▀
+░█▀▄░█░█░▀▀█░░█░░█▀▀░▀▀█
+░▀░▀░▀▀▀░▀▀▀░░▀░░▀░░░▀▀▀
+
+"#;
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
houseme	4ac9143165	fix	2025-12-27 01:22:03 +08:00
houseme	a5695f35d2	add ipnet crates	2025-12-27 01:08:52 +08:00
houseme	cd9b5ad3f3	fix	2025-12-25 11:35:40 +08:00
houseme	04833f35cf	Security fix: Add trusted proxy validation for SourceIP bypass vulnerability	2025-12-25 01:11:05 +08:00
houseme	82a6e78845	Inject `GlobalReadiness` into HTTP server pipeline and gate traffic until `FullReady` (#1255 )	2025-12-25 00:19:03 +08:00
houseme	7e75c9b1f5	remove unlinked file (#1258 )	2025-12-24 23:37:43 +08:00
weisd	8bdff3fbcb	fix: Add retry mechanism for GLOBAL_CONFIG_SYS initialization (#1252 )	2025-12-24 16:38:28 +08:00
Andrea Manzi	65d32e693f	add ca-certificates in mcp-server Dockerfile (#1248 ) Signed-off-by: Andrea Manzi <andrea.manzi@gmail.com>	2025-12-24 08:36:14 +08:00
Michele Zanotti	1ff28b3157	helm: expose init container parameters as helm values (#1232 ) Co-authored-by: houseme <housemecn@gmail.com>	2025-12-23 21:31:28 +08:00
Juri Malinovski	2186f46ea3	helm: fix service/containers ports, fix podAntiAffinity (#1230 ) Co-authored-by: majinghe <42570491+majinghe@users.noreply.github.com>	2025-12-23 20:36:33 +08:00
唐小鸭	add6453aea	feat: add seek support for small objects in rustfs (#1231 ) Co-authored-by: loverustfs <hello@rustfs.com>	2025-12-23 20:27:34 +08:00
yxrxy	4418c882ad	Revert "fix(iam): store previous credentials in .rustfs.sys bucket to… (#1238 ) Co-authored-by: loverustfs <hello@rustfs.com>	2025-12-23 19:37:39 +08:00
Muhammed Hussain Karimi	00c607b5ce	🧑‍💻 Fix nix develop problem with Git-Based dependecies on nix develop shell (#1243 ) Signed-off-by: Muhammed Hussain Karimi <info@karimi.dev>	2025-12-23 19:26:50 +08:00
majinghe	79585f98e0	delete userless helm chart file (#1245 )	2025-12-23 19:15:29 +08:00
majinghe	2a3517f1d5	Custom annotation (#1242 )	2025-12-23 17:31:01 +08:00
tryao	3942e07487	console port is 9001 (#1235 ) Signed-off-by: tryao <yaotairan@gmail.com>	2025-12-23 13:36:38 +08:00
houseme	04811c0006	update s3s version (#1237 )	2025-12-23 13:09:57 +08:00
Ali Mehraji	73c15d6be1	Add: rust installation in Makefile (#1188 ) Signed-off-by: Ali Mehraji <a.mehraji75@gmail.com> Signed-off-by: houseme <housemecn@gmail.com> Co-authored-by: houseme <housemecn@gmail.com> Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>	2025-12-23 08:51:04 +08:00
loverustfs	af5c0b13ef	fix: HeadObject returns 404 for deleted objects with versioning enabled (#1229 ) Co-authored-by: houseme <housemecn@gmail.com>	2025-12-22 20:43:00 +08:00
Juri Malinovski	f17990f746	helm: allow to define additional config variables (#1220 ) Signed-off-by: Juri Malinovski <juri.malinovski@coolbet.com>	2025-12-22 20:25:23 +08:00
weisd	80cfb4feab	Add Disk Timeout and Health Check Functionality (#1196 ) Signed-off-by: weisd <im@weisd.in> Co-authored-by: loverustfs <hello@rustfs.com>	2025-12-22 17:15:19 +08:00
houseme	08f1a31f3f	Fix notification event stream cleanup, add bounded send concurrency, and reduce overhead (#1224 )	2025-12-22 00:57:05 +08:00
loverustfs	1c51e204ab	ci: reduce cargo build jobs to 2 for standard-2 runner	2025-12-21 23:54:40 +08:00
loverustfs	958f054123	ci: update all workflows to use ubicloud-standard-2 runner	2025-12-21 23:43:12 +08:00