fix:rm object versions (#385)

* chore: upgrade actions/checkout from v4 to v5

- Update GitHub Actions checkout action version
- Ensure compatibility with latest workflow features
- Maintain existing checkout behavior and configuration

* upgrade version

.dockerignore

@@ -0,0 +1 @@
+target

.github/pull_request_template.md

@@ -19,9 +19,7 @@ Pull Request Template for RustFS
 
 ## Checklist
 - [ ] I have read and followed the [CONTRIBUTING.md](CONTRIBUTING.md) guidelines
-- [ ] Code is formatted with `cargo fmt --all`
-- [ ] Passed `cargo clippy --all-targets --all-features -- -D warnings`
-- [ ] Passed `cargo check --all-targets`
+- [ ] Passed `make pre-commit`
 - [ ] Added/updated necessary tests
 - [ ] Documentation updated (if needed)
 - [ ] CI/CD passed (if applicable)

.github/workflows/audit.yml

@@ -16,13 +16,13 @@ name: Security Audit
 
 on:
   push:
-    branches: [main]
+    branches: [ main ]
     paths:
       - '**/Cargo.toml'
       - '**/Cargo.lock'
       - '.github/workflows/audit.yml'
   pull_request:
-    branches: [main]
+    branches: [ main ]
     paths:
       - '**/Cargo.toml'
       - '**/Cargo.lock'
@@ -41,7 +41,7 @@ jobs:
     timeout-minutes: 15
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@v5
 
       - name: Install cargo-audit
         uses: taiki-e/install-action@v2
@@ -69,7 +69,7 @@ jobs:
       pull-requests: write
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@v5
 
       - name: Dependency Review
         uses: actions/dependency-review-action@v4

.github/workflows/build.yml

@@ -28,8 +28,8 @@ name: Build and Release
 
 on:
   push:
-    tags: ["*.*.*"]
-    branches: [main]
+    tags: [ "*.*.*" ]
+    branches: [ main ]
     paths-ignore:
       - "**.md"
       - "**.txt"
@@ -45,7 +45,7 @@ on:
       - ".gitignore"
       - ".dockerignore"
   pull_request:
-    branches: [main]
+    branches: [ main ]
     paths-ignore:
       - "**.md"
       - "**.txt"
@@ -89,7 +89,7 @@ jobs:
       is_prerelease: ${{ steps.check.outputs.is_prerelease }}
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@v5
         with:
           fetch-depth: 0
 
@@ -153,7 +153,7 @@ jobs:
   # Build RustFS binaries
   build-rustfs:
     name: Build RustFS
-    needs: [build-check]
+    needs: [ build-check ]
     if: needs.build-check.outputs.should_build == 'true'
     runs-on: ${{ matrix.os }}
     timeout-minutes: 60
@@ -200,7 +200,7 @@ jobs:
           #  platform: windows
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@v5
         with:
           fetch-depth: 0
 
@@ -527,7 +527,7 @@ jobs:
   # Build summary
   build-summary:
     name: Build Summary
-    needs: [build-check, build-rustfs]
+    needs: [ build-check, build-rustfs ]
     if: always() && needs.build-check.outputs.should_build == 'true'
     runs-on: ubuntu-latest
     steps:
@@ -579,7 +579,7 @@ jobs:
   # Create GitHub Release (only for tag pushes)
   create-release:
     name: Create GitHub Release
-    needs: [build-check, build-rustfs]
+    needs: [ build-check, build-rustfs ]
     if: startsWith(github.ref, 'refs/tags/') && needs.build-check.outputs.build_type != 'development'
     runs-on: ubuntu-latest
     permissions:
@@ -589,7 +589,7 @@ jobs:
       release_url: ${{ steps.create.outputs.release_url }}
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@v5
         with:
           fetch-depth: 0
 
@@ -665,7 +665,7 @@ jobs:
   # Prepare and upload release assets
   upload-release-assets:
     name: Upload Release Assets
-    needs: [build-check, build-rustfs, create-release]
+    needs: [ build-check, build-rustfs, create-release ]
     if: startsWith(github.ref, 'refs/tags/') && needs.build-check.outputs.build_type != 'development'
     runs-on: ubuntu-latest
     permissions:
@@ -673,10 +673,10 @@ jobs:
       actions: read
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@v5
 
       - name: Download all build artifacts
-        uses: actions/download-artifact@v4
+        uses: actions/download-artifact@v5
         with:
           path: ./artifacts
           pattern: rustfs-*
@@ -746,7 +746,7 @@ jobs:
   # Update latest.json for stable releases only
   update-latest-version:
     name: Update Latest Version
-    needs: [build-check, upload-release-assets]
+    needs: [ build-check, upload-release-assets ]
     if: startsWith(github.ref, 'refs/tags/')
     runs-on: ubuntu-latest
     steps:
@@ -796,14 +796,14 @@ jobs:
   # Publish release (remove draft status)
   publish-release:
     name: Publish Release
-    needs: [build-check, create-release, upload-release-assets]
+    needs: [ build-check, create-release, upload-release-assets ]
     if: startsWith(github.ref, 'refs/tags/') && needs.build-check.outputs.build_type != 'development'
     runs-on: ubuntu-latest
     permissions:
       contents: write
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@v5
 
       - name: Update release notes and publish
         env:

.github/workflows/ci.yml

@@ -16,7 +16,7 @@ name: Continuous Integration
 
 on:
   push:
-    branches: [main]
+    branches: [ main ]
     paths-ignore:
       - "**.md"
       - "**.txt"
@@ -36,7 +36,7 @@ on:
       - ".github/workflows/audit.yml"
       - ".github/workflows/performance.yml"
   pull_request:
-    branches: [main]
+    branches: [ main ]
     paths-ignore:
       - "**.md"
       - "**.txt"
@@ -88,7 +88,7 @@ jobs:
     name: Typos
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v5
       - uses: dtolnay/rust-toolchain@stable
       - name: Typos check with custom config file
         uses: crate-ci/typos@master
@@ -101,7 +101,7 @@ jobs:
     timeout-minutes: 60
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@v5
 
       - name: Setup Rust environment
         uses: ./.github/actions/setup
@@ -130,7 +130,7 @@ jobs:
     timeout-minutes: 30
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@v5
 
       - name: Setup Rust environment
         uses: ./.github/actions/setup

.github/workflows/docker.yml

@@ -36,8 +36,8 @@ permissions:
 on:
   # Automatically triggered when build workflow completes
   workflow_run:
-    workflows: ["Build and Release"]
-    types: [completed]
+    workflows: [ "Build and Release" ]
+    types: [ completed ]
   # Manual trigger with same parameters for consistency
   workflow_dispatch:
     inputs:
@@ -79,7 +79,7 @@ jobs:
       create_latest: ${{ steps.check.outputs.create_latest }}
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@v5
         with:
           fetch-depth: 0
           # For workflow_run events, checkout the specific commit that triggered the workflow
@@ -250,7 +250,7 @@ jobs:
     timeout-minutes: 60
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@v5
 
       - name: Login to Docker Hub
         uses: docker/login-action@v3
@@ -382,7 +382,7 @@ jobs:
   # Docker build summary
   docker-summary:
     name: Docker Build Summary
-    needs: [build-check, build-docker]
+    needs: [ build-check, build-docker ]
     if: always() && needs.build-check.outputs.should_build == 'true'
     runs-on: ubuntu-latest
     steps:

.github/workflows/performance.yml

@@ -16,7 +16,7 @@ name: Performance Testing
 
 on:
   push:
-    branches: [main]
+    branches: [ main ]
     paths:
       - "**/*.rs"
       - "**/Cargo.toml"
@@ -41,7 +41,7 @@ jobs:
     timeout-minutes: 30
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@v5
 
       - name: Setup Rust environment
         uses: ./.github/actions/setup
@@ -116,7 +116,7 @@ jobs:
     timeout-minutes: 45
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@v5
 
       - name: Setup Rust environment
         uses: ./.github/actions/setup

.gitignore

@@ -20,3 +20,4 @@ profile.json
 .docker/openobserve-otel/data
 *.zst
 .secrets
+*.go

Cargo.toml

@@ -90,32 +90,29 @@ rustfs-checksums = { path = "crates/checksums", version = "0.0.5" }
 rustfs-workers = { path = "crates/workers", version = "0.0.5" }
 rustfs-mcp = { path = "crates/mcp", version = "0.0.5" }
 aes-gcm = { version = "0.10.3", features = ["std"] }
-anyhow = "1.0.98"
+anyhow = "1.0.99"
 arc-swap = "1.7.1"
 argon2 = { version = "0.5.3", features = ["std"] }
 atoi = "2.0.0"
 async-channel = "2.5.0"
 async-recursion = "1.1.1"
 async-trait = "0.1.88"
-async-compression = { version = "0.4.0" }
+async-compression = { version = "0.4.19" }
 atomic_enum = "0.3.0"
-aws-config = { version = "1.8.3" }
-aws-sdk-s3 = "1.100.0"
+aws-config = { version = "1.8.4" }
+aws-sdk-s3 = "1.101.0"
 axum = "0.8.4"
-axum-extra = "0.10.1"
-axum-server = { version = "0.7.2", features = ["tls-rustls"] }
 base64-simd = "0.8.0"
 base64 = "0.22.1"
 brotli = "8.0.1"
 bytes = { version = "1.10.1", features = ["serde"] }
 bytesize = "2.0.1"
 byteorder = "1.5.0"
-bytes-utils = "0.1.4"
 cfg-if = "1.0.1"
-crc-fast = "1.3.0"
+crc-fast = "1.4.0"
 chacha20poly1305 = { version = "0.10.1" }
 chrono = { version = "0.4.41", features = ["serde"] }
-clap = { version = "4.5.42", features = ["derive", "env"] }
+clap = { version = "4.5.44", features = ["derive", "env"] }
 const-str = { version = "0.6.4", features = ["std", "proc"] }
 crc32fast = "1.5.0"
 criterion = { version = "0.7", features = ["html_reports"] }
@@ -132,7 +129,7 @@ form_urlencoded = "1.2.1"
 futures = "0.3.31"
 futures-core = "0.3.31"
 futures-util = "0.3.31"
-glob = "0.3.2"
+glob = "0.3.3"
 hex = "0.4.3"
 hex-simd = "0.8.0"
 highway = { version = "1.3.0" }
@@ -156,7 +153,6 @@ keyring = { version = "3.6.3", features = [
 ] }
 lazy_static = "1.5.0"
 libsystemd = { version = "0.7.2" }
-lru = "0.16"
 local-ip-address = "0.6.5"
 lz4 = "1.28.1"
 matchit = "0.8.4"
@@ -192,7 +188,7 @@ percent-encoding = "2.3.1"
 pin-project-lite = "0.2.16"
 prost = "0.14.1"
 pretty_assertions = "1.4.1"
-quick-xml = "0.38.0"
+quick-xml = "0.38.1"
 rand = "0.9.2"
 rdkafka = { version = "0.38.0", features = ["tokio"] }
 reed-solomon-simd = { version = "3.0.1" }
@@ -210,7 +206,7 @@ rfd = { version = "0.15.4", default-features = false, features = [
     "xdg-portal",
     "tokio",
 ] }
-rmcp = { version = "0.3.2" }
+rmcp = { version = "0.5.0" }
 rmp = "0.8.14"
 rmp-serde = "1.3.0"
 rsa = "0.9.8"
@@ -236,12 +232,12 @@ snafu = "0.8.6"
 snap = "1.1.1"
 socket2 = "0.6.0"
 strum = { version = "0.27.2", features = ["derive"] }
-sysinfo = "0.36.1"
+sysinfo = "0.37.0"
 sysctl = "0.6.0"
 tempfile = "3.20.0"
 temp-env = "0.3.6"
 test-case = "3.3.1"
-thiserror = "2.0.12"
+thiserror = "2.0.14"
 time = { version = "0.3.41", features = [
     "std",
     "parsing",
@@ -254,10 +250,10 @@ tokio-rustls = { version = "0.26.2", default-features = false }
 tokio-stream = { version = "0.1.17" }
 tokio-tar = "0.3.1"
 tokio-test = "0.4.4"
-tokio-util = { version = "0.7.15", features = ["io", "compat"] }
-tonic = { version = "0.14.0", features = ["gzip"] }
-tonic-prost = { version = "0.14.0" }
-tonic-prost-build = { version = "0.14.0" }
+tokio-util = { version = "0.7.16", features = ["io", "compat"] }
+tonic = { version = "0.14.1", features = ["gzip"] }
+tonic-prost = { version = "0.14.1" }
+tonic-prost-build = { version = "0.14.1" }
 tower = { version = "0.5.2", features = ["timeout"] }
 tower-http = { version = "0.6.6", features = ["cors"] }
 tracing = "0.1.41"
@@ -266,11 +262,10 @@ tracing-core = "0.1.34"
 tracing-error = "0.2.1"
 tracing-opentelemetry = "0.31.0"
 tracing-subscriber = { version = "0.3.19", features = ["env-filter", "time"] }
-tracing-test = "0.2.5"
 transform-stream = "0.3.1"
 url = "2.5.4"
 urlencoding = "2.1.3"
-uuid = { version = "1.17.0", features = [
+uuid = { version = "1.18.0", features = [
     "v4",
     "fast-rng",
     "macro-diagnostics",
@@ -283,7 +278,7 @@ zstd = "0.13.3"
 
 
 [workspace.metadata.cargo-shear]
-ignored = ["rustfs", "rust-i18n"]
+ignored = ["rustfs", "rust-i18n", "rustfs-mcp"]
 
 [profile.wasm-dev]
 inherits = "dev"

Justfile

@@ -0,0 +1,258 @@
+DOCKER_CLI := env("DOCKER_CLI", "docker")
+IMAGE_NAME := env("IMAGE_NAME", "rustfs:v1.0.0")
+DOCKERFILE_SOURCE := env("DOCKERFILE_SOURCE", "Dockerfile.source")
+DOCKERFILE_PRODUCTION := env("DOCKERFILE_PRODUCTION", "Dockerfile")
+CONTAINER_NAME := env("CONTAINER_NAME", "rustfs-dev")
+
+[group("📒 Help")]
+[private]
+default:
+    @just --list --list-heading $'🦀 RustFS justfile manual page:\n'
+
+[doc("show help")]
+[group("📒 Help")]
+help: default
+
+[doc("run `cargo fmt` to format codes")]
+[group("👆 Code Quality")]
+fmt:
+    @echo "🔧 Formatting code..."
+    cargo fmt --all
+
+[doc("run `cargo fmt` in check mode")]
+[group("👆 Code Quality")]
+fmt-check:
+    @echo "📝 Checking code formatting..."
+    cargo fmt --all --check
+
+[doc("run `cargo clippy`")]
+[group("👆 Code Quality")]
+clippy:
+    @echo "🔍 Running clippy checks..."
+    cargo clippy --all-targets --all-features --fix --allow-dirty -- -D warnings
+
+[doc("run `cargo check`")]
+[group("👆 Code Quality")]
+check:
+    @echo "🔨 Running compilation check..."
+    cargo check --all-targets
+
+[doc("run `cargo test`")]
+[group("👆 Code Quality")]
+test:
+    @echo "🧪 Running tests..."
+    cargo nextest run --all --exclude e2e_test
+    cargo test --all --doc
+
+[doc("run `fmt` `clippy` `check` `test` at once")]
+[group("👆 Code Quality")]
+pre-commit: fmt clippy check test
+    @echo "✅ All pre-commit checks passed!"
+
+[group("🤔 Git")]
+setup-hooks:
+    @echo "🔧 Setting up git hooks..."
+    chmod +x .git/hooks/pre-commit
+    @echo "✅ Git hooks setup complete!"
+
+[doc("use `release` mode for building")]
+[group("🔨 Build")]
+build:
+    @echo "🔨 Building RustFS using build-rustfs.sh script..."
+    ./build-rustfs.sh
+
+[doc("use `debug` mode for building")]
+[group("🔨 Build")]
+build-dev:
+    @echo "🔨 Building RustFS in development mode..."
+    ./build-rustfs.sh --dev
+
+[group("🔨 Build")]
+[private]
+build-target target:
+    @echo "🔨 Building rustfs for {{ target }}..."
+    @echo "💡 On macOS/Windows, use 'make build-docker' or 'make docker-dev' instead"
+    ./build-rustfs.sh --platform {{ target }}
+
+[doc("use `x86_64-unknown-linux-musl` target for building")]
+[group("🔨 Build")]
+build-musl: (build-target "x86_64-unknown-linux-musl")
+
+[doc("use `x86_64-unknown-linux-gnu` target for building")]
+[group("🔨 Build")]
+build-gnu: (build-target "x86_64-unknown-linux-gnu")
+
+[doc("use `aarch64-unknown-linux-musl` target for building")]
+[group("🔨 Build")]
+build-musl-arm64: (build-target "aarch64-unknown-linux-musl")
+
+[doc("use `aarch64-unknown-linux-gnu` target for building")]
+[group("🔨 Build")]
+build-gnu-arm64: (build-target "aarch64-unknown-linux-gnu")
+
+[doc("build and deploy to server")]
+[group("🔨 Build")]
+deploy-dev ip: build-musl
+    @echo "🚀 Deploying to dev server: {{ ip }}"
+    ./scripts/dev_deploy.sh {{ ip }}
+
+[group("🔨 Build")]
+[private]
+build-cross-all-pre:
+    @echo "🔧 Building all target architectures..."
+    @echo "💡 On macOS/Windows, use 'make docker-dev' for reliable multi-arch builds"
+    @echo "🔨 Generating protobuf code..."
+    -cargo run --bin gproto
+
+[doc("build all targets at once")]
+[group("🔨 Build")]
+build-cross-all: build-cross-all-pre && build-gnu build-gnu-arm64 build-musl build-musl-arm64
+
+# ========================================================================================
+# Docker Multi-Architecture Builds (Primary Methods)
+# ========================================================================================
+
+[doc("build an image and run it")]
+[group("🐳 Build Image")]
+build-docker os="rockylinux9.3" cli=(DOCKER_CLI) dockerfile=(DOCKERFILE_SOURCE):
+    #!/usr/bin/env bash
+    SOURCE_BUILD_IMAGE_NAME="rustfs/rustfs-{{ os }}:v1"
+    SOURCE_BUILD_CONTAINER_NAME="rustfs-{{ os }}-build"
+    BUILD_CMD="/root/.cargo/bin/cargo build --release --bin rustfs --target-dir /root/s3-rustfs/target/{{ os }}"
+    echo "🐳 Building RustFS using Docker ({{ os }})..."
+    {{ cli }} buildx build -t $SOURCE_BUILD_IMAGE_NAME -f {{ dockerfile }} .
+    {{ cli }} run --rm --name $SOURCE_BUILD_CONTAINER_NAME -v $(pwd):/root/s3-rustfs -it $SOURCE_BUILD_IMAGE_NAME $BUILD_CMD
+
+[doc("build an image")]
+[group("🐳 Build Image")]
+docker-buildx:
+    @echo "🏗️ Building multi-architecture production Docker images with buildx..."
+    ./docker-buildx.sh
+
+[doc("build an image and push it")]
+[group("🐳 Build Image")]
+docker-buildx-push:
+    @echo "🚀 Building and pushing multi-architecture production Docker images with buildx..."
+    ./docker-buildx.sh --push
+
+[doc("build an image with a version")]
+[group("🐳 Build Image")]
+docker-buildx-version version:
+    @echo "🏗️ Building multi-architecture production Docker images (version: {{ version }}..."
+    ./docker-buildx.sh --release {{ version }}
+
+[doc("build an image with a version and push it")]
+[group("🐳 Build Image")]
+docker-buildx-push-version version:
+    @echo "🚀 Building and pushing multi-architecture production Docker images (version: {{ version }}..."
+    ./docker-buildx.sh --release {{ version }} --push
+
+[doc("build an image with a version and push it to registry")]
+[group("🐳 Build Image")]
+docker-dev-push registry cli=(DOCKER_CLI) source=(DOCKERFILE_SOURCE):
+    @echo "🚀 Building and pushing multi-architecture development Docker images..."
+    @echo "💡 push to registry: {{ registry }}"
+    {{ cli }} buildx build \
+    	--platform linux/amd64,linux/arm64 \
+    	--file {{ source }} \
+    	--tag {{ registry }}/rustfs:source-latest \
+    	--tag {{ registry }}/rustfs:dev-latest \
+    	--push \
+    	.
+
+# Local production builds using direct buildx (alternative to docker-buildx.sh)
+
+[group("🐳 Build Image")]
+docker-buildx-production-local cli=(DOCKER_CLI) source=(DOCKERFILE_PRODUCTION):
+    @echo "🏗️ Building single-architecture production Docker image locally..."
+    @echo "💡 Alternative to docker-buildx.sh for local testing"
+    {{ cli }} buildx build \
+    	--file {{ source }} \
+    	--tag rustfs:production-latest \
+    	--tag rustfs:latest \
+    	--load \
+    	--build-arg RELEASE=latest \
+    	.
+
+# Development/Source builds using direct buildx commands
+
+[group("🐳 Build Image")]
+docker-dev cli=(DOCKER_CLI) source=(DOCKERFILE_SOURCE):
+    @echo "🏗️ Building multi-architecture development Docker images with buildx..."
+    @echo "💡 This builds from source code and is intended for local development and testing"
+    @echo "⚠️  Multi-arch images cannot be loaded locally, use docker-dev-push to push to registry"
+    {{ cli }} buildx build \
+    	--platform linux/amd64,linux/arm64 \
+    	--file {{ source }} \
+    	--tag rustfs:source-latest \
+    	--tag rustfs:dev-latest \
+    	.
+
+[group("🐳 Build Image")]
+docker-dev-local cli=(DOCKER_CLI) source=(DOCKERFILE_SOURCE):
+    @echo "🏗️ Building single-architecture development Docker image for local use..."
+    @echo "💡 This builds from source code for the current platform and loads locally"
+    {{ cli }} buildx build \
+    	--file {{ source }} \
+    	--tag rustfs:source-latest \
+    	--tag rustfs:dev-latest \
+    	--load \
+    	.
+
+# ========================================================================================
+# Single Architecture Docker Builds (Traditional)
+# ========================================================================================
+
+[group("🐳 Build Image")]
+docker-build-production cli=(DOCKER_CLI) source=(DOCKERFILE_PRODUCTION):
+    @echo "🏗️ Building single-architecture production Docker image..."
+    @echo "💡 Consider using 'make docker-buildx-production-local' for multi-arch support"
+    {{ cli }} build -f {{ source }} -t rustfs:latest .
+
+[group("🐳 Build Image")]
+docker-build-source cli=(DOCKER_CLI) source=(DOCKERFILE_SOURCE):
+    @echo "🏗️ Building single-architecture source Docker image..."
+    @echo "💡 Consider using 'make docker-dev-local' for multi-arch support"
+    {{ cli }} build -f {{ source }} -t rustfs:source .
+
+# ========================================================================================
+# Development Environment
+# ========================================================================================
+
+[group("🏃 Running")]
+dev-env-start cli=(DOCKER_CLI) source=(DOCKERFILE_SOURCE) container=(CONTAINER_NAME):
+    @echo "🚀 Starting development environment..."
+    {{ cli }} buildx build \
+    	--file {{ source }} \
+    	--tag rustfs:dev \
+    	--load \
+    	.
+    -{{ cli }} stop {{ container }} 2>/dev/null
+    -{{ cli }} rm {{ container }} 2>/dev/null
+    {{ cli }} run -d --name {{ container }} \
+    	-p 9010:9010 -p 9000:9000 \
+    	-v {{ invocation_directory() }}:/workspace \
+    	-it rustfs:dev
+
+[group("🏃 Running")]
+dev-env-stop cli=(DOCKER_CLI) container=(CONTAINER_NAME):
+    @echo "🛑 Stopping development environment..."
+    -{{ cli }} stop {{ container }} 2>/dev/null
+    -{{ cli }}  rm {{ container }} 2>/dev/null
+
+[group("🏃 Running")]
+dev-env-restart: dev-env-stop dev-env-start
+
+[group("👍 E2E")]
+e2e-server:
+    sh scripts/run.sh
+
+[group("👍 E2E")]
+probe-e2e:
+    sh scripts/probe.sh
+
+[doc("inspect one image")]
+[group("🚚 Other")]
+docker-inspect-multiarch image cli=(DOCKER_CLI):
+    @echo "🔍 Inspecting multi-architecture image: {{ image }}"
+    {{ cli }} buildx imagetools inspect {{ image }}

Makefile

@@ -76,7 +76,7 @@ build-docker: SOURCE_BUILD_CONTAINER_NAME = rustfs-$(BUILD_OS)-build
 build-docker: BUILD_CMD = /root/.cargo/bin/cargo build --release --bin rustfs --target-dir /root/s3-rustfs/target/$(BUILD_OS)
 build-docker:
 	@echo "🐳 Building RustFS using Docker ($(BUILD_OS))..."
-	$(DOCKER_CLI) build -t $(SOURCE_BUILD_IMAGE_NAME) -f $(DOCKERFILE_SOURCE) .
+	$(DOCKER_CLI) buildx build -t $(SOURCE_BUILD_IMAGE_NAME) -f $(DOCKERFILE_SOURCE) .
 	$(DOCKER_CLI) run --rm --name $(SOURCE_BUILD_CONTAINER_NAME) -v $(shell pwd):/root/s3-rustfs -it $(SOURCE_BUILD_IMAGE_NAME) $(BUILD_CMD)
 
 .PHONY: build-musl

crates/ahm/Cargo.toml

@@ -24,24 +24,19 @@ tracing = { workspace = true }
 serde = { workspace = true, features = ["derive"] }
 serde_json = { workspace = true }
 thiserror = { workspace = true }
-bytes = { workspace = true }
-time = { workspace = true, features = ["serde"] }
 uuid = { workspace = true, features = ["v4", "serde"] }
 anyhow = { workspace = true }
 async-trait = { workspace = true }
 futures = { workspace = true }
 url = { workspace = true }
 rustfs-lock = { workspace = true }
-
+s3s = { workspace = true }
 lazy_static = { workspace = true }
 chrono = { workspace = true }
 
 [dev-dependencies]
-rmp-serde = { workspace = true }
-tokio-test = { workspace = true }
 serde_json = { workspace = true }
 serial_test = "3.2.0"
-once_cell = { workspace = true }
 tracing-subscriber = { workspace = true }
 walkdir = "2.5.0"
 tempfile = { workspace = true }

crates/ahm/src/heal/storage.rs

@@ -133,8 +133,14 @@ impl HealStorageAPI for ECStoreHealStorage {
         match self.ecstore.get_object_info(bucket, object, &Default::default()).await {
             Ok(info) => Ok(Some(info)),
             Err(e) => {
-                error!("Failed to get object meta: {}/{} - {}", bucket, object, e);
-                Err(Error::other(e))
+                // Map ObjectNotFound to None to align with Option return type
+                if matches!(e, rustfs_ecstore::error::StorageError::ObjectNotFound(_, _)) {
+                    debug!("Object meta not found: {}/{}", bucket, object);
+                    Ok(None)
+                } else {
+                    error!("Failed to get object meta: {}/{} - {}", bucket, object, e);
+                    Err(Error::other(e))
+                }
             }
         }
     }
@@ -142,22 +148,47 @@ impl HealStorageAPI for ECStoreHealStorage {
     async fn get_object_data(&self, bucket: &str, object: &str) -> Result<Option<Vec<u8>>> {
         debug!("Getting object data: {}/{}", bucket, object);
 
-        match (*self.ecstore)
+        let reader = match (*self.ecstore)
             .get_object_reader(bucket, object, None, Default::default(), &Default::default())
             .await
         {
-            Ok(mut reader) => match reader.read_all().await {
-                Ok(data) => Ok(Some(data)),
-                Err(e) => {
-                    error!("Failed to read object data: {}/{} - {}", bucket, object, e);
-                    Err(Error::other(e))
-                }
-            },
+            Ok(reader) => reader,
             Err(e) => {
                 error!("Failed to get object: {}/{} - {}", bucket, object, e);
-                Err(Error::other(e))
+                return Err(Error::other(e));
+            }
+        };
+
+        // WARNING: Returning Vec<u8> for large objects is dangerous. To avoid OOM, cap the read size.
+        // If needed, refactor callers to stream instead of buffering entire object.
+        const MAX_READ_BYTES: usize = 16 * 1024 * 1024; // 16 MiB cap
+        let mut buf = Vec::with_capacity(1024 * 1024);
+        use tokio::io::AsyncReadExt as _;
+        let mut n_read: usize = 0;
+        let mut stream = reader.stream;
+        loop {
+            // Read in chunks
+            let mut chunk = vec![0u8; 1024 * 1024];
+            match stream.read(&mut chunk).await {
+                Ok(0) => break,
+                Ok(n) => {
+                    buf.extend_from_slice(&chunk[..n]);
+                    n_read += n;
+                    if n_read > MAX_READ_BYTES {
+                        warn!(
+                            "Object data exceeds cap ({} bytes), aborting full read to prevent OOM: {}/{}",
+                            MAX_READ_BYTES, bucket, object
+                        );
+                        return Ok(None);
+                    }
+                }
+                Err(e) => {
+                    error!("Failed to read object data: {}/{} - {}", bucket, object, e);
+                    return Err(Error::other(e));
+                }
             }
         }
+        Ok(Some(buf))
     }
 
     async fn put_object_data(&self, bucket: &str, object: &str, data: &[u8]) -> Result<()> {
@@ -197,27 +228,34 @@ impl HealStorageAPI for ECStoreHealStorage {
     async fn verify_object_integrity(&self, bucket: &str, object: &str) -> Result<bool> {
         debug!("Verifying object integrity: {}/{}", bucket, object);
 
-        // Try to get object info and data to verify integrity
+        // Check object metadata first
         match self.get_object_meta(bucket, object).await? {
             Some(obj_info) => {
-                // Check if object has valid metadata
                 if obj_info.size < 0 {
                     warn!("Object has invalid size: {}/{}", bucket, object);
                     return Ok(false);
                 }
 
-                // Try to read object data to verify it's accessible
-                match self.get_object_data(bucket, object).await {
-                    Ok(Some(_)) => {
-                        info!("Object integrity check passed: {}/{}", bucket, object);
-                        Ok(true)
+                // Stream-read the object to a sink to avoid loading into memory
+                match (*self.ecstore)
+                    .get_object_reader(bucket, object, None, Default::default(), &Default::default())
+                    .await
+                {
+                    Ok(reader) => {
+                        let mut stream = reader.stream;
+                        match tokio::io::copy(&mut stream, &mut tokio::io::sink()).await {
+                            Ok(_) => {
+                                info!("Object integrity check passed: {}/{}", bucket, object);
+                                Ok(true)
+                            }
+                            Err(e) => {
+                                warn!("Object stream read failed: {}/{} - {}", bucket, object, e);
+                                Ok(false)
+                            }
+                        }
                     }
-                    Ok(None) => {
-                        warn!("Object data not found: {}/{}", bucket, object);
-                        Ok(false)
-                    }
-                    Err(_) => {
-                        warn!("Object data read failed: {}/{}", bucket, object);
+                    Err(e) => {
+                        warn!("Failed to get object reader: {}/{} - {}", bucket, object, e);
                         Ok(false)
                     }
                 }

crates/ahm/src/scanner/data_scanner.rs

@@ -23,22 +23,23 @@ use ecstore::{
     set_disk::SetDisks,
 };
 use rustfs_ecstore::{self as ecstore, StorageAPI, data_usage::store_data_usage_in_backend};
-use rustfs_filemeta::MetacacheReader;
+use rustfs_filemeta::{MetacacheReader, VersionType};
 use tokio::sync::{Mutex, RwLock};
 use tokio_util::sync::CancellationToken;
 use tracing::{debug, error, info, warn};
 
 use super::metrics::{BucketMetrics, DiskMetrics, MetricsCollector, ScannerMetrics};
 use crate::heal::HealManager;
+use crate::scanner::lifecycle::ScannerItem;
 use crate::{
     HealRequest,
     error::{Error, Result},
     get_ahm_services_cancel_token,
 };
-use rustfs_common::{
-    data_usage::DataUsageInfo,
-    metrics::{Metric, Metrics, globalMetrics},
-};
+
+use rustfs_common::data_usage::DataUsageInfo;
+use rustfs_common::metrics::{Metric, Metrics, globalMetrics};
+use rustfs_ecstore::cmd::bucket_targets::VersioningConfig;
 
 use rustfs_ecstore::disk::RUSTFS_META_BUCKET;
 
@@ -290,7 +291,7 @@ impl Scanner {
 
     /// Get global metrics from common crate
     pub async fn get_global_metrics(&self) -> rustfs_madmin::metrics::ScannerMetrics {
-        globalMetrics.report().await
+        (*globalMetrics).report().await
     }
 
     /// Perform a single scan cycle
@@ -317,7 +318,7 @@ impl Scanner {
             cycle_completed: vec![chrono::Utc::now()],
             started: chrono::Utc::now(),
         };
-        globalMetrics.set_cycle(Some(cycle_info)).await;
+        (*globalMetrics).set_cycle(Some(cycle_info)).await;
 
         self.metrics.set_current_cycle(self.state.read().await.current_cycle);
         self.metrics.increment_total_cycles();
@@ -431,8 +432,27 @@ impl Scanner {
         }
 
         if let Some(ecstore) = rustfs_ecstore::new_object_layer_fn() {
-            // First try the standard integrity check
+            // First check whether the object still logically exists.
+            // If it's already deleted (e.g., non-versioned bucket), do not trigger heal.
             let object_opts = ecstore::store_api::ObjectOptions::default();
+            match ecstore.get_object_info(bucket, object, &object_opts).await {
+                Ok(_) => {
+                    // Object exists logically, continue with verification below
+                }
+                Err(e) => {
+                    if matches!(e, ecstore::error::StorageError::ObjectNotFound(_, _)) {
+                        debug!(
+                            "Object {}/{} not found logically (likely deleted), skip integrity check & heal",
+                            bucket, object
+                        );
+                        return Ok(());
+                    } else {
+                        debug!("get_object_info error for {}/{}: {}", bucket, object, e);
+                        // Fall through to existing logic which will handle accordingly
+                    }
+                }
+            }
+            // First try the standard integrity check
             let mut integrity_failed = false;
 
             debug!("Running standard object verification for {}/{}", bucket, object);
@@ -1160,6 +1180,19 @@ impl Scanner {
     /// This method collects all objects from a disk for a specific bucket.
     /// It returns a map of object names to their metadata for later analysis.
     async fn scan_volume(&self, disk: &DiskStore, bucket: &str) -> Result<HashMap<String, rustfs_filemeta::FileMeta>> {
+        let ecstore = match rustfs_ecstore::new_object_layer_fn() {
+            Some(ecstore) => ecstore,
+            None => {
+                error!("ECStore not available");
+                return Err(Error::Other("ECStore not available".to_string()));
+            }
+        };
+        let bucket_info = ecstore.get_bucket_info(bucket, &Default::default()).await.ok();
+        let versioning_config = bucket_info.map(|bi| Arc::new(VersioningConfig { enabled: bi.versioning }));
+        let lifecycle_config = rustfs_ecstore::bucket::metadata_sys::get_lifecycle_config(bucket)
+            .await
+            .ok()
+            .map(|(c, _)| Arc::new(c));
         // Start global metrics collection for volume scan
         let stop_fn = Metrics::time(Metric::ScanObject);
 
@@ -1247,6 +1280,15 @@ impl Scanner {
                             }
                         }
                     } else {
+                        // Apply lifecycle actions
+                        if let Some(lifecycle_config) = &lifecycle_config {
+                            let mut scanner_item =
+                                ScannerItem::new(bucket.to_string(), Some(lifecycle_config.clone()), versioning_config.clone());
+                            if let Err(e) = scanner_item.apply_actions(&entry.name, entry.clone()).await {
+                                error!("Failed to apply lifecycle actions for {}/{}: {}", bucket, entry.name, e);
+                            }
+                        }
+
                         // Store object metadata for later analysis
                         object_metadata.insert(entry.name.clone(), file_meta.clone());
                     }
@@ -1375,8 +1417,64 @@ impl Scanner {
                 let empty_vec = Vec::new();
                 let locations = object_locations.get(&key).unwrap_or(&empty_vec);
 
+                // If any disk reports this object as a latest delete marker (tombstone),
+                // it's a legitimate deletion. Skip missing-object heal to avoid recreating
+                // deleted objects. Optional: a metadata heal could be submitted to fan-out
+                // the delete marker, but we keep it conservative here.
+                let mut has_latest_delete_marker = false;
+                for &disk_idx in locations {
+                    if let Some(bucket_map) = all_disk_objects.get(disk_idx) {
+                        if let Some(file_map) = bucket_map.get(bucket) {
+                            if let Some(fm) = file_map.get(object_name) {
+                                if let Some(first_ver) = fm.versions.first() {
+                                    if first_ver.header.version_type == VersionType::Delete {
+                                        has_latest_delete_marker = true;
+                                        break;
+                                    }
+                                }
+                            }
+                        }
+                    }
+                }
+                if has_latest_delete_marker {
+                    debug!(
+                        "Object {}/{} is a delete marker on some disk(s), skipping heal for missing parts",
+                        bucket, object_name
+                    );
+                    continue;
+                }
+
                 // Check if object is missing from some disks
                 if locations.len() < disks.len() {
+                    // Before submitting heal, confirm the object still exists logically.
+                    let should_heal = if let Some(store) = rustfs_ecstore::new_object_layer_fn() {
+                        match store.get_object_info(bucket, object_name, &Default::default()).await {
+                            Ok(_) => true, // exists -> propagate by heal
+                            Err(e) => {
+                                if matches!(e, rustfs_ecstore::error::StorageError::ObjectNotFound(_, _)) {
+                                    debug!(
+                                        "Object {}/{} not found logically (deleted), skip missing-disks heal",
+                                        bucket, object_name
+                                    );
+                                    false
+                                } else {
+                                    debug!(
+                                        "Object {}/{} get_object_info errored ({}), conservatively skip heal",
+                                        bucket, object_name, e
+                                    );
+                                    false
+                                }
+                            }
+                        }
+                    } else {
+                        // No store available; be conservative and skip to avoid recreating deletions
+                        debug!("No ECStore available to confirm existence, skip heal for {}/{}", bucket, object_name);
+                        false
+                    };
+
+                    if !should_heal {
+                        continue;
+                    }
                     objects_needing_heal += 1;
                     let missing_disks: Vec<usize> = (0..disks.len()).filter(|&i| !locations.contains(&i)).collect();
                     warn!("Object {}/{} missing from disks: {:?}", bucket, object_name, missing_disks);

crates/ahm/src/scanner/lifecycle.rs

@@ -0,0 +1,125 @@
+// 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::Arc;
+
+use rustfs_common::metrics::IlmAction;
+use rustfs_ecstore::bucket::lifecycle::bucket_lifecycle_audit::LcEventSrc;
+use rustfs_ecstore::bucket::lifecycle::bucket_lifecycle_ops::{apply_lifecycle_action, eval_action_from_lifecycle};
+use rustfs_ecstore::bucket::metadata_sys::get_object_lock_config;
+use rustfs_ecstore::cmd::bucket_targets::VersioningConfig;
+use rustfs_ecstore::store_api::ObjectInfo;
+use rustfs_filemeta::FileMetaVersion;
+use rustfs_filemeta::metacache::MetaCacheEntry;
+use s3s::dto::BucketLifecycleConfiguration as LifecycleConfig;
+use tracing::info;
+
+#[derive(Clone)]
+pub struct ScannerItem {
+    bucket: String,
+    lifecycle: Option<Arc<LifecycleConfig>>,
+    versioning: Option<Arc<VersioningConfig>>,
+}
+
+impl ScannerItem {
+    pub fn new(bucket: String, lifecycle: Option<Arc<LifecycleConfig>>, versioning: Option<Arc<VersioningConfig>>) -> Self {
+        Self {
+            bucket,
+            lifecycle,
+            versioning,
+        }
+    }
+
+    pub async fn apply_actions(&mut self, object: &str, mut meta: MetaCacheEntry) -> anyhow::Result<()> {
+        info!("apply_actions called for object: {}", object);
+        if self.lifecycle.is_none() {
+            info!("No lifecycle config for object: {}", object);
+            return Ok(());
+        }
+        info!("Lifecycle config exists for object: {}", object);
+
+        let file_meta = match meta.xl_meta() {
+            Ok(meta) => meta,
+            Err(e) => {
+                tracing::error!("Failed to get xl_meta for {}: {}", object, e);
+                return Ok(());
+            }
+        };
+
+        let latest_version = file_meta.versions.first().cloned().unwrap_or_default();
+        let file_meta_version = FileMetaVersion::try_from(latest_version.meta.as_slice()).unwrap_or_default();
+
+        let obj_info = ObjectInfo {
+            bucket: self.bucket.clone(),
+            name: object.to_string(),
+            version_id: latest_version.header.version_id,
+            mod_time: latest_version.header.mod_time,
+            size: file_meta_version.object.as_ref().map_or(0, |o| o.size),
+            user_defined: serde_json::from_slice(file_meta.data.as_slice()).unwrap_or_default(),
+            ..Default::default()
+        };
+
+        self.apply_lifecycle(&obj_info).await;
+
+        Ok(())
+    }
+
+    async fn apply_lifecycle(&mut self, oi: &ObjectInfo) -> (IlmAction, i64) {
+        let size = oi.size;
+        if self.lifecycle.is_none() {
+            return (IlmAction::NoneAction, size);
+        }
+
+        let (olcfg, rcfg) = if self.bucket != ".minio.sys" {
+            (
+                get_object_lock_config(&self.bucket).await.ok(),
+                None, // FIXME: replication config
+            )
+        } else {
+            (None, None)
+        };
+
+        let lc_evt = eval_action_from_lifecycle(
+            self.lifecycle.as_ref().unwrap(),
+            olcfg
+                .as_ref()
+                .and_then(|(c, _)| c.rule.as_ref().and_then(|r| r.default_retention.clone())),
+            rcfg.clone(),
+            oi,
+        )
+        .await;
+
+        info!("lifecycle: {} Initial scan: {}", oi.name, lc_evt.action);
+
+        let mut new_size = size;
+        match lc_evt.action {
+            IlmAction::DeleteVersionAction | IlmAction::DeleteAllVersionsAction | IlmAction::DelMarkerDeleteAllVersionsAction => {
+                new_size = 0;
+            }
+            IlmAction::DeleteAction => {
+                if let Some(vcfg) = &self.versioning {
+                    if !vcfg.is_enabled() {
+                        new_size = 0;
+                    }
+                } else {
+                    new_size = 0;
+                }
+            }
+            _ => (),
+        }
+
+        apply_lifecycle_action(&lc_evt, &LcEventSrc::Scanner, oi).await;
+        (lc_evt.action, new_size)
+    }
+}

crates/ahm/src/scanner/mod.rs

@@ -14,6 +14,7 @@
 
 pub mod data_scanner;
 pub mod histogram;
+pub mod lifecycle;
 pub mod metrics;
 
 pub use data_scanner::Scanner;

crates/ahm/tests/lifecycle_integration_test.rs

@@ -0,0 +1,243 @@
+use rustfs_ahm::scanner::{Scanner, data_scanner::ScannerConfig};
+use rustfs_ecstore::{
+    bucket::metadata::BUCKET_LIFECYCLE_CONFIG,
+    bucket::metadata_sys,
+    disk::endpoint::Endpoint,
+    endpoints::{EndpointServerPools, Endpoints, PoolEndpoints},
+    store::ECStore,
+    store_api::{ObjectIO, ObjectOptions, PutObjReader, StorageAPI},
+};
+use serial_test::serial;
+use std::sync::Once;
+use std::sync::OnceLock;
+use std::{path::PathBuf, sync::Arc, time::Duration};
+use tokio::fs;
+use tracing::info;
+
+static GLOBAL_ENV: OnceLock<(Vec<PathBuf>, Arc<ECStore>)> = OnceLock::new();
+static INIT: Once = Once::new();
+
+fn init_tracing() {
+    INIT.call_once(|| {
+        let _ = tracing_subscriber::fmt::try_init();
+    });
+}
+
+/// Test helper: Create test environment with ECStore
+async fn setup_test_env() -> (Vec<PathBuf>, Arc<ECStore>) {
+    init_tracing();
+
+    // Fast path: already initialized, just clone and return
+    if let Some((paths, ecstore)) = GLOBAL_ENV.get() {
+        return (paths.clone(), ecstore.clone());
+    }
+
+    // create temp dir as 4 disks with unique base dir
+    let test_base_dir = format!("/tmp/rustfs_ahm_lifecycle_test_{}", uuid::Uuid::new_v4());
+    let temp_dir = std::path::PathBuf::from(&test_base_dir);
+    if temp_dir.exists() {
+        fs::remove_dir_all(&temp_dir).await.ok();
+    }
+    fs::create_dir_all(&temp_dir).await.unwrap();
+
+    // create 4 disk dirs
+    let disk_paths = vec![
+        temp_dir.join("disk1"),
+        temp_dir.join("disk2"),
+        temp_dir.join("disk3"),
+        temp_dir.join("disk4"),
+    ];
+
+    for disk_path in &disk_paths {
+        fs::create_dir_all(disk_path).await.unwrap();
+    }
+
+    // create EndpointServerPools
+    let mut endpoints = Vec::new();
+    for (i, disk_path) in disk_paths.iter().enumerate() {
+        let mut endpoint = Endpoint::try_from(disk_path.to_str().unwrap()).unwrap();
+        // set correct index
+        endpoint.set_pool_index(0);
+        endpoint.set_set_index(0);
+        endpoint.set_disk_index(i);
+        endpoints.push(endpoint);
+    }
+
+    let pool_endpoints = PoolEndpoints {
+        legacy: false,
+        set_count: 1,
+        drives_per_set: 4,
+        endpoints: Endpoints::from(endpoints),
+        cmd_line: "test".to_string(),
+        platform: format!("OS: {} | Arch: {}", std::env::consts::OS, std::env::consts::ARCH),
+    };
+
+    let endpoint_pools = EndpointServerPools(vec![pool_endpoints]);
+
+    // format disks (only first time)
+    rustfs_ecstore::store::init_local_disks(endpoint_pools.clone()).await.unwrap();
+
+    // create ECStore with dynamic port 0 (let OS assign) or fixed 9002 if free
+    let port = 9002; // for simplicity
+    let server_addr: std::net::SocketAddr = format!("127.0.0.1:{port}").parse().unwrap();
+    let ecstore = ECStore::new(server_addr, endpoint_pools).await.unwrap();
+
+    // init bucket metadata system
+    let buckets_list = ecstore
+        .list_bucket(&rustfs_ecstore::store_api::BucketOptions {
+            no_metadata: true,
+            ..Default::default()
+        })
+        .await
+        .unwrap();
+    let buckets = buckets_list.into_iter().map(|v| v.name).collect();
+    rustfs_ecstore::bucket::metadata_sys::init_bucket_metadata_sys(ecstore.clone(), buckets).await;
+
+    // Initialize background expiry workers
+    rustfs_ecstore::bucket::lifecycle::bucket_lifecycle_ops::init_background_expiry(ecstore.clone()).await;
+
+    // Store in global once lock
+    let _ = GLOBAL_ENV.set((disk_paths.clone(), ecstore.clone()));
+
+    (disk_paths, ecstore)
+}
+
+/// Test helper: Create a test bucket
+async fn create_test_bucket(ecstore: &Arc<ECStore>, bucket_name: &str) {
+    (**ecstore)
+        .make_bucket(bucket_name, &Default::default())
+        .await
+        .expect("Failed to create test bucket");
+    info!("Created test bucket: {}", bucket_name);
+}
+
+/// Test helper: Upload test object
+async fn upload_test_object(ecstore: &Arc<ECStore>, bucket: &str, object: &str, data: &[u8]) {
+    let mut reader = PutObjReader::from_vec(data.to_vec());
+    let object_info = (**ecstore)
+        .put_object(bucket, object, &mut reader, &ObjectOptions::default())
+        .await
+        .expect("Failed to upload test object");
+
+    info!("Uploaded test object: {}/{} ({} bytes)", bucket, object, object_info.size);
+}
+
+/// Test helper: Set bucket lifecycle configuration
+async fn set_bucket_lifecycle(bucket_name: &str) -> Result<(), Box<dyn std::error::Error>> {
+    // Create a simple lifecycle configuration XML with 0 days expiry for immediate testing
+    let lifecycle_xml = r#"<?xml version="1.0" encoding="UTF-8"?>
+<LifecycleConfiguration>
+    <Rule>
+        <ID>test-rule</ID>
+        <Status>Enabled</Status>
+        <Filter>
+            <Prefix>test/</Prefix>
+        </Filter>
+        <Expiration>
+            <Days>0</Days>
+        </Expiration>
+    </Rule>
+</LifecycleConfiguration>"#;
+
+    metadata_sys::update(bucket_name, BUCKET_LIFECYCLE_CONFIG, lifecycle_xml.as_bytes().to_vec()).await?;
+
+    Ok(())
+}
+
+/// Test helper: Check if object exists
+async fn object_exists(ecstore: &Arc<ECStore>, bucket: &str, object: &str) -> bool {
+    ((**ecstore).get_object_info(bucket, object, &ObjectOptions::default()).await).is_ok()
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 4)]
+#[serial]
+async fn test_lifecycle_expiry_basic() {
+    let (_disk_paths, ecstore) = setup_test_env().await;
+
+    // Create test bucket and object
+    let bucket_name = "test-lifecycle-bucket";
+    let object_name = "test/object.txt"; // Match the lifecycle rule prefix "test/"
+    let test_data = b"Hello, this is test data for lifecycle expiry!";
+
+    create_test_bucket(&ecstore, bucket_name).await;
+    upload_test_object(&ecstore, bucket_name, object_name, test_data).await;
+
+    // Verify object exists initially
+    assert!(object_exists(&ecstore, bucket_name, object_name).await);
+    println!("✅ Object exists before lifecycle processing");
+
+    // Set lifecycle configuration with very short expiry (0 days = immediate expiry)
+    set_bucket_lifecycle(bucket_name)
+        .await
+        .expect("Failed to set lifecycle configuration");
+    println!("✅ Lifecycle configuration set for bucket: {bucket_name}");
+
+    // Verify lifecycle configuration was set
+    match rustfs_ecstore::bucket::metadata_sys::get(bucket_name).await {
+        Ok(bucket_meta) => {
+            assert!(bucket_meta.lifecycle_config.is_some());
+            println!("✅ Bucket metadata retrieved successfully");
+        }
+        Err(e) => {
+            println!("❌ Error retrieving bucket metadata: {e:?}");
+        }
+    }
+
+    // Create scanner with very short intervals for testing
+    let scanner_config = ScannerConfig {
+        scan_interval: Duration::from_millis(100),
+        deep_scan_interval: Duration::from_millis(500),
+        max_concurrent_scans: 1,
+        ..Default::default()
+    };
+
+    let scanner = Scanner::new(Some(scanner_config), None);
+
+    // Start scanner
+    scanner.start().await.expect("Failed to start scanner");
+    println!("✅ Scanner started");
+
+    // Wait for scanner to process lifecycle rules
+    tokio::time::sleep(Duration::from_secs(2)).await;
+
+    // Manually trigger a scan cycle to ensure lifecycle processing
+    scanner.scan_cycle().await.expect("Failed to trigger scan cycle");
+    println!("✅ Manual scan cycle completed");
+
+    // Wait a bit more for background workers to process expiry tasks
+    tokio::time::sleep(Duration::from_secs(5)).await;
+
+    // Check if object has been expired (deleted)
+    let object_still_exists = object_exists(&ecstore, bucket_name, object_name).await;
+    println!("Object exists after lifecycle processing: {object_still_exists}");
+
+    if object_still_exists {
+        println!("❌ Object was not deleted by lifecycle processing");
+        // Let's try to get object info to see its details
+        match ecstore
+            .get_object_info(bucket_name, object_name, &rustfs_ecstore::store_api::ObjectOptions::default())
+            .await
+        {
+            Ok(obj_info) => {
+                println!(
+                    "Object info: name={}, size={}, mod_time={:?}",
+                    obj_info.name, obj_info.size, obj_info.mod_time
+                );
+            }
+            Err(e) => {
+                println!("Error getting object info: {e:?}");
+            }
+        }
+    } else {
+        println!("✅ Object was successfully deleted by lifecycle processing");
+    }
+
+    assert!(!object_still_exists);
+    println!("✅ Object successfully expired");
+
+    // Stop scanner
+    let _ = scanner.stop().await;
+    println!("✅ Scanner stopped");
+
+    println!("Lifecycle expiry basic test completed");
+}

crates/checksums/Cargo.toml

@@ -28,18 +28,11 @@ documentation = "https://docs.rs/rustfs-signer/latest/rustfs_checksum/"
 [dependencies]
 bytes = { workspace = true }
 crc-fast = { workspace = true }
-hex = { workspace = true }
 http = { workspace = true }
-http-body = { workspace = true }
 base64-simd = { workspace = true }
 md-5 = { workspace = true }
-pin-project-lite = { workspace = true }
 sha1 = { workspace = true }
 sha2 = { workspace = true }
-tracing = { workspace = true }
 
 [dev-dependencies]
-bytes-utils = { workspace = true }
 pretty_assertions = { workspace = true }
-tracing-test = { workspace = true }
-tokio = { workspace = true, features = ["macros", "rt"] }

crates/checksums/src/http.rs

@@ -24,6 +24,7 @@ pub const SHA_1_HEADER_NAME: &str = "x-amz-checksum-sha1";
 pub const SHA_256_HEADER_NAME: &str = "x-amz-checksum-sha256";
 pub const CRC_64_NVME_HEADER_NAME: &str = "x-amz-checksum-crc64nvme";
 
+#[allow(dead_code)]
 pub(crate) static MD5_HEADER_NAME: &str = "content-md5";
 
 pub const CHECKSUM_ALGORITHMS_IN_PRIORITY_ORDER: [&str; 5] =

crates/checksums/src/lib.rs

@@ -294,7 +294,7 @@ impl Checksum for Sha256 {
         Self::size()
     }
 }
-
+#[allow(dead_code)]
 #[derive(Debug, Default)]
 struct Md5 {
     hasher: md5::Md5,

crates/config/src/constants/env.rs

@@ -19,3 +19,265 @@ pub const ENV_WORD_DELIMITER: &str = "_";
 /// Medium-drawn lines separator
 /// This is used to separate words in environment variable names.
 pub const ENV_WORD_DELIMITER_DASH: &str = "-";
+
+#[derive(Debug, PartialEq, Eq, Clone, Copy, Default)]
+pub enum EnableState {
+    True,
+    False,
+    #[default]
+    Empty,
+    Yes,
+    No,
+    On,
+    Off,
+    Enabled,
+    Disabled,
+    Ok,
+    NotOk,
+    Success,
+    Failure,
+    Active,
+    Inactive,
+    One,
+    Zero,
+}
+impl std::fmt::Display for EnableState {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        f.write_str(self.as_str())
+    }
+}
+impl std::str::FromStr for EnableState {
+    type Err = ();
+    fn from_str(s: &str) -> Result<Self, Self::Err> {
+        match s.trim() {
+            s if s.eq_ignore_ascii_case("true") => Ok(EnableState::True),
+            s if s.eq_ignore_ascii_case("false") => Ok(EnableState::False),
+            "" => Ok(EnableState::Empty),
+            s if s.eq_ignore_ascii_case("yes") => Ok(EnableState::Yes),
+            s if s.eq_ignore_ascii_case("no") => Ok(EnableState::No),
+            s if s.eq_ignore_ascii_case("on") => Ok(EnableState::On),
+            s if s.eq_ignore_ascii_case("off") => Ok(EnableState::Off),
+            s if s.eq_ignore_ascii_case("enabled") => Ok(EnableState::Enabled),
+            s if s.eq_ignore_ascii_case("disabled") => Ok(EnableState::Disabled),
+            s if s.eq_ignore_ascii_case("ok") => Ok(EnableState::Ok),
+            s if s.eq_ignore_ascii_case("not_ok") => Ok(EnableState::NotOk),
+            s if s.eq_ignore_ascii_case("success") => Ok(EnableState::Success),
+            s if s.eq_ignore_ascii_case("failure") => Ok(EnableState::Failure),
+            s if s.eq_ignore_ascii_case("active") => Ok(EnableState::Active),
+            s if s.eq_ignore_ascii_case("inactive") => Ok(EnableState::Inactive),
+            "1" => Ok(EnableState::One),
+            "0" => Ok(EnableState::Zero),
+            _ => Err(()),
+        }
+    }
+}
+
+impl EnableState {
+    /// Returns the default value for the enum.
+    pub fn get_default() -> Self {
+        Self::default()
+    }
+    /// Returns the string representation of the enum.
+    pub fn as_str(&self) -> &str {
+        match self {
+            EnableState::True => "true",
+            EnableState::False => "false",
+            EnableState::Empty => "",
+            EnableState::Yes => "yes",
+            EnableState::No => "no",
+            EnableState::On => "on",
+            EnableState::Off => "off",
+            EnableState::Enabled => "enabled",
+            EnableState::Disabled => "disabled",
+            EnableState::Ok => "ok",
+            EnableState::NotOk => "not_ok",
+            EnableState::Success => "success",
+            EnableState::Failure => "failure",
+            EnableState::Active => "active",
+            EnableState::Inactive => "inactive",
+            EnableState::One => "1",
+            EnableState::Zero => "0",
+        }
+    }
+
+    /// is_enabled checks if the state represents an enabled condition.
+    pub fn is_enabled(self) -> bool {
+        matches!(
+            self,
+            EnableState::True
+                | EnableState::Yes
+                | EnableState::On
+                | EnableState::Enabled
+                | EnableState::Ok
+                | EnableState::Success
+                | EnableState::Active
+                | EnableState::One
+        )
+    }
+
+    /// is_disabled checks if the state represents a disabled condition.
+    pub fn is_disabled(self) -> bool {
+        matches!(
+            self,
+            EnableState::False
+                | EnableState::No
+                | EnableState::Off
+                | EnableState::Disabled
+                | EnableState::NotOk
+                | EnableState::Failure
+                | EnableState::Inactive
+                | EnableState::Zero
+                | EnableState::Empty
+        )
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use std::str::FromStr;
+    #[test]
+    fn test_enable_state_display_and_fromstr() {
+        let cases = [
+            (EnableState::True, "true"),
+            (EnableState::False, "false"),
+            (EnableState::Empty, ""),
+            (EnableState::Yes, "yes"),
+            (EnableState::No, "no"),
+            (EnableState::On, "on"),
+            (EnableState::Off, "off"),
+            (EnableState::Enabled, "enabled"),
+            (EnableState::Disabled, "disabled"),
+            (EnableState::Ok, "ok"),
+            (EnableState::NotOk, "not_ok"),
+            (EnableState::Success, "success"),
+            (EnableState::Failure, "failure"),
+            (EnableState::Active, "active"),
+            (EnableState::Inactive, "inactive"),
+            (EnableState::One, "1"),
+            (EnableState::Zero, "0"),
+        ];
+        for (variant, string) in cases.iter() {
+            assert_eq!(&variant.to_string(), string);
+            assert_eq!(EnableState::from_str(string).unwrap(), *variant);
+        }
+        // Test invalid string
+        assert!(EnableState::from_str("invalid").is_err());
+    }
+    #[test]
+    fn test_enable_state_enum() {
+        let cases = [
+            (EnableState::True, "true"),
+            (EnableState::False, "false"),
+            (EnableState::Empty, ""),
+            (EnableState::Yes, "yes"),
+            (EnableState::No, "no"),
+            (EnableState::On, "on"),
+            (EnableState::Off, "off"),
+            (EnableState::Enabled, "enabled"),
+            (EnableState::Disabled, "disabled"),
+            (EnableState::Ok, "ok"),
+            (EnableState::NotOk, "not_ok"),
+            (EnableState::Success, "success"),
+            (EnableState::Failure, "failure"),
+            (EnableState::Active, "active"),
+            (EnableState::Inactive, "inactive"),
+            (EnableState::One, "1"),
+            (EnableState::Zero, "0"),
+        ];
+        for (variant, string) in cases.iter() {
+            assert_eq!(variant.to_string(), *string);
+        }
+    }
+
+    #[test]
+    fn test_enable_state_enum_from_str() {
+        let cases = [
+            ("true", EnableState::True),
+            ("false", EnableState::False),
+            ("", EnableState::Empty),
+            ("yes", EnableState::Yes),
+            ("no", EnableState::No),
+            ("on", EnableState::On),
+            ("off", EnableState::Off),
+            ("enabled", EnableState::Enabled),
+            ("disabled", EnableState::Disabled),
+            ("ok", EnableState::Ok),
+            ("not_ok", EnableState::NotOk),
+            ("success", EnableState::Success),
+            ("failure", EnableState::Failure),
+            ("active", EnableState::Active),
+            ("inactive", EnableState::Inactive),
+            ("1", EnableState::One),
+            ("0", EnableState::Zero),
+        ];
+        for (string, variant) in cases.iter() {
+            assert_eq!(EnableState::from_str(string).unwrap(), *variant);
+        }
+    }
+
+    #[test]
+    fn test_enable_state_default() {
+        let default_state = EnableState::get_default();
+        assert_eq!(default_state, EnableState::Empty);
+        assert_eq!(default_state.as_str(), "");
+    }
+
+    #[test]
+    fn test_enable_state_as_str() {
+        let cases = [
+            (EnableState::True, "true"),
+            (EnableState::False, "false"),
+            (EnableState::Empty, ""),
+            (EnableState::Yes, "yes"),
+            (EnableState::No, "no"),
+            (EnableState::On, "on"),
+            (EnableState::Off, "off"),
+            (EnableState::Enabled, "enabled"),
+            (EnableState::Disabled, "disabled"),
+            (EnableState::Ok, "ok"),
+            (EnableState::NotOk, "not_ok"),
+            (EnableState::Success, "success"),
+            (EnableState::Failure, "failure"),
+            (EnableState::Active, "active"),
+            (EnableState::Inactive, "inactive"),
+            (EnableState::One, "1"),
+            (EnableState::Zero, "0"),
+        ];
+        for (variant, string) in cases.iter() {
+            assert_eq!(variant.as_str(), *string);
+        }
+    }
+
+    #[test]
+    fn test_enable_state_is_enabled() {
+        let enabled_states = [
+            EnableState::True,
+            EnableState::Yes,
+            EnableState::On,
+            EnableState::Enabled,
+            EnableState::Ok,
+            EnableState::Success,
+            EnableState::Active,
+            EnableState::One,
+        ];
+        for state in enabled_states.iter() {
+            assert!(state.is_enabled());
+        }
+
+        let disabled_states = [
+            EnableState::False,
+            EnableState::No,
+            EnableState::Off,
+            EnableState::Disabled,
+            EnableState::NotOk,
+            EnableState::Failure,
+            EnableState::Inactive,
+            EnableState::Zero,
+            EnableState::Empty,
+        ];
+        for state in disabled_states.iter() {
+            assert!(state.is_disabled());
+        }
+    }
+}

crates/config/src/constants/mod.rs

@@ -12,5 +12,6 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-pub(crate) mod app;
-pub(crate) mod env;
+pub mod app;
+pub mod env;
+pub mod tls;

crates/config/src/constants/tls.rs

@@ -0,0 +1,15 @@
+//  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 ENV_TLS_KEYLOG: &str = "RUSTFS_TLS_KEYLOG";

crates/config/src/lib.rs

@@ -18,6 +18,8 @@ pub mod constants;
 pub use constants::app::*;
 #[cfg(feature = "constants")]
 pub use constants::env::*;
+#[cfg(feature = "constants")]
+pub use constants::tls::*;
 #[cfg(feature = "notify")]
 pub mod notify;
 #[cfg(feature = "observability")]

crates/e2e_test/src/reliant/lifecycle.rs

@@ -0,0 +1,133 @@
+#![cfg(test)]
+// 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 aws_config::meta::region::RegionProviderChain;
+use aws_sdk_s3::Client;
+use aws_sdk_s3::config::{Credentials, Region};
+use bytes::Bytes;
+use serial_test::serial;
+use std::error::Error;
+use tokio::time::sleep;
+
+const ENDPOINT: &str = "http://localhost:9000";
+const ACCESS_KEY: &str = "rustfsadmin";
+const SECRET_KEY: &str = "rustfsadmin";
+const BUCKET: &str = "test-basic-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)
+}
+
+async fn setup_test_bucket(client: &Client) -> Result<(), Box<dyn Error>> {
+    match client.create_bucket().bucket(BUCKET).send().await {
+        Ok(_) => {}
+        Err(e) => {
+            let error_str = e.to_string();
+            if !error_str.contains("BucketAlreadyOwnedByYou") && !error_str.contains("BucketAlreadyExists") {
+                return Err(e.into());
+            }
+        }
+    }
+    Ok(())
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+#[serial]
+#[ignore = "requires running RustFS server at localhost:9000"]
+async fn test_bucket_lifecycle_configuration() -> Result<(), Box<dyn std::error::Error>> {
+    use aws_sdk_s3::types::{BucketLifecycleConfiguration, LifecycleExpiration, LifecycleRule, LifecycleRuleFilter};
+    use tokio::time::Duration;
+
+    let client = create_aws_s3_client().await?;
+    setup_test_bucket(&client).await?;
+
+    // Upload test object first
+    let test_content = "Test object for lifecycle expiration";
+    let lifecycle_object_key = "lifecycle-test-object.txt";
+    client
+        .put_object()
+        .bucket(BUCKET)
+        .key(lifecycle_object_key)
+        .body(Bytes::from(test_content.as_bytes()).into())
+        .send()
+        .await?;
+
+    // Verify object exists initially
+    let resp = client.get_object().bucket(BUCKET).key(lifecycle_object_key).send().await?;
+    assert!(resp.content_length().unwrap_or(0) > 0);
+
+    // Configure lifecycle rule: expire after current time + 3 seconds
+    let expiration = LifecycleExpiration::builder().days(0).build();
+    let filter = LifecycleRuleFilter::builder().prefix(lifecycle_object_key).build();
+    let rule = LifecycleRule::builder()
+        .id("expire-test-object")
+        .filter(filter)
+        .expiration(expiration)
+        .status(aws_sdk_s3::types::ExpirationStatus::Enabled)
+        .build()?;
+    let lifecycle = BucketLifecycleConfiguration::builder().rules(rule).build()?;
+
+    client
+        .put_bucket_lifecycle_configuration()
+        .bucket(BUCKET)
+        .lifecycle_configuration(lifecycle)
+        .send()
+        .await?;
+
+    // Verify lifecycle configuration was set
+    let resp = client.get_bucket_lifecycle_configuration().bucket(BUCKET).send().await?;
+    let rules = resp.rules();
+    assert!(rules.iter().any(|r| r.id().unwrap_or("") == "expire-test-object"));
+
+    // Wait for lifecycle processing (scanner runs every 1 second)
+    sleep(Duration::from_secs(3)).await;
+
+    // After lifecycle processing, the object should be deleted by the lifecycle rule
+    let get_result = client.get_object().bucket(BUCKET).key(lifecycle_object_key).send().await;
+
+    match get_result {
+        Ok(_) => {
+            panic!("Expected object to be deleted by lifecycle rule, but it still exists");
+        }
+        Err(e) => {
+            if let Some(service_error) = e.as_service_error() {
+                if service_error.is_no_such_key() {
+                    println!("Lifecycle configuration test completed - object was successfully deleted by lifecycle rule");
+                } else {
+                    panic!("Expected NoSuchKey error, but got: {e:?}");
+                }
+            } else {
+                panic!("Expected service error, but got: {e:?}");
+            }
+        }
+    }
+
+    println!("Lifecycle configuration test completed.");
+    Ok(())
+}

crates/e2e_test/src/reliant/lock.rs

@@ -38,6 +38,79 @@ fn get_cluster_endpoints() -> Vec<Endpoint> {
     }]
 }
 
+#[tokio::test]
+#[serial]
+#[ignore = "requires running RustFS server at localhost:9000"]
+async fn test_guard_drop_releases_exclusive_lock_local() -> Result<(), Box<dyn Error>> {
+    // Single local client; no external server required
+    let client: Arc<dyn LockClient> = Arc::new(LocalClient::new());
+    let ns_lock = NamespaceLock::with_clients("e2e_guard_local".to_string(), vec![client]);
+
+    // Acquire exclusive guard
+    let g1 = ns_lock
+        .lock_guard("guard_exclusive", "owner1", Duration::from_millis(100), Duration::from_secs(5))
+        .await?;
+    assert!(g1.is_some(), "first guard acquisition should succeed");
+
+    // While g1 is alive, second exclusive acquisition should fail
+    let g2 = ns_lock
+        .lock_guard("guard_exclusive", "owner2", Duration::from_millis(50), Duration::from_secs(5))
+        .await?;
+    assert!(g2.is_none(), "second guard acquisition should fail while first is held");
+
+    // Drop first guard to trigger background release
+    drop(g1);
+    // Give the background unlock worker a short moment to process
+    sleep(Duration::from_millis(80)).await;
+
+    // Now acquisition should succeed
+    let g3 = ns_lock
+        .lock_guard("guard_exclusive", "owner2", Duration::from_millis(100), Duration::from_secs(5))
+        .await?;
+    assert!(g3.is_some(), "acquisition should succeed after guard drop releases the lock");
+    drop(g3);
+
+    Ok(())
+}
+
+#[tokio::test]
+#[serial]
+#[ignore = "requires running RustFS server at localhost:9000"]
+async fn test_guard_shared_then_write_after_drop() -> Result<(), Box<dyn Error>> {
+    // Two shared read guards should coexist; write should be blocked until they drop
+    let client: Arc<dyn LockClient> = Arc::new(LocalClient::new());
+    let ns_lock = NamespaceLock::with_clients("e2e_guard_rw".to_string(), vec![client]);
+
+    // Acquire two read guards
+    let r1 = ns_lock
+        .rlock_guard("rw_resource", "reader1", Duration::from_millis(100), Duration::from_secs(5))
+        .await?;
+    let r2 = ns_lock
+        .rlock_guard("rw_resource", "reader2", Duration::from_millis(100), Duration::from_secs(5))
+        .await?;
+    assert!(r1.is_some() && r2.is_some(), "both read guards should be acquired");
+
+    // Attempt write while readers hold the lock should fail
+    let w_fail = ns_lock
+        .lock_guard("rw_resource", "writer", Duration::from_millis(50), Duration::from_secs(5))
+        .await?;
+    assert!(w_fail.is_none(), "write should be blocked when read guards are active");
+
+    // Drop read guards to release
+    drop(r1);
+    drop(r2);
+    sleep(Duration::from_millis(80)).await;
+
+    // Now write should succeed
+    let w_ok = ns_lock
+        .lock_guard("rw_resource", "writer", Duration::from_millis(150), Duration::from_secs(5))
+        .await?;
+    assert!(w_ok.is_some(), "write should succeed after read guards are dropped");
+    drop(w_ok);
+
+    Ok(())
+}
+
 #[tokio::test]
 #[serial]
 #[ignore = "requires running RustFS server at localhost:9000"]

crates/e2e_test/src/reliant/mod.rs

@@ -12,6 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+mod lifecycle;
 mod lock;
 mod node_interact_test;
 mod sql;

crates/ecstore/Cargo.toml

@@ -69,7 +69,6 @@ hmac = { workspace = true }
 sha1 = { workspace = true }
 sha2 = { workspace = true }
 hex-simd = { workspace = true }
-path-clean = { workspace = true }
 tempfile.workspace = true
 hyper.workspace = true
 hyper-util.workspace = true
@@ -123,4 +122,4 @@ harness = false
 
 [[bench]]
 name = "comparison_benchmark"
-harness = false
+harness = false

crates/ecstore/src/bucket/lifecycle/bucket_lifecycle_ops.rs

@@ -516,7 +516,7 @@ impl TransitionState {
                         if let Err(err) = transition_object(api.clone(), &task.obj_info, LcAuditEvent::new(task.event.clone(), task.src.clone())).await {
                             if !is_err_version_not_found(&err) && !is_err_object_not_found(&err) && !is_network_or_host_down(&err.to_string(), false) && !err.to_string().contains("use of closed network connection") {
                                 error!("Transition to {} failed for {}/{} version:{} with {}",
-                                    task.event.storage_class, task.obj_info.bucket, task.obj_info.name, task.obj_info.version_id.expect("err"), err.to_string());
+                                    task.event.storage_class, task.obj_info.bucket, task.obj_info.name, task.obj_info.version_id.map(|v| v.to_string()).unwrap_or_default(), err.to_string());
                             }
                         } else {
                             let mut ts = TierStats {
@@ -743,7 +743,7 @@ pub async fn transition_object(api: Arc<ECStore>, oi: &ObjectInfo, lae: LcAuditE
             ..Default::default()
         },
         //lifecycle_audit_event: lae,
-        version_id: Some(oi.version_id.expect("err").to_string()),
+        version_id: oi.version_id.map(|v| v.to_string()),
         versioned: BucketVersioningSys::prefix_enabled(&oi.bucket, &oi.name).await,
         version_suspended: BucketVersioningSys::prefix_suspended(&oi.bucket, &oi.name).await,
         mod_time: oi.mod_time,
@@ -808,7 +808,7 @@ impl LifecycleOps for ObjectInfo {
         lifecycle::ObjectOpts {
             name: self.name.clone(),
             user_tags: self.user_tags.clone(),
-            version_id: self.version_id.expect("err").to_string(),
+            version_id: self.version_id.map(|v| v.to_string()).unwrap_or_default(),
             mod_time: self.mod_time,
             size: self.size as usize,
             is_latest: self.is_latest,
@@ -874,7 +874,11 @@ pub async fn eval_action_from_lifecycle(
             if lock_enabled && enforce_retention_for_deletion(oi) {
                 //if serverDebugLog {
                 if oi.version_id.is_some() {
-                    info!("lifecycle: {} v({}) is locked, not deleting", oi.name, oi.version_id.expect("err"));
+                    info!(
+                        "lifecycle: {} v({}) is locked, not deleting",
+                        oi.name,
+                        oi.version_id.map(|v| v.to_string()).unwrap_or_default()
+                    );
                 } else {
                     info!("lifecycle: {} is locked, not deleting", oi.name);
                 }
@@ -928,7 +932,7 @@ pub async fn apply_expiry_on_non_transitioned_objects(
     };
 
     if lc_event.action.delete_versioned() {
-        opts.version_id = Some(oi.version_id.expect("err").to_string());
+        opts.version_id = oi.version_id.map(|v| v.to_string());
     }
 
     opts.versioned = BucketVersioningSys::prefix_enabled(&oi.bucket, &oi.name).await;

crates/ecstore/src/bucket/lifecycle/lifecycle.rs

@@ -27,6 +27,7 @@ use std::env;
 use std::fmt::Display;
 use time::macros::{datetime, offset};
 use time::{self, Duration, OffsetDateTime};
+use tracing::info;
 
 use crate::bucket::lifecycle::rule::TransitionOps;
 
@@ -279,7 +280,12 @@ impl Lifecycle for BucketLifecycleConfiguration {
 
     async fn eval_inner(&self, obj: &ObjectOpts, now: OffsetDateTime) -> Event {
         let mut events = Vec::<Event>::new();
+        info!(
+            "eval_inner: object={}, mod_time={:?}, now={:?}, is_latest={}, delete_marker={}",
+            obj.name, obj.mod_time, now, obj.is_latest, obj.delete_marker
+        );
         if obj.mod_time.expect("err").unix_timestamp() == 0 {
+            info!("eval_inner: mod_time is 0, returning default event");
             return Event::default();
         }
 
@@ -418,7 +424,16 @@ impl Lifecycle for BucketLifecycleConfiguration {
                     }
                 }
 
-                if obj.is_latest && !obj.delete_marker {
+                info!(
+                    "eval_inner: checking expiration condition - is_latest={}, delete_marker={}, version_id={:?}, condition_met={}",
+                    obj.is_latest,
+                    obj.delete_marker,
+                    obj.version_id,
+                    (obj.is_latest || obj.version_id.is_empty()) && !obj.delete_marker
+                );
+                // Allow expiration for latest objects OR non-versioned objects (empty version_id)
+                if (obj.is_latest || obj.version_id.is_empty()) && !obj.delete_marker {
+                    info!("eval_inner: entering expiration check");
                     if let Some(ref expiration) = rule.expiration {
                         if let Some(ref date) = expiration.date {
                             let date0 = OffsetDateTime::from(date.clone());
@@ -435,22 +450,29 @@ impl Lifecycle for BucketLifecycleConfiguration {
                                 });
                             }
                         } else if let Some(days) = expiration.days {
-                            if days != 0 {
-                                let expected_expiry: OffsetDateTime = expected_expiry_time(obj.mod_time.expect("err!"), days);
-                                if now.unix_timestamp() == 0 || now.unix_timestamp() > expected_expiry.unix_timestamp() {
-                                    let mut event = Event {
-                                        action: IlmAction::DeleteAction,
-                                        rule_id: rule.id.clone().expect("err!"),
-                                        due: Some(expected_expiry),
-                                        noncurrent_days: 0,
-                                        newer_noncurrent_versions: 0,
-                                        storage_class: "".into(),
-                                    };
-                                    /*if rule.expiration.expect("err!").delete_all.val {
-                                        event.action = IlmAction::DeleteAllVersionsAction
-                                    }*/
-                                    events.push(event);
-                                }
+                            let expected_expiry: OffsetDateTime = expected_expiry_time(obj.mod_time.expect("err!"), days);
+                            info!(
+                                "eval_inner: expiration check - days={}, obj_time={:?}, expiry_time={:?}, now={:?}, should_expire={}",
+                                days,
+                                obj.mod_time.expect("err!"),
+                                expected_expiry,
+                                now,
+                                now.unix_timestamp() > expected_expiry.unix_timestamp()
+                            );
+                            if now.unix_timestamp() == 0 || now.unix_timestamp() > expected_expiry.unix_timestamp() {
+                                info!("eval_inner: object should expire, adding DeleteAction");
+                                let mut event = Event {
+                                    action: IlmAction::DeleteAction,
+                                    rule_id: rule.id.clone().expect("err!"),
+                                    due: Some(expected_expiry),
+                                    noncurrent_days: 0,
+                                    newer_noncurrent_versions: 0,
+                                    storage_class: "".into(),
+                                };
+                                /*if rule.expiration.expect("err!").delete_all.val {
+                                    event.action = IlmAction::DeleteAllVersionsAction
+                                }*/
+                                events.push(event);
                             }
                         }
                     }
@@ -598,7 +620,7 @@ impl LifecycleCalculate for Transition {
 
 pub fn expected_expiry_time(mod_time: OffsetDateTime, days: i32) -> OffsetDateTime {
     if days == 0 {
-        return mod_time;
+        return OffsetDateTime::UNIX_EPOCH; // Return epoch time to ensure immediate expiry
     }
     let t = mod_time
         .to_offset(offset!(-0:00:00))

crates/ecstore/src/bucket/object_lock/objectlock.rs

@@ -54,8 +54,8 @@ pub fn get_object_retention_meta(meta: HashMap<String, String>) -> ObjectLockRet
     }
     if let Some(till_str) = till_str {
         let t = OffsetDateTime::parse(till_str, &format_description::well_known::Iso8601::DEFAULT);
-        if t.is_err() {
-            retain_until_date = Date::from(t.expect("err")); //TODO: utc
+        if let Ok(parsed_time) = t {
+            retain_until_date = Date::from(parsed_time);
         }
     }
     ObjectLockRetention {

crates/ecstore/src/cmd/bucket_replication.rs

@@ -1897,7 +1897,7 @@ impl ReplicationState {
         } else if !self.replica_status.is_empty() {
             self.replica_status.clone()
         } else {
-            return ReplicationStatusType::Unknown;
+            ReplicationStatusType::Unknown
         }
     }
 

crates/ecstore/src/disk/local.rs

@@ -953,9 +953,8 @@ impl LocalDisk {
             let name = path_join_buf(&[current, entry]);
 
             if !dir_stack.is_empty() {
-                if let Some(pop) = dir_stack.pop() {
+                if let Some(pop) = dir_stack.last().cloned() {
                     if pop < name {
-                        //
                         out.write_obj(&MetaCacheEntry {
                             name: pop.clone(),
                             ..Default::default()
@@ -969,6 +968,7 @@ impl LocalDisk {
                                 error!("scan_dir err {:?}", er);
                             }
                         }
+                        dir_stack.pop();
                     }
                 }
             }

crates/ecstore/src/error.rs

@@ -156,11 +156,12 @@ pub enum StorageError {
     #[error("Object exists on :{0} as directory {1}")]
     ObjectExistsAsDirectory(String, String),
 
-    // #[error("Storage resources are insufficient for the read operation")]
-    // InsufficientReadQuorum,
+    #[error("Storage resources are insufficient for the read operation: {0}/{1}")]
+    InsufficientReadQuorum(String, String),
+
+    #[error("Storage resources are insufficient for the write operation: {0}/{1}")]
+    InsufficientWriteQuorum(String, String),
 
-    // #[error("Storage resources are insufficient for the write operation")]
-    // InsufficientWriteQuorum,
     #[error("Decommission not started")]
     DecommissionNotStarted,
     #[error("Decommission already running")]
@@ -413,6 +414,8 @@ impl Clone for StorageError {
             StorageError::TooManyOpenFiles => StorageError::TooManyOpenFiles,
             StorageError::NoHealRequired => StorageError::NoHealRequired,
             StorageError::Lock(e) => StorageError::Lock(e.clone()),
+            StorageError::InsufficientReadQuorum(a, b) => StorageError::InsufficientReadQuorum(a.clone(), b.clone()),
+            StorageError::InsufficientWriteQuorum(a, b) => StorageError::InsufficientWriteQuorum(a.clone(), b.clone()),
         }
     }
 }
@@ -476,6 +479,8 @@ impl StorageError {
             StorageError::TooManyOpenFiles => 0x36,
             StorageError::NoHealRequired => 0x37,
             StorageError::Lock(_) => 0x38,
+            StorageError::InsufficientReadQuorum(_, _) => 0x39,
+            StorageError::InsufficientWriteQuorum(_, _) => 0x3A,
         }
     }
 
@@ -541,6 +546,8 @@ impl StorageError {
             0x36 => Some(StorageError::TooManyOpenFiles),
             0x37 => Some(StorageError::NoHealRequired),
             0x38 => Some(StorageError::Lock(rustfs_lock::LockError::internal("Generic lock error".to_string()))),
+            0x39 => Some(StorageError::InsufficientReadQuorum(Default::default(), Default::default())),
+            0x3A => Some(StorageError::InsufficientWriteQuorum(Default::default(), Default::default())),
             _ => None,
         }
     }
@@ -753,6 +760,17 @@ pub fn to_object_err(err: Error, params: Vec<&str>) -> Error {
             StorageError::PrefixAccessDenied(bucket, object)
         }
 
+        StorageError::ErasureReadQuorum => {
+            let bucket = params.first().cloned().unwrap_or_default().to_owned();
+            let object = params.get(1).cloned().map(decode_dir_object).unwrap_or_default();
+            StorageError::InsufficientReadQuorum(bucket, object)
+        }
+        StorageError::ErasureWriteQuorum => {
+            let bucket = params.first().cloned().unwrap_or_default().to_owned();
+            let object = params.get(1).cloned().map(decode_dir_object).unwrap_or_default();
+            StorageError::InsufficientWriteQuorum(bucket, object)
+        }
+
         _ => err,
     }
 }

crates/ecstore/src/set_disk.rs

@@ -17,6 +17,8 @@
 
 use crate::bitrot::{create_bitrot_reader, create_bitrot_writer};
 use crate::bucket::lifecycle::lifecycle::TRANSITION_COMPLETE;
+use crate::bucket::versioning::VersioningApi;
+use crate::bucket::versioning_sys::BucketVersioningSys;
 use crate::client::{object_api_utils::extract_etag, transition_api::ReaderImpl};
 use crate::disk::STORAGE_FORMAT_FILE;
 use crate::disk::error_reduce::{OBJECT_OP_IGNORED_ERRS, reduce_read_quorum_errs, reduce_write_quorum_errs};
@@ -2027,6 +2029,24 @@ impl SetDisks {
 
         Ok((fi, parts_metadata, op_online_disks))
     }
+    async fn get_object_info_and_quorum(&self, bucket: &str, object: &str, opts: &ObjectOptions) -> Result<(ObjectInfo, usize)> {
+        let (fi, _, _) = self.get_object_fileinfo(bucket, object, opts, false).await?;
+
+        let write_quorum = fi.write_quorum(self.default_write_quorum());
+
+        let oi = ObjectInfo::from_file_info(&fi, bucket, object, opts.versioned || opts.version_suspended);
+        // TODO: replicatio
+
+        if fi.deleted {
+            if opts.version_id.is_none() || opts.delete_marker {
+                return Err(to_object_err(StorageError::FileNotFound, vec![bucket, object]));
+            } else {
+                return Err(to_object_err(StorageError::MethodNotAllowed, vec![bucket, object]));
+            }
+        }
+
+        Ok((oi, write_quorum))
+    }
 
     #[allow(clippy::too_many_arguments)]
     #[tracing::instrument(
@@ -3211,6 +3231,20 @@ impl ObjectIO for SetDisks {
         h: HeaderMap,
         opts: &ObjectOptions,
     ) -> Result<GetObjectReader> {
+        // Acquire a shared read-lock early to protect read consistency
+        let mut _read_lock_guard: Option<rustfs_lock::LockGuard> = None;
+        if !opts.no_lock {
+            let guard_opt = self
+                .namespace_lock
+                .rlock_guard(object, &self.locker_owner, Duration::from_secs(5), Duration::from_secs(10))
+                .await?;
+
+            if guard_opt.is_none() {
+                return Err(Error::other("can not get lock. please retry".to_string()));
+            }
+            _read_lock_guard = guard_opt;
+        }
+
         let (fi, files, disks) = self
             .get_object_fileinfo(bucket, object, opts, true)
             .await
@@ -3256,7 +3290,10 @@ impl ObjectIO for SetDisks {
         let object = object.to_owned();
         let set_index = self.set_index;
         let pool_index = self.pool_index;
+        // Move the read-lock guard into the task so it lives for the duration of the read
+        let _guard_to_hold = _read_lock_guard; // moved into closure below
         tokio::spawn(async move {
+            let _guard = _guard_to_hold; // keep guard alive until task ends
             if let Err(e) = Self::get_object_with_fileinfo(
                 &bucket,
                 &object,
@@ -3284,16 +3321,18 @@ impl ObjectIO for SetDisks {
     async fn put_object(&self, bucket: &str, object: &str, data: &mut PutObjReader, opts: &ObjectOptions) -> Result<ObjectInfo> {
         let disks = self.disks.read().await;
 
+        // Acquire per-object exclusive lock via RAII guard. It auto-releases asynchronously on drop.
+        let mut _object_lock_guard: Option<rustfs_lock::LockGuard> = None;
         if !opts.no_lock {
-            let paths = vec![object.to_string()];
-            let lock_acquired = self
+            let guard_opt = self
                 .namespace_lock
-                .lock_batch(&paths, &self.locker_owner, Duration::from_secs(5), Duration::from_secs(10))
+                .lock_guard(object, &self.locker_owner, Duration::from_secs(5), Duration::from_secs(10))
                 .await?;
 
-            if !lock_acquired {
+            if guard_opt.is_none() {
                 return Err(Error::other("can not get lock. please retry".to_string()));
             }
+            _object_lock_guard = guard_opt;
         }
 
         let mut user_defined = opts.user_defined.clone();
@@ -3500,14 +3539,6 @@ impl ObjectIO for SetDisks {
 
         self.delete_all(RUSTFS_META_TMP_BUCKET, &tmp_dir).await?;
 
-        // Release lock if it was acquired
-        if !opts.no_lock {
-            let paths = vec![object.to_string()];
-            if let Err(err) = self.namespace_lock.unlock_batch(&paths, &self.locker_owner).await {
-                error!("Failed to unlock object {}: {}", object, err);
-            }
-        }
-
         for (i, op_disk) in online_disks.iter().enumerate() {
             if let Some(disk) = op_disk {
                 if disk.is_online().await {
@@ -3583,6 +3614,19 @@ impl StorageAPI for SetDisks {
             return Err(StorageError::NotImplemented);
         }
 
+        // Guard lock for source object metadata update
+        let mut _lock_guard: Option<rustfs_lock::LockGuard> = None;
+        {
+            let guard_opt = self
+                .namespace_lock
+                .lock_guard(src_object, &self.locker_owner, Duration::from_secs(5), Duration::from_secs(10))
+                .await?;
+            if guard_opt.is_none() {
+                return Err(Error::other("can not get lock. please retry".to_string()));
+            }
+            _lock_guard = guard_opt;
+        }
+
         let disks = self.get_disks_internal().await;
 
         let (mut metas, errs) = {
@@ -3676,6 +3720,18 @@ impl StorageAPI for SetDisks {
     }
     #[tracing::instrument(skip(self))]
     async fn delete_object_version(&self, bucket: &str, object: &str, fi: &FileInfo, force_del_marker: bool) -> Result<()> {
+        // Guard lock for single object delete-version
+        let mut _lock_guard: Option<rustfs_lock::LockGuard> = None;
+        {
+            let guard_opt = self
+                .namespace_lock
+                .lock_guard(object, &self.locker_owner, Duration::from_secs(5), Duration::from_secs(10))
+                .await?;
+            if guard_opt.is_none() {
+                return Err(Error::other("can not get lock. please retry".to_string()));
+            }
+            _lock_guard = guard_opt;
+        }
         let disks = self.get_disks(0, 0).await?;
         let write_quorum = disks.len() / 2 + 1;
 
@@ -3732,23 +3788,48 @@ impl StorageAPI for SetDisks {
             del_errs.push(None)
         }
 
+        // Per-object guards to keep until function end
+        let mut _guards: HashMap<String, rustfs_lock::LockGuard> = HashMap::new();
+        // Acquire locks for all objects first; mark errors for failures
+        for (i, dobj) in objects.iter().enumerate() {
+            if !_guards.contains_key(&dobj.object_name) {
+                match self
+                    .namespace_lock
+                    .lock_guard(&dobj.object_name, &self.locker_owner, Duration::from_secs(5), Duration::from_secs(10))
+                    .await?
+                {
+                    Some(g) => {
+                        _guards.insert(dobj.object_name.clone(), g);
+                    }
+                    None => {
+                        del_errs[i] = Some(Error::other("can not get lock. please retry"));
+                    }
+                }
+            }
+        }
+
         // let mut del_fvers = Vec::with_capacity(objects.len());
 
+        let ver_cfg = BucketVersioningSys::get(bucket).await.unwrap_or_default();
+
         let mut vers_map: HashMap<&String, FileInfoVersions> = HashMap::new();
 
         for (i, dobj) in objects.iter().enumerate() {
             let mut vr = FileInfo {
                 name: dobj.object_name.clone(),
                 version_id: dobj.version_id,
+                idx: i,
                 ..Default::default()
             };
 
-            // 删除
-            del_objects[i].object_name.clone_from(&vr.name);
-            del_objects[i].version_id = vr.version_id.map(|v| v.to_string());
+            vr.set_tier_free_version_id(&Uuid::new_v4().to_string());
 
-            if del_objects[i].version_id.is_none() {
-                let (suspended, versioned) = (opts.version_suspended, opts.versioned);
+            // 删除
+            // del_objects[i].object_name.clone_from(&vr.name);
+            // del_objects[i].version_id = vr.version_id.map(|v| v.to_string());
+
+            if dobj.version_id.is_none() {
+                let (suspended, versioned) = (ver_cfg.suspended(), ver_cfg.prefix_enabled(dobj.object_name.as_str()));
                 if suspended || versioned {
                     vr.mod_time = Some(OffsetDateTime::now_utc());
                     vr.deleted = true;
@@ -3788,13 +3869,23 @@ impl StorageAPI for SetDisks {
                 }
             }
 
-            vers_map.insert(&dobj.object_name, v);
+            // Only add to vers_map if we hold the lock
+            if _guards.contains_key(&dobj.object_name) {
+                vers_map.insert(&dobj.object_name, v);
+            }
         }
 
         let mut vers = Vec::with_capacity(vers_map.len());
 
-        for (_, ver) in vers_map {
-            vers.push(ver);
+        for (_, mut fi_vers) in vers_map {
+            fi_vers.versions.sort_by(|a, b| a.deleted.cmp(&b.deleted));
+            fi_vers.versions.reverse();
+
+            if let Some(index) = fi_vers.versions.iter().position(|fi| fi.deleted) {
+                fi_vers.versions.truncate(index + 1);
+            }
+
+            vers.push(fi_vers);
         }
 
         let disks = self.disks.read().await;
@@ -3830,6 +3921,18 @@ impl StorageAPI for SetDisks {
 
     #[tracing::instrument(skip(self))]
     async fn delete_object(&self, bucket: &str, object: &str, opts: ObjectOptions) -> Result<ObjectInfo> {
+        // Guard lock for single object delete
+        let mut _lock_guard: Option<rustfs_lock::LockGuard> = None;
+        if !opts.delete_prefix {
+            let guard_opt = self
+                .namespace_lock
+                .lock_guard(object, &self.locker_owner, Duration::from_secs(5), Duration::from_secs(10))
+                .await?;
+            if guard_opt.is_none() {
+                return Err(Error::other("can not get lock. please retry".to_string()));
+            }
+            _lock_guard = guard_opt;
+        }
         if opts.delete_prefix {
             self.delete_prefix(bucket, object)
                 .await
@@ -3837,7 +3940,145 @@ impl StorageAPI for SetDisks {
 
             return Ok(ObjectInfo::default());
         }
-        unimplemented!()
+
+        let (oi, write_quorum) = match self.get_object_info_and_quorum(bucket, object, &opts).await {
+            Ok((oi, wq)) => (oi, wq),
+            Err(e) => {
+                return Err(to_object_err(e, vec![bucket, object]));
+            }
+        };
+
+        let mark_delete = oi.version_id.is_some();
+
+        let mut delete_marker = opts.versioned;
+
+        let mod_time = if let Some(mt) = opts.mod_time {
+            mt
+        } else {
+            OffsetDateTime::now_utc()
+        };
+
+        let find_vid = Uuid::new_v4();
+
+        if mark_delete && (opts.versioned || opts.version_suspended) {
+            if !delete_marker {
+                delete_marker = opts.version_suspended && opts.version_id.is_none();
+            }
+
+            let mut fi = FileInfo {
+                name: object.to_string(),
+                deleted: delete_marker,
+                mark_deleted: mark_delete,
+                mod_time: Some(mod_time),
+                ..Default::default() // TODO: replication
+            };
+
+            fi.set_tier_free_version_id(&find_vid.to_string());
+
+            if opts.skip_free_version {
+                fi.set_skip_tier_free_version();
+            }
+
+            fi.version_id = if let Some(vid) = opts.version_id {
+                Some(Uuid::parse_str(vid.as_str())?)
+            } else if opts.versioned {
+                Some(Uuid::new_v4())
+            } else {
+                None
+            };
+
+            self.delete_object_version(bucket, object, &fi, opts.delete_marker)
+                .await
+                .map_err(|e| to_object_err(e, vec![bucket, object]))?;
+
+            return Ok(ObjectInfo::from_file_info(&fi, bucket, object, opts.versioned || opts.version_suspended));
+        }
+
+        let version_id = opts.version_id.as_ref().and_then(|v| Uuid::parse_str(v).ok());
+
+        // Create a single object deletion request
+        let mut vr = FileInfo {
+            name: object.to_string(),
+            version_id: opts.version_id.as_ref().and_then(|v| Uuid::parse_str(v).ok()),
+            ..Default::default()
+        };
+
+        // Handle versioning
+        let (suspended, versioned) = (opts.version_suspended, opts.versioned);
+        if opts.version_id.is_none() && (suspended || versioned) {
+            vr.mod_time = Some(OffsetDateTime::now_utc());
+            vr.deleted = true;
+            if versioned {
+                vr.version_id = Some(Uuid::new_v4());
+            }
+        }
+
+        let vers = vec![FileInfoVersions {
+            name: vr.name.clone(),
+            versions: vec![vr.clone()],
+            ..Default::default()
+        }];
+
+        let disks = self.disks.read().await;
+        let disks = disks.clone();
+        let write_quorum = disks.len() / 2 + 1;
+
+        let mut futures = Vec::with_capacity(disks.len());
+        let mut errs = Vec::with_capacity(disks.len());
+
+        for disk in disks.iter() {
+            let vers = vers.clone();
+            futures.push(async move {
+                if let Some(disk) = disk {
+                    disk.delete_versions(bucket, vers, DeleteOptions::default()).await
+                } else {
+                    Err(DiskError::DiskNotFound)
+                }
+            });
+        }
+
+        let results = join_all(futures).await;
+
+        for result in results {
+            match result {
+                Ok(disk_errs) => {
+                    // Handle errors from disk operations
+                    for err in disk_errs.iter().flatten() {
+                        warn!("delete_object disk error: {:?}", err);
+                    }
+                    errs.push(None);
+                }
+                Err(e) => {
+                    errs.push(Some(e));
+                }
+            }
+        }
+
+        // Check write quorum
+        if let Some(err) = reduce_write_quorum_errs(&errs, OBJECT_OP_IGNORED_ERRS, write_quorum) {
+            return Err(to_object_err(err.into(), vec![bucket, object]));
+        }
+
+        // Create result ObjectInfo
+        let result_info = if vr.deleted {
+            ObjectInfo {
+                bucket: bucket.to_string(),
+                name: object.to_string(),
+                delete_marker: true,
+                mod_time: vr.mod_time,
+                version_id: vr.version_id,
+                ..Default::default()
+            }
+        } else {
+            ObjectInfo {
+                bucket: bucket.to_string(),
+                name: object.to_string(),
+                version_id: vr.version_id,
+                ..Default::default()
+            }
+        };
+
+        Ok(result_info)
     }
 
     #[tracing::instrument(skip(self))]
@@ -3869,33 +4110,18 @@ impl StorageAPI for SetDisks {
 
     #[tracing::instrument(skip(self))]
     async fn get_object_info(&self, bucket: &str, object: &str, opts: &ObjectOptions) -> Result<ObjectInfo> {
-        // let mut _ns = None;
-        // if !opts.no_lock {
-        //     let paths = vec![object.to_string()];
-        //     let ns_lock = new_nslock(
-        //         Arc::clone(&self.ns_mutex),
-        //         self.locker_owner.clone(),
-        //         bucket.to_string(),
-        //         paths,
-        //         self.lockers.clone(),
-        //     )
-        //     .await;
-        //     if !ns_lock
-        //         .0
-        //         .write()
-        //         .await
-        //         .get_lock(&Options {
-        //             timeout: Duration::from_secs(5),
-        //             retry_interval: Duration::from_secs(1),
-        //         })
-        //         .await
-        //         .map_err(|err| Error::other(err.to_string()))?
-        //     {
-        //         return Err(Error::other("can not get lock. please retry".to_string()));
-        //     }
-
-        //     _ns = Some(ns_lock);
-        // }
+        // Acquire a shared read-lock to protect consistency during info fetch
+        let mut _read_lock_guard: Option<rustfs_lock::LockGuard> = None;
+        if !opts.no_lock {
+            let guard_opt = self
+                .namespace_lock
+                .rlock_guard(object, &self.locker_owner, Duration::from_secs(5), Duration::from_secs(10))
+                .await?;
+            if guard_opt.is_none() {
+                return Err(Error::other("can not get lock. please retry".to_string()));
+            }
+            _read_lock_guard = guard_opt;
+        }
 
         let (fi, _, _) = self
             .get_object_fileinfo(bucket, object, opts, false)
@@ -3927,6 +4153,19 @@ impl StorageAPI for SetDisks {
     async fn put_object_metadata(&self, bucket: &str, object: &str, opts: &ObjectOptions) -> Result<ObjectInfo> {
         // TODO: nslock
 
+        // Guard lock for metadata update
+        let mut _lock_guard: Option<rustfs_lock::LockGuard> = None;
+        if !opts.no_lock {
+            let guard_opt = self
+                .namespace_lock
+                .lock_guard(object, &self.locker_owner, Duration::from_secs(5), Duration::from_secs(10))
+                .await?;
+            if guard_opt.is_none() {
+                return Err(Error::other("can not get lock. please retry".to_string()));
+            }
+            _lock_guard = guard_opt;
+        }
+
         let disks = self.get_disks_internal().await;
 
         let (metas, errs) = {
@@ -4017,12 +4256,18 @@ impl StorageAPI for SetDisks {
             }
         };
 
-        /*if !opts.no_lock {
-            let lk = self.new_ns_lock(bucket, object);
-            let lkctx = lk.get_lock(globalDeleteOperationTimeout)?;
-            //ctx = lkctx.Context()
-            //defer lk.Unlock(lkctx)
-        }*/
+        // Acquire write-lock early; hold for the whole transition operation scope
+        let mut _lock_guard: Option<rustfs_lock::LockGuard> = None;
+        if !opts.no_lock {
+            let guard_opt = self
+                .namespace_lock
+                .lock_guard(object, &self.locker_owner, Duration::from_secs(5), Duration::from_secs(10))
+                .await?;
+            if guard_opt.is_none() {
+                return Err(Error::other("can not get lock. please retry".to_string()));
+            }
+            _lock_guard = guard_opt;
+        }
 
         let (mut fi, meta_arr, online_disks) = self.get_object_fileinfo(bucket, object, opts, true).await?;
         /*if err != nil {
@@ -4140,6 +4385,18 @@ impl StorageAPI for SetDisks {
 
     #[tracing::instrument(level = "debug", skip(self))]
     async fn restore_transitioned_object(&self, bucket: &str, object: &str, opts: &ObjectOptions) -> Result<()> {
+        // Acquire write-lock early for the restore operation
+        let mut _lock_guard: Option<rustfs_lock::LockGuard> = None;
+        if !opts.no_lock {
+            let guard_opt = self
+                .namespace_lock
+                .lock_guard(object, &self.locker_owner, Duration::from_secs(5), Duration::from_secs(10))
+                .await?;
+            if guard_opt.is_none() {
+                return Err(Error::other("can not get lock. please retry".to_string()));
+            }
+            _lock_guard = guard_opt;
+        }
         let set_restore_header_fn = async move |oi: &mut ObjectInfo, rerr: Option<Error>| -> Result<()> {
             if rerr.is_none() {
                 return Ok(());
@@ -4213,6 +4470,18 @@ impl StorageAPI for SetDisks {
 
     #[tracing::instrument(level = "debug", skip(self))]
     async fn put_object_tags(&self, bucket: &str, object: &str, tags: &str, opts: &ObjectOptions) -> Result<ObjectInfo> {
+        // Acquire write-lock for tag update (metadata write)
+        let mut _lock_guard: Option<rustfs_lock::LockGuard> = None;
+        if !opts.no_lock {
+            let guard_opt = self
+                .namespace_lock
+                .lock_guard(object, &self.locker_owner, Duration::from_secs(5), Duration::from_secs(10))
+                .await?;
+            if guard_opt.is_none() {
+                return Err(Error::other("can not get lock. please retry".to_string()));
+            }
+            _lock_guard = guard_opt;
+        }
         let (mut fi, _, disks) = self.get_object_fileinfo(bucket, object, opts, false).await?;
 
         fi.metadata.insert(AMZ_OBJECT_TAGGING.to_owned(), tags.to_owned());
@@ -5175,9 +5444,10 @@ impl StorageAPI for SetDisks {
 
     #[tracing::instrument(skip(self))]
     async fn verify_object_integrity(&self, bucket: &str, object: &str, opts: &ObjectOptions) -> Result<()> {
-        let mut get_object_reader =
-            <Self as ObjectIO>::get_object_reader(self, bucket, object, None, HeaderMap::new(), opts).await?;
-        let _ = get_object_reader.read_all().await?;
+        let get_object_reader = <Self as ObjectIO>::get_object_reader(self, bucket, object, None, HeaderMap::new(), opts).await?;
+        // Stream to sink to avoid loading entire object into memory during verification
+        let mut reader = get_object_reader.stream;
+        tokio::io::copy(&mut reader, &mut tokio::io::sink()).await?;
         Ok(())
     }
 }

crates/ecstore/src/sets.rs

@@ -165,7 +165,13 @@ impl Sets {
 
             let lock_clients = create_unique_clients(&set_endpoints).await?;
 
-            let namespace_lock = rustfs_lock::NamespaceLock::with_clients(format!("set-{i}"), lock_clients);
+            // Bind lock quorum to EC write quorum for this set: data_shards (+1 if equal to parity) per default_write_quorum()
+            let mut write_quorum = set_drive_count - parity_count;
+            if write_quorum == parity_count {
+                write_quorum += 1;
+            }
+            let namespace_lock =
+                rustfs_lock::NamespaceLock::with_clients_and_quorum(format!("set-{i}"), lock_clients, write_quorum);
 
             let set_disks = SetDisks::new(
                 Arc::new(namespace_lock),
@@ -876,11 +882,15 @@ impl StorageAPI for Sets {
         unimplemented!()
     }
 
-    #[tracing::instrument(skip(self))]
+    #[tracing::instrument(level = "debug", skip(self))]
     async fn verify_object_integrity(&self, bucket: &str, object: &str, opts: &ObjectOptions) -> Result<()> {
-        self.get_disks_by_key(object)
-            .verify_object_integrity(bucket, object, opts)
-            .await
+        let gor = self.get_object_reader(bucket, object, None, HeaderMap::new(), opts).await?;
+        let mut reader = gor.stream;
+
+        // Stream data to sink instead of reading all into memory to prevent OOM
+        tokio::io::copy(&mut reader, &mut tokio::io::sink()).await?;
+
+        Ok(())
     }
 }
 

crates/ecstore/src/store.rs

@@ -2238,9 +2238,10 @@ impl StorageAPI for ECStore {
     }
 
     async fn verify_object_integrity(&self, bucket: &str, object: &str, opts: &ObjectOptions) -> Result<()> {
-        let mut get_object_reader =
-            <Self as ObjectIO>::get_object_reader(self, bucket, object, None, HeaderMap::new(), opts).await?;
-        let _ = get_object_reader.read_all().await?;
+        let get_object_reader = <Self as ObjectIO>::get_object_reader(self, bucket, object, None, HeaderMap::new(), opts).await?;
+        // Stream to sink to avoid loading entire object into memory during verification
+        let mut reader = get_object_reader.stream;
+        tokio::io::copy(&mut reader, &mut tokio::io::sink()).await?;
         Ok(())
     }
 }

crates/ecstore/src/store_api.rs

@@ -310,6 +310,8 @@ pub struct ObjectOptions {
     pub replication_request: bool,
     pub delete_marker: bool,
 
+    pub skip_free_version: bool,
+
     pub transition: TransitionOptions,
     pub expiration: ExpirationOptions,
     pub lifecycle_audit_event: LcAuditEvent,

crates/ecstore/src/store_init.rs

@@ -139,8 +139,8 @@ async fn init_format_erasure(
             let idx = i * set_drive_count + j;
             let mut newfm = fm.clone();
             newfm.erasure.this = fm.erasure.sets[i][j];
-            if deployment_id.is_some() {
-                newfm.id = deployment_id.unwrap();
+            if let Some(id) = deployment_id {
+                newfm.id = id;
             }
 
             fms[idx] = Some(newfm);

crates/filemeta/src/filemeta.rs

@@ -496,39 +496,32 @@ impl FileMeta {
     }
 
     pub fn add_version_filemata(&mut self, ver: FileMetaVersion) -> Result<()> {
-        let mod_time = ver.get_mod_time().unwrap().nanosecond();
         if !ver.valid() {
             return Err(Error::other("attempted to add invalid version"));
         }
-        let encoded = ver.marshal_msg()?;
 
-        if self.versions.len() + 1 > 100 {
+        if self.versions.len() + 1 >= 100 {
             return Err(Error::other(
                 "You've exceeded the limit on the number of versions you can create on this object",
             ));
         }
 
-        self.versions.push(FileMetaShallowVersion {
-            header: FileMetaVersionHeader {
-                mod_time: Some(OffsetDateTime::from_unix_timestamp(-1)?),
-                ..Default::default()
-            },
-            ..Default::default()
-        });
+        let mod_time = ver.get_mod_time();
+        let encoded = ver.marshal_msg()?;
+        let new_version = FileMetaShallowVersion {
+            header: ver.header(),
+            meta: encoded,
+        };
 
-        let len = self.versions.len();
-        for (i, existing) in self.versions.iter().enumerate() {
-            if existing.header.mod_time.unwrap().nanosecond() <= mod_time {
-                let vers = self.versions[i..len - 1].to_vec();
-                self.versions[i + 1..].clone_from_slice(vers.as_slice());
-                self.versions[i] = FileMetaShallowVersion {
-                    header: ver.header(),
-                    meta: encoded,
-                };
-                return Ok(());
-            }
-        }
-        Err(Error::other("addVersion: Internal error, unable to add version"))
+        // Find the insertion position: insert before the first element with mod_time >= new mod_time
+        // This maintains descending order by mod_time (newest first)
+        let insert_pos = self
+            .versions
+            .iter()
+            .position(|existing| existing.header.mod_time <= mod_time)
+            .unwrap_or(self.versions.len());
+        self.versions.insert(insert_pos, new_version);
+        Ok(())
     }
 
     // delete_version deletes version, returns data_dir
@@ -554,7 +547,15 @@ impl FileMeta {
 
             match ver.header.version_type {
                 VersionType::Invalid | VersionType::Legacy => return Err(Error::other("invalid file meta version")),
-                VersionType::Delete => return Ok(None),
+                VersionType::Delete => {
+                    self.versions.remove(i);
+                    if fi.deleted && fi.version_id.is_none() {
+                        self.add_version_filemata(ventry)?;
+                        return Ok(None);
+                    }
+
+                    return Ok(None);
+                }
                 VersionType::Object => {
                     let v = self.get_idx(i)?;
 
@@ -600,6 +601,7 @@ impl FileMeta {
 
         if fi.deleted {
             self.add_version_filemata(ventry)?;
+            return Ok(None);
         }
 
         Err(Error::FileVersionNotFound)
@@ -961,7 +963,8 @@ impl FileMetaVersion {
 
     pub fn get_version_id(&self) -> Option<Uuid> {
         match self.version_type {
-            VersionType::Object | VersionType::Delete => self.object.as_ref().map(|v| v.version_id).unwrap_or_default(),
+            VersionType::Object => self.object.as_ref().map(|v| v.version_id).unwrap_or_default(),
+            VersionType::Delete => self.delete_marker.as_ref().map(|v| v.version_id).unwrap_or_default(),
             _ => None,
         }
     }
@@ -2363,7 +2366,7 @@ mod test {
         assert!(stats.delete_markers > 0, "应该有删除标记");
 
         // 测试版本合并功能
-        let merged = merge_file_meta_versions(1, false, 0, &[fm.versions.clone()]);
+        let merged = merge_file_meta_versions(1, false, 0, std::slice::from_ref(&fm.versions));
         assert!(!merged.is_empty(), "合并后应该有版本");
     }
 

crates/filemeta/src/lib.rs

@@ -17,7 +17,7 @@ pub mod fileinfo;
 mod filemeta;
 mod filemeta_inline;
 pub mod headers;
-mod metacache;
+pub mod metacache;
 
 pub mod test_data;
 

crates/filemeta/src/metacache.rs

@@ -795,24 +795,26 @@ impl<T: Clone + Debug + Send + 'static> Cache<T> {
             }
         }
 
-        if self.opts.no_wait && v.is_some() && now - self.last_update_ms.load(AtomicOrdering::SeqCst) < self.ttl.as_secs() * 2 {
-            if self.updating.try_lock().is_ok() {
-                let this = Arc::clone(&self);
-                spawn(async move {
-                    let _ = this.update().await;
-                });
+        if self.opts.no_wait && now - self.last_update_ms.load(AtomicOrdering::SeqCst) < self.ttl.as_secs() * 2 {
+            if let Some(value) = v {
+                if self.updating.try_lock().is_ok() {
+                    let this = Arc::clone(&self);
+                    spawn(async move {
+                        let _ = this.update().await;
+                    });
+                }
+                return Ok(value);
             }
-
-            return Ok(v.unwrap());
         }
 
         let _ = self.updating.lock().await;
 
-        if let Ok(duration) =
-            SystemTime::now().duration_since(UNIX_EPOCH + Duration::from_secs(self.last_update_ms.load(AtomicOrdering::SeqCst)))
-        {
+        if let (Ok(duration), Some(value)) = (
+            SystemTime::now().duration_since(UNIX_EPOCH + Duration::from_secs(self.last_update_ms.load(AtomicOrdering::SeqCst))),
+            v,
+        ) {
             if duration < self.ttl {
-                return Ok(v.unwrap());
+                return Ok(value);
             }
         }
 

crates/lock/Cargo.toml

@@ -32,9 +32,7 @@ workspace = true
 async-trait.workspace = true
 bytes.workspace = true
 futures.workspace = true
-lazy_static.workspace = true
 rustfs-protos.workspace = true
-rand.workspace = true
 serde.workspace = true
 serde_json.workspace = true
 tokio.workspace = true
@@ -44,4 +42,3 @@ url.workspace = true
 uuid.workspace = true
 thiserror.workspace = true
 once_cell.workspace = true
-lru.workspace = true

crates/lock/src/guard.rs

@@ -0,0 +1,120 @@
+// 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::Arc;
+
+use once_cell::sync::Lazy;
+use tokio::sync::mpsc;
+
+use crate::{client::LockClient, types::LockId};
+
+#[derive(Debug, Clone)]
+struct UnlockJob {
+    lock_id: LockId,
+    clients: Vec<Arc<dyn LockClient>>, // cloned Arcs; cheap and shares state
+}
+
+#[derive(Debug)]
+struct UnlockRuntime {
+    tx: mpsc::Sender<UnlockJob>,
+}
+
+// Global unlock runtime with background worker
+static UNLOCK_RUNTIME: Lazy<UnlockRuntime> = Lazy::new(|| {
+    // Larger buffer to reduce contention during bursts
+    let (tx, mut rx) = mpsc::channel::<UnlockJob>(8192);
+
+    // Spawn background worker when first used; assumes a Tokio runtime is available
+    tokio::spawn(async move {
+        while let Some(job) = rx.recv().await {
+            // Best-effort release across clients; try all, success if any succeeds
+            let mut any_ok = false;
+            let lock_id = job.lock_id.clone();
+            for client in job.clients.into_iter() {
+                if client.release(&lock_id).await.unwrap_or(false) {
+                    any_ok = true;
+                }
+            }
+            if !any_ok {
+                tracing::warn!("LockGuard background release failed for {}", lock_id);
+            } else {
+                tracing::debug!("LockGuard background released {}", lock_id);
+            }
+        }
+    });
+
+    UnlockRuntime { tx }
+});
+
+/// A RAII guard that releases the lock asynchronously when dropped.
+#[derive(Debug)]
+pub struct LockGuard {
+    lock_id: LockId,
+    clients: Vec<Arc<dyn LockClient>>,
+    /// If true, Drop will not try to release (used if user manually released).
+    disarmed: bool,
+}
+
+impl LockGuard {
+    pub(crate) fn new(lock_id: LockId, clients: Vec<Arc<dyn LockClient>>) -> Self {
+        Self {
+            lock_id,
+            clients,
+            disarmed: false,
+        }
+    }
+
+    /// Get the lock id associated with this guard
+    pub fn lock_id(&self) -> &LockId {
+        &self.lock_id
+    }
+
+    /// Manually disarm the guard so dropping it won't release the lock.
+    /// Call this if you explicitly released the lock elsewhere.
+    pub fn disarm(&mut self) {
+        self.disarmed = true;
+    }
+}
+
+impl Drop for LockGuard {
+    fn drop(&mut self) {
+        if self.disarmed {
+            return;
+        }
+
+        let job = UnlockJob {
+            lock_id: self.lock_id.clone(),
+            clients: self.clients.clone(),
+        };
+
+        // Try a non-blocking send to avoid panics in Drop
+        if let Err(err) = UNLOCK_RUNTIME.tx.try_send(job) {
+            // Channel full or closed; best-effort fallback: spawn a detached task
+            let lock_id = self.lock_id.clone();
+            let clients = self.clients.clone();
+            tracing::warn!("LockGuard channel send failed ({}), spawning fallback unlock task for {}", err, lock_id);
+
+            // If runtime is not available, this will panic; but in RustFS we are inside Tokio contexts.
+            let handle = tokio::spawn(async move {
+                let futures_iter = clients.into_iter().map(|client| {
+                    let id = lock_id.clone();
+                    async move { client.release(&id).await.unwrap_or(false) }
+                });
+                let _ = futures::future::join_all(futures_iter).await;
+            });
+            // Explicitly drop the JoinHandle to acknowledge detaching the task.
+            std::mem::drop(handle);
+        }
+    }
+}

crates/lock/src/lib.rs

@@ -27,6 +27,7 @@ pub mod local;
 
 // Core Modules
 pub mod error;
+pub mod guard;
 pub mod types;
 
 // ============================================================================
@@ -39,6 +40,7 @@ pub use crate::{
     client::{LockClient, local::LocalClient, remote::RemoteClient},
     // Error types
     error::{LockError, Result},
+    guard::LockGuard,
     local::LocalLockMap,
     // Main components
     namespace::{NamespaceLock, NamespaceLockManager},

crates/lock/src/local.rs

@@ -12,11 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use std::collections::HashMap;
+use std::collections::{BTreeMap, HashMap};
 use std::sync::Arc;
 use std::sync::atomic::{AtomicBool, Ordering};
 use std::time::{Duration, Instant};
-use tokio::sync::RwLock;
+use tokio::sync::{Mutex, Notify, RwLock};
 
 use crate::LockRequest;
 
@@ -29,6 +29,11 @@ pub struct LocalLockEntry {
     pub readers: HashMap<String, usize>,
     /// lock expiration time
     pub expires_at: Option<Instant>,
+    /// number of writers waiting (for simple fairness against reader storms)
+    pub writer_pending: usize,
+    /// notifiers for readers/writers
+    pub notify_readers: Arc<Notify>,
+    pub notify_writers: Arc<Notify>,
 }
 
 /// local lock map
@@ -38,6 +43,10 @@ pub struct LocalLockMap {
     pub locks: Arc<RwLock<HashMap<crate::types::LockId, Arc<RwLock<LocalLockEntry>>>>>,
     /// Shutdown flag for background tasks
     shutdown: Arc<AtomicBool>,
+    /// expiration schedule map: when -> lock_ids
+    expirations: Arc<Mutex<BTreeMap<Instant, Vec<crate::types::LockId>>>>,
+    /// notify expiry task when new earlier deadline arrives
+    exp_notify: Arc<Notify>,
 }
 
 impl Default for LocalLockMap {
@@ -52,6 +61,8 @@ impl LocalLockMap {
         let map = Self {
             locks: Arc::new(RwLock::new(HashMap::new())),
             shutdown: Arc::new(AtomicBool::new(false)),
+            expirations: Arc::new(Mutex::new(BTreeMap::new())),
+            exp_notify: Arc::new(Notify::new()),
         };
         map.spawn_expiry_task();
         map
@@ -61,56 +72,115 @@ impl LocalLockMap {
     fn spawn_expiry_task(&self) {
         let locks = self.locks.clone();
         let shutdown = self.shutdown.clone();
+        let expirations = self.expirations.clone();
+        let exp_notify = self.exp_notify.clone();
         tokio::spawn(async move {
-            let mut interval = tokio::time::interval(Duration::from_secs(1));
             loop {
-                interval.tick().await;
-
                 if shutdown.load(Ordering::Relaxed) {
                     tracing::debug!("Expiry task shutting down");
                     break;
                 }
 
-                let now = Instant::now();
-                let mut to_remove = Vec::new();
+                // Find next deadline and drain due ids
+                let (due_ids, wait_duration) = {
+                    let mut due = Vec::new();
+                    let mut guard = expirations.lock().await;
+                    let now = Instant::now();
+                    let next_deadline = guard.first_key_value().map(|(k, _)| *k);
+                    // drain all <= now
+                    let mut keys_to_remove = Vec::new();
+                    for (k, v) in guard.range(..=now).map(|(k, v)| (*k, v.clone())) {
+                        due.extend(v);
+                        keys_to_remove.push(k);
+                    }
+                    for k in keys_to_remove {
+                        guard.remove(&k);
+                    }
+                    let wait = if due.is_empty() {
+                        next_deadline.map(|dl| if dl > now { dl - now } else { Duration::from_millis(0) })
+                    } else {
+                        Some(Duration::from_millis(0))
+                    };
+                    (due, wait)
+                };
 
-                {
-                    let locks_guard = locks.read().await;
-                    for (key, entry) in locks_guard.iter() {
-                        if let Ok(mut entry_guard) = entry.try_write() {
-                            if let Some(exp) = entry_guard.expires_at {
-                                if exp <= now {
-                                    entry_guard.writer = None;
-                                    entry_guard.readers.clear();
-                                    entry_guard.expires_at = None;
+                if !due_ids.is_empty() {
+                    // process due ids without holding the map lock during awaits
+                    let now = Instant::now();
+                    // collect entries to process
+                    let entries: Vec<(crate::types::LockId, Arc<RwLock<LocalLockEntry>>)> = {
+                        let locks_guard = locks.read().await;
+                        due_ids
+                            .into_iter()
+                            .filter_map(|id| locks_guard.get(&id).cloned().map(|e| (id, e)))
+                            .collect()
+                    };
 
-                                    if entry_guard.writer.is_none() && entry_guard.readers.is_empty() {
-                                        to_remove.push(key.clone());
-                                    }
+                    let mut to_remove = Vec::new();
+                    for (lock_id, entry) in entries {
+                        let mut entry_guard = entry.write().await;
+                        if let Some(exp) = entry_guard.expires_at {
+                            if exp <= now {
+                                entry_guard.writer = None;
+                                entry_guard.readers.clear();
+                                entry_guard.expires_at = None;
+                                entry_guard.notify_writers.notify_waiters();
+                                entry_guard.notify_readers.notify_waiters();
+                                if entry_guard.writer.is_none() && entry_guard.readers.is_empty() {
+                                    to_remove.push(lock_id);
                                 }
                             }
                         }
                     }
+                    if !to_remove.is_empty() {
+                        let mut locks_w = locks.write().await;
+                        for id in to_remove {
+                            let _ = locks_w.remove(&id);
+                        }
+                    }
+                    continue; // immediately look for next
                 }
 
-                if !to_remove.is_empty() {
-                    let mut locks_guard = locks.write().await;
-                    for key in to_remove {
-                        locks_guard.remove(&key);
+                // nothing due; wait for next deadline or notification
+                if let Some(dur) = wait_duration {
+                    tokio::select! {
+                        _ = tokio::time::sleep(dur) => {},
+                        _ = exp_notify.notified() => {},
                     }
+                } else {
+                    // no deadlines, wait for new schedule or shutdown tick
+                    exp_notify.notified().await;
                 }
             }
         });
     }
 
+    /// schedule an expiry time for the given lock id (inline, avoid per-acquisition spawn)
+    async fn schedule_expiry(&self, id: crate::types::LockId, exp: Instant) {
+        let mut guard = self.expirations.lock().await;
+        let is_earliest = match guard.first_key_value() {
+            Some((k, _)) => exp < *k,
+            None => true,
+        };
+        guard.entry(exp).or_insert_with(Vec::new).push(id);
+        drop(guard);
+        if is_earliest {
+            self.exp_notify.notify_waiters();
+        }
+    }
+
     /// write lock with TTL, support timeout, use LockRequest
     pub async fn lock_with_ttl_id(&self, request: &LockRequest) -> std::io::Result<bool> {
         let start = Instant::now();
-        let expires_at = Some(Instant::now() + request.ttl);
 
         loop {
-            // get or create lock entry
-            let entry = {
+            // get or create lock entry (double-checked to reduce write-lock contention)
+            let entry = if let Some(e) = {
+                let locks_guard = self.locks.read().await;
+                locks_guard.get(&request.lock_id).cloned()
+            } {
+                e
+            } else {
                 let mut locks_guard = self.locks.write().await;
                 locks_guard
                     .entry(request.lock_id.clone())
@@ -119,13 +189,17 @@ impl LocalLockMap {
                             writer: None,
                             readers: HashMap::new(),
                             expires_at: None,
+                            writer_pending: 0,
+                            notify_readers: Arc::new(Notify::new()),
+                            notify_writers: Arc::new(Notify::new()),
                         }))
                     })
                     .clone()
             };
 
-            // try to get write lock to modify state
-            if let Ok(mut entry_guard) = entry.try_write() {
+            // attempt acquisition or wait using Notify
+            let notify_to_wait = {
+                let mut entry_guard = entry.write().await;
                 // check expired state
                 let now = Instant::now();
                 if let Some(exp) = entry_guard.expires_at {
@@ -136,30 +210,68 @@ impl LocalLockMap {
                     }
                 }
 
-                // check if can get write lock
+                // try acquire
                 if entry_guard.writer.is_none() && entry_guard.readers.is_empty() {
                     entry_guard.writer = Some(request.owner.clone());
-                    entry_guard.expires_at = expires_at;
+                    let expires_at = Instant::now() + request.ttl;
+                    entry_guard.expires_at = Some(expires_at);
                     tracing::debug!("Write lock acquired for resource '{}' by owner '{}'", request.resource, request.owner);
+                    {
+                        drop(entry_guard);
+                        self.schedule_expiry(request.lock_id.clone(), expires_at).await;
+                    }
                     return Ok(true);
                 }
-            }
+                // couldn't acquire now, mark as pending writer and choose notifier
+                entry_guard.writer_pending = entry_guard.writer_pending.saturating_add(1);
+                entry_guard.notify_writers.clone()
+            };
 
-            if start.elapsed() >= request.acquire_timeout {
+            // wait with remaining timeout
+            let elapsed = start.elapsed();
+            if elapsed >= request.acquire_timeout {
+                // best-effort decrement pending counter
+                if let Ok(mut eg) = entry.try_write() {
+                    eg.writer_pending = eg.writer_pending.saturating_sub(1);
+                } else {
+                    let mut eg = entry.write().await;
+                    eg.writer_pending = eg.writer_pending.saturating_sub(1);
+                }
                 return Ok(false);
             }
-            tokio::time::sleep(Duration::from_millis(10)).await;
+            let remaining = request.acquire_timeout - elapsed;
+            if tokio::time::timeout(remaining, notify_to_wait.notified()).await.is_err() {
+                // timeout; decrement pending before returning
+                if let Ok(mut eg) = entry.try_write() {
+                    eg.writer_pending = eg.writer_pending.saturating_sub(1);
+                } else {
+                    let mut eg = entry.write().await;
+                    eg.writer_pending = eg.writer_pending.saturating_sub(1);
+                }
+                return Ok(false);
+            }
+            // woke up; decrement pending before retrying
+            if let Ok(mut eg) = entry.try_write() {
+                eg.writer_pending = eg.writer_pending.saturating_sub(1);
+            } else {
+                let mut eg = entry.write().await;
+                eg.writer_pending = eg.writer_pending.saturating_sub(1);
+            }
         }
     }
 
     /// read lock with TTL, support timeout, use LockRequest
     pub async fn rlock_with_ttl_id(&self, request: &LockRequest) -> std::io::Result<bool> {
         let start = Instant::now();
-        let expires_at = Some(Instant::now() + request.ttl);
 
         loop {
-            // get or create lock entry
-            let entry = {
+            // get or create lock entry (double-checked to reduce write-lock contention)
+            let entry = if let Some(e) = {
+                let locks_guard = self.locks.read().await;
+                locks_guard.get(&request.lock_id).cloned()
+            } {
+                e
+            } else {
                 let mut locks_guard = self.locks.write().await;
                 locks_guard
                     .entry(request.lock_id.clone())
@@ -168,13 +280,17 @@ impl LocalLockMap {
                             writer: None,
                             readers: HashMap::new(),
                             expires_at: None,
+                            writer_pending: 0,
+                            notify_readers: Arc::new(Notify::new()),
+                            notify_writers: Arc::new(Notify::new()),
                         }))
                     })
                     .clone()
             };
 
-            // try to get write lock to modify state
-            if let Ok(mut entry_guard) = entry.try_write() {
+            // attempt acquisition or wait using Notify
+            let notify_to_wait = {
+                let mut entry_guard = entry.write().await;
                 // check expired state
                 let now = Instant::now();
                 if let Some(exp) = entry_guard.expires_at {
@@ -185,189 +301,247 @@ impl LocalLockMap {
                     }
                 }
 
-                // check if can get read lock
-                if entry_guard.writer.is_none() {
-                    // increase read lock count
+                if entry_guard.writer.is_none() && entry_guard.writer_pending == 0 {
                     *entry_guard.readers.entry(request.owner.clone()).or_insert(0) += 1;
-                    entry_guard.expires_at = expires_at;
+                    let expires_at = Instant::now() + request.ttl;
+                    entry_guard.expires_at = Some(expires_at);
                     tracing::debug!("Read lock acquired for resource '{}' by owner '{}'", request.resource, request.owner);
+                    {
+                        drop(entry_guard);
+                        self.schedule_expiry(request.lock_id.clone(), expires_at).await;
+                    }
                     return Ok(true);
                 }
-            }
 
-            if start.elapsed() >= request.acquire_timeout {
+                // choose notifier: prefer waiting on writers if writers pending, else readers
+                if entry_guard.writer_pending > 0 {
+                    entry_guard.notify_writers.clone()
+                } else {
+                    entry_guard.notify_readers.clone()
+                }
+            };
+
+            // wait with remaining timeout
+            let elapsed = start.elapsed();
+            if elapsed >= request.acquire_timeout {
+                return Ok(false);
+            }
+            let remaining = request.acquire_timeout - elapsed;
+            if tokio::time::timeout(remaining, notify_to_wait.notified()).await.is_err() {
                 return Ok(false);
             }
-            tokio::time::sleep(Duration::from_millis(10)).await;
         }
     }
 
     /// unlock by LockId and owner - need to specify owner to correctly unlock
     pub async fn unlock_by_id_and_owner(&self, lock_id: &crate::types::LockId, owner: &str) -> std::io::Result<()> {
-        println!("Unlocking lock_id: {lock_id:?}, owner: {owner}");
-        let mut need_remove = false;
-
-        {
+        // first, get the entry without holding the write lock on the map
+        let entry = {
             let locks_guard = self.locks.read().await;
-            if let Some(entry) = locks_guard.get(lock_id) {
-                println!("Found lock entry, attempting to acquire write lock...");
-                match entry.try_write() {
-                    Ok(mut entry_guard) => {
-                        println!("Successfully acquired write lock for unlock");
-                        // try to release write lock
-                        if entry_guard.writer.as_ref() == Some(&owner.to_string()) {
-                            println!("Releasing write lock for owner: {owner}");
-                            entry_guard.writer = None;
-                        }
-                        // try to release read lock
-                        else if let Some(count) = entry_guard.readers.get_mut(owner) {
-                            println!("Releasing read lock for owner: {owner} (count: {count})");
-                            *count -= 1;
-                            if *count == 0 {
-                                entry_guard.readers.remove(owner);
-                                println!("Removed owner {owner} from readers");
-                            }
-                        } else {
-                            println!("Owner {owner} not found in writers or readers");
-                        }
-                        // check if need to remove
-                        if entry_guard.readers.is_empty() && entry_guard.writer.is_none() {
-                            println!("Lock entry is empty, marking for removal");
-                            entry_guard.expires_at = None;
-                            need_remove = true;
-                        } else {
-                            println!(
-                                "Lock entry still has content: writer={:?}, readers={:?}",
-                                entry_guard.writer, entry_guard.readers
-                            );
-                        }
-                    }
-                    Err(_) => {
-                        println!("Failed to acquire write lock for unlock - this is the problem!");
-                        return Err(std::io::Error::new(
-                            std::io::ErrorKind::WouldBlock,
-                            "Failed to acquire write lock for unlock",
-                        ));
-                    }
+            match locks_guard.get(lock_id) {
+                Some(e) => e.clone(),
+                None => return Err(std::io::Error::new(std::io::ErrorKind::NotFound, "Lock entry not found")),
+            }
+        };
+
+        let mut need_remove = false;
+        let (notify_writers, notify_readers, writer_pending, writer_none) = {
+            let mut entry_guard = entry.write().await;
+
+            // try to release write lock
+            if entry_guard.writer.as_ref() == Some(&owner.to_string()) {
+                entry_guard.writer = None;
+            }
+            // try to release read lock
+            else if let Some(count) = entry_guard.readers.get_mut(owner) {
+                *count -= 1;
+                if *count == 0 {
+                    entry_guard.readers.remove(owner);
                 }
             } else {
-                println!("Lock entry not found for lock_id: {lock_id:?}");
+                // owner not found, treat as no-op
             }
+
+            // check if need to remove
+            if entry_guard.readers.is_empty() && entry_guard.writer.is_none() {
+                entry_guard.expires_at = None;
+                need_remove = true;
+            }
+
+            // capture notifications and state
+            (
+                entry_guard.notify_writers.clone(),
+                entry_guard.notify_readers.clone(),
+                entry_guard.writer_pending,
+                entry_guard.writer.is_none(),
+            )
+        };
+
+        if writer_pending > 0 && writer_none {
+            // Wake a single writer to preserve fairness and avoid thundering herd
+            notify_writers.notify_one();
+        } else if writer_none {
+            // No writers waiting, allow readers to proceed
+            notify_readers.notify_waiters();
         }
 
-        // only here, entry's Ref is really dropped, can safely remove
         if need_remove {
-            println!("Removing lock entry from map...");
             let mut locks_guard = self.locks.write().await;
-            let removed = locks_guard.remove(lock_id);
-            println!("Lock entry removed: {:?}", removed.is_some());
+            let _ = locks_guard.remove(lock_id);
         }
-        println!("Unlock operation completed");
         Ok(())
     }
 
     /// unlock by LockId - smart release (compatible with old interface, but may be inaccurate)
     pub async fn unlock_by_id(&self, lock_id: &crate::types::LockId) -> std::io::Result<()> {
-        let mut need_remove = false;
-
-        {
+        let entry = {
             let locks_guard = self.locks.read().await;
-            if let Some(entry) = locks_guard.get(lock_id) {
-                if let Ok(mut entry_guard) = entry.try_write() {
-                    // release write lock first
-                    if entry_guard.writer.is_some() {
-                        entry_guard.writer = None;
-                    }
-                    // if no write lock, release first read lock
-                    else if let Some((owner, _)) = entry_guard.readers.iter().next() {
-                        let owner = owner.clone();
-                        if let Some(count) = entry_guard.readers.get_mut(&owner) {
-                            *count -= 1;
-                            if *count == 0 {
-                                entry_guard.readers.remove(&owner);
-                            }
-                        }
-                    }
+            match locks_guard.get(lock_id) {
+                Some(e) => e.clone(),
+                None => return Ok(()), // nothing to do
+            }
+        };
 
-                    // if completely idle, clean entry
-                    if entry_guard.readers.is_empty() && entry_guard.writer.is_none() {
-                        entry_guard.expires_at = None;
-                        need_remove = true;
+        let mut need_remove = false;
+        let (notify_writers, notify_readers, writer_pending, writer_none) = {
+            let mut entry_guard = entry.write().await;
+
+            // release write lock first
+            if entry_guard.writer.is_some() {
+                entry_guard.writer = None;
+            }
+            // if no write lock, release first read lock
+            else if let Some((owner, _)) = entry_guard.readers.iter().next() {
+                let owner = owner.clone();
+                if let Some(count) = entry_guard.readers.get_mut(&owner) {
+                    *count -= 1;
+                    if *count == 0 {
+                        entry_guard.readers.remove(&owner);
                     }
                 }
             }
+
+            if entry_guard.readers.is_empty() && entry_guard.writer.is_none() {
+                entry_guard.expires_at = None;
+                need_remove = true;
+            }
+
+            (
+                entry_guard.notify_writers.clone(),
+                entry_guard.notify_readers.clone(),
+                entry_guard.writer_pending,
+                entry_guard.writer.is_none(),
+            )
+        };
+
+        if writer_pending > 0 && writer_none {
+            notify_writers.notify_one();
+        } else if writer_none {
+            notify_readers.notify_waiters();
         }
 
         if need_remove {
             let mut locks_guard = self.locks.write().await;
-            locks_guard.remove(lock_id);
+            let _ = locks_guard.remove(lock_id);
         }
         Ok(())
     }
 
     /// runlock by LockId and owner - need to specify owner to correctly unlock read lock
     pub async fn runlock_by_id_and_owner(&self, lock_id: &crate::types::LockId, owner: &str) -> std::io::Result<()> {
-        let mut need_remove = false;
-
-        {
+        let entry = {
             let locks_guard = self.locks.read().await;
-            if let Some(entry) = locks_guard.get(lock_id) {
-                if let Ok(mut entry_guard) = entry.try_write() {
-                    // release read lock
-                    if let Some(count) = entry_guard.readers.get_mut(owner) {
-                        *count -= 1;
-                        if *count == 0 {
-                            entry_guard.readers.remove(owner);
-                        }
-                    }
+            match locks_guard.get(lock_id) {
+                Some(e) => e.clone(),
+                None => return Ok(()),
+            }
+        };
 
-                    // if completely idle, clean entry
-                    if entry_guard.readers.is_empty() && entry_guard.writer.is_none() {
-                        entry_guard.expires_at = None;
-                        need_remove = true;
-                    }
+        let mut need_remove = false;
+        let (notify_writers, notify_readers, writer_pending, writer_none) = {
+            let mut entry_guard = entry.write().await;
+
+            // release read lock
+            if let Some(count) = entry_guard.readers.get_mut(owner) {
+                *count -= 1;
+                if *count == 0 {
+                    entry_guard.readers.remove(owner);
                 }
             }
+
+            if entry_guard.readers.is_empty() && entry_guard.writer.is_none() {
+                entry_guard.expires_at = None;
+                need_remove = true;
+            }
+
+            (
+                entry_guard.notify_writers.clone(),
+                entry_guard.notify_readers.clone(),
+                entry_guard.writer_pending,
+                entry_guard.writer.is_none(),
+            )
+        };
+
+        if writer_pending > 0 && writer_none {
+            notify_writers.notify_waiters();
+        } else if writer_none {
+            notify_readers.notify_waiters();
         }
 
         if need_remove {
             let mut locks_guard = self.locks.write().await;
-            locks_guard.remove(lock_id);
+            let _ = locks_guard.remove(lock_id);
         }
         Ok(())
     }
 
     /// runlock by LockId - smart release read lock (compatible with old interface)
     pub async fn runlock_by_id(&self, lock_id: &crate::types::LockId) -> std::io::Result<()> {
-        let mut need_remove = false;
-
-        {
+        let entry = {
             let locks_guard = self.locks.read().await;
-            if let Some(entry) = locks_guard.get(lock_id) {
-                if let Ok(mut entry_guard) = entry.try_write() {
-                    // release first read lock
-                    if let Some((owner, _)) = entry_guard.readers.iter().next() {
-                        let owner = owner.clone();
-                        if let Some(count) = entry_guard.readers.get_mut(&owner) {
-                            *count -= 1;
-                            if *count == 0 {
-                                entry_guard.readers.remove(&owner);
-                            }
-                        }
-                    }
+            match locks_guard.get(lock_id) {
+                Some(e) => e.clone(),
+                None => return Ok(()),
+            }
+        };
 
-                    // if completely idle, clean entry
-                    if entry_guard.readers.is_empty() && entry_guard.writer.is_none() {
-                        entry_guard.expires_at = None;
-                        need_remove = true;
+        let mut need_remove = false;
+        let (notify_writers, notify_readers, writer_pending, writer_none) = {
+            let mut entry_guard = entry.write().await;
+
+            // release first read lock
+            if let Some((owner, _)) = entry_guard.readers.iter().next() {
+                let owner = owner.clone();
+                if let Some(count) = entry_guard.readers.get_mut(&owner) {
+                    *count -= 1;
+                    if *count == 0 {
+                        entry_guard.readers.remove(&owner);
                     }
                 }
             }
+
+            if entry_guard.readers.is_empty() && entry_guard.writer.is_none() {
+                entry_guard.expires_at = None;
+                need_remove = true;
+            }
+
+            (
+                entry_guard.notify_writers.clone(),
+                entry_guard.notify_readers.clone(),
+                entry_guard.writer_pending,
+                entry_guard.writer.is_none(),
+            )
+        };
+
+        if writer_pending > 0 && writer_none {
+            notify_writers.notify_waiters();
+        } else if writer_none {
+            notify_readers.notify_waiters();
         }
 
         if need_remove {
             let mut locks_guard = self.locks.write().await;
-            locks_guard.remove(lock_id);
+            let _ = locks_guard.remove(lock_id);
         }
         Ok(())
     }

crates/lock/src/namespace.rs

@@ -19,6 +19,7 @@ use std::time::Duration;
 use crate::{
     client::LockClient,
     error::{LockError, Result},
+    guard::LockGuard,
     types::{LockId, LockInfo, LockRequest, LockResponse, LockStatus, LockType},
 };
 
@@ -60,6 +61,22 @@ impl NamespaceLock {
         }
     }
 
+    /// Create namespace lock with clients and an explicit quorum size.
+    /// Quorum will be clamped into [1, clients.len()]. For single client, quorum is always 1.
+    pub fn with_clients_and_quorum(namespace: String, clients: Vec<Arc<dyn LockClient>>, quorum: usize) -> Self {
+        let q = if clients.len() <= 1 {
+            1
+        } else {
+            quorum.clamp(1, clients.len())
+        };
+
+        Self {
+            clients,
+            namespace,
+            quorum: q,
+        }
+    }
+
     /// Create namespace lock with client (compatibility)
     pub fn with_client(client: Arc<dyn LockClient>) -> Self {
         Self::with_clients("default".to_string(), vec![client])
@@ -86,54 +103,77 @@ impl NamespaceLock {
             return self.clients[0].acquire_lock(request).await;
         }
 
-        // Two-phase commit for distributed lock acquisition
-        self.acquire_lock_with_2pc(request).await
+        // Quorum-based acquisition for distributed mode
+        let (resp, _idxs) = self.acquire_lock_quorum(request).await?;
+        Ok(resp)
     }
 
-    /// Two-phase commit lock acquisition: all nodes must succeed or all fail
-    async fn acquire_lock_with_2pc(&self, request: &LockRequest) -> Result<LockResponse> {
-        // Phase 1: Prepare - try to acquire lock on all clients
-        let futures: Vec<_> = self
+    /// Acquire a lock and return a RAII guard that will release asynchronously on Drop.
+    /// This is a thin wrapper around `acquire_lock` and will only create a guard when acquisition succeeds.
+    pub async fn acquire_guard(&self, request: &LockRequest) -> Result<Option<LockGuard>> {
+        if self.clients.is_empty() {
+            return Err(LockError::internal("No lock clients available"));
+        }
+
+        if self.clients.len() == 1 {
+            let resp = self.clients[0].acquire_lock(request).await?;
+            if resp.success {
+                return Ok(Some(LockGuard::new(
+                    LockId::new_deterministic(&request.resource),
+                    vec![self.clients[0].clone()],
+                )));
+            }
+            return Ok(None);
+        }
+
+        let (resp, idxs) = self.acquire_lock_quorum(request).await?;
+        if resp.success {
+            let subset: Vec<_> = idxs.into_iter().filter_map(|i| self.clients.get(i).cloned()).collect();
+            Ok(Some(LockGuard::new(LockId::new_deterministic(&request.resource), subset)))
+        } else {
+            Ok(None)
+        }
+    }
+
+    /// Convenience: acquire exclusive lock as a guard
+    pub async fn lock_guard(&self, resource: &str, owner: &str, timeout: Duration, ttl: Duration) -> Result<Option<LockGuard>> {
+        let req = LockRequest::new(self.get_resource_key(resource), LockType::Exclusive, owner)
+            .with_acquire_timeout(timeout)
+            .with_ttl(ttl);
+        self.acquire_guard(&req).await
+    }
+
+    /// Convenience: acquire shared lock as a guard
+    pub async fn rlock_guard(&self, resource: &str, owner: &str, timeout: Duration, ttl: Duration) -> Result<Option<LockGuard>> {
+        let req = LockRequest::new(self.get_resource_key(resource), LockType::Shared, owner)
+            .with_acquire_timeout(timeout)
+            .with_ttl(ttl);
+        self.acquire_guard(&req).await
+    }
+
+    /// Quorum-based lock acquisition: success if at least `self.quorum` clients succeed.
+    /// Returns the LockResponse and the indices of clients that acquired the lock.
+    async fn acquire_lock_quorum(&self, request: &LockRequest) -> Result<(LockResponse, Vec<usize>)> {
+        let futs: Vec<_> = self
             .clients
             .iter()
             .enumerate()
-            .map(|(idx, client)| async move {
-                let result = client.acquire_lock(request).await;
-                (idx, result)
-            })
+            .map(|(idx, client)| async move { (idx, client.acquire_lock(request).await) })
             .collect();
 
-        let results = futures::future::join_all(futures).await;
+        let results = futures::future::join_all(futs).await;
         let mut successful_clients = Vec::new();
-        let mut failed_clients = Vec::new();
 
-        // Collect results
-        for (idx, result) in results {
-            match result {
-                Ok(response) if response.success => {
+        for (idx, res) in results {
+            if let Ok(resp) = res {
+                if resp.success {
                     successful_clients.push(idx);
                 }
-                _ => {
-                    failed_clients.push(idx);
-                }
             }
         }
 
-        // Check if we have enough successful acquisitions for quorum
         if successful_clients.len() >= self.quorum {
-            // Phase 2a: Commit - we have quorum, but need to ensure consistency
-            // If not all clients succeeded, we need to rollback for consistency
-            if successful_clients.len() < self.clients.len() {
-                // Rollback all successful acquisitions to maintain consistency
-                self.rollback_acquisitions(request, &successful_clients).await;
-                return Ok(LockResponse::failure(
-                    "Partial success detected, rolled back for consistency".to_string(),
-                    Duration::ZERO,
-                ));
-            }
-
-            // All clients succeeded - lock acquired successfully
-            Ok(LockResponse::success(
+            let resp = LockResponse::success(
                 LockInfo {
                     id: LockId::new_deterministic(&request.resource),
                     resource: request.resource.clone(),
@@ -148,16 +188,17 @@ impl NamespaceLock {
                     wait_start_time: None,
                 },
                 Duration::ZERO,
-            ))
+            );
+            Ok((resp, successful_clients))
         } else {
-            // Phase 2b: Abort - insufficient quorum, rollback any successful acquisitions
             if !successful_clients.is_empty() {
                 self.rollback_acquisitions(request, &successful_clients).await;
             }
-            Ok(LockResponse::failure(
+            let resp = LockResponse::failure(
                 format!("Failed to acquire quorum: {}/{} required", successful_clients.len(), self.quorum),
                 Duration::ZERO,
-            ))
+            );
+            Ok((resp, Vec::new()))
         }
     }
 
@@ -420,6 +461,33 @@ mod tests {
         assert!(result.is_ok());
     }
 
+    #[tokio::test]
+    async fn test_guard_acquire_and_drop_release() {
+        let ns_lock = NamespaceLock::with_client(Arc::new(LocalClient::new()));
+
+        // Acquire guard
+        let guard = ns_lock
+            .lock_guard("guard-resource", "owner", Duration::from_millis(100), Duration::from_secs(5))
+            .await
+            .unwrap();
+        assert!(guard.is_some());
+        let lock_id = guard.as_ref().unwrap().lock_id().clone();
+
+        // Drop guard to trigger background release
+        drop(guard);
+
+        // Give background worker a moment to process
+        tokio::time::sleep(Duration::from_millis(50)).await;
+
+        // Re-acquire should succeed (previous lock released)
+        let req = LockRequest::new(&lock_id.resource, LockType::Exclusive, "owner").with_ttl(Duration::from_secs(2));
+        let resp = ns_lock.acquire_lock(&req).await.unwrap();
+        assert!(resp.success);
+
+        // Cleanup
+        let _ = ns_lock.release_lock(&LockId::new_deterministic(&lock_id.resource)).await;
+    }
+
     #[tokio::test]
     async fn test_connection_health() {
         let local_lock = NamespaceLock::new("test-namespace".to_string());
@@ -502,9 +570,11 @@ mod tests {
         let client2: Arc<dyn LockClient> = Arc::new(LocalClient::new());
         let clients = vec![client1, client2];
 
-        let ns_lock = NamespaceLock::with_clients("test-namespace".to_string(), clients);
+        // LocalClient shares a global in-memory map. For exclusive locks, only one can acquire at a time.
+        // In real distributed setups the quorum should be tied to EC write quorum. Here we use quorum=1 for success.
+        let ns_lock = NamespaceLock::with_clients_and_quorum("test-namespace".to_string(), clients, 1);
 
-        let request = LockRequest::new("test-resource", LockType::Exclusive, "test_owner").with_ttl(Duration::from_secs(10));
+        let request = LockRequest::new("test-resource", LockType::Shared, "test_owner").with_ttl(Duration::from_secs(2));
 
         // This should succeed only if ALL clients can acquire the lock
         let response = ns_lock.acquire_lock(&request).await.unwrap();

crates/mcp/Dockerfile

@@ -0,0 +1,15 @@
+FROM rust:1.88 AS builder
+
+WORKDIR /build
+
+COPY . .
+
+RUN cargo build --release -p rustfs-mcp
+
+FROM debian:bookworm-slim
+
+WORKDIR /app
+
+COPY --from=builder /build/target/release/rustfs-mcp /app/
+
+ENTRYPOINT ["/app/rustfs-mcp"]

crates/mcp/README.md

@@ -98,7 +98,9 @@ rustfs-mcp --log-level debug --region us-west-2
 ```
 
 ### Integration with chat client
+
 #### Option 1: Using Command Line Arguments
+
 ```json
 {
   "mcpServers": {
@@ -116,6 +118,7 @@ rustfs-mcp --log-level debug --region us-west-2
 ```
 
 #### Option 2: Using Environment Variables
+
 ```json
 {
   "mcpServers": {
@@ -130,26 +133,84 @@ rustfs-mcp --log-level debug --region us-west-2
   }
 }
 ```
+
+### Using MCP with Docker
+
+#### Docker image build
+
+Using MCP with docker will simply the usage of rustfs mcp. Building the docker image with below command:
+
+```
+docker build -f Dockerfile -t rustfs/rustfs-mcp ../../
+```
+
+Alternatively, if you want to build the image from the rustfs codebase root directory,run the command:
+
+```
+docker build -f crates/mcp/Dockerfile -t rustfs/rustfs-mcp .
+```
+
+#### IDE Configuration
+
+Adding the following content in IDE MCP settings:
+
+```
+{
+  "mcpServers": {
+    "rustfs-mcp": {
+      "command": "docker",
+      "args": [
+        "run",
+        "--rm",
+        "-i",
+        "-e",
+        "AWS_ACCESS_KEY_ID",
+        "-e",
+        "AWS_SECRET_ACCESS_KEY",
+        "-e",
+        "AWS_REGION",
+        "-e",
+        "AWS_ENDPOINT_URL",
+        "rustfs/rustfs-mcp"
+      ],
+      "env": {
+        "AWS_ACCESS_KEY_ID": "rustfs_access_key",
+        "AWS_SECRET_ACCESS_KEY": "rustfs_secret_key",
+        "AWS_REGION": "cn-east-1",
+        "AWS_ENDPOINT_URL": "rustfs_instance_url"
+      }
+    }
+  }
+}
+```
+
+If success, MCP configure page will show the [available tools](#️-available-tools).
+
 ## 🛠️ Available Tools
 
 The MCP server exposes the following tools that AI assistants can use:
 
 ### `list_buckets`
+
 List all S3 buckets accessible with the configured credentials.
 
 **Parameters:** None
 
 ### `list_objects`
+
 List objects in an S3 bucket with optional prefix filtering.
 
 **Parameters:**
+
 - `bucket_name` (string): Name of the S3 bucket
 - `prefix` (string, optional): Prefix to filter objects
 
 ### `upload_file`
+
 Upload a local file to S3 with automatic MIME type detection.
 
 **Parameters:**
+
 - `local_file_path` (string): Path to the local file
 - `bucket_name` (string): Target S3 bucket
 - `object_key` (string): S3 object key (destination path)
@@ -158,9 +219,11 @@ Upload a local file to S3 with automatic MIME type detection.
 - `cache_control` (string, optional): Cache control header
 
 ### `get_object`
+
 Retrieve an object from S3 with two operation modes: read content directly or download to a file.
 
 **Parameters:**
+
 - `bucket_name` (string): Source S3 bucket
 - `object_key` (string): S3 object key
 - `version_id` (string, optional): Version ID for versioned objects

crates/notify/Cargo.toml

@@ -26,7 +26,7 @@ categories = ["web-programming", "development-tools", "filesystem"]
 documentation = "https://docs.rs/rustfs-notify/latest/rustfs_notify/"
 
 [dependencies]
-rustfs-config = { workspace = true, features = ["notify"] }
+rustfs-config = { workspace = true, features = ["notify", "constants"] }
 rustfs-ecstore = { workspace = true }
 rustfs-utils = { workspace = true, features = ["path", "sys"] }
 async-trait = { workspace = true }

crates/notify/src/registry.rs

@@ -19,7 +19,7 @@ use crate::{
     target::Target,
 };
 use futures::stream::{FuturesUnordered, StreamExt};
-use rustfs_config::notify::{ENABLE_KEY, ENABLE_ON, NOTIFY_ROUTE_PREFIX};
+use rustfs_config::notify::{ENABLE_KEY, NOTIFY_ROUTE_PREFIX};
 use rustfs_config::{DEFAULT_DELIMITER, ENV_PREFIX};
 use rustfs_ecstore::config::{Config, KVS};
 use std::collections::{HashMap, HashSet};
@@ -111,10 +111,10 @@ impl TargetRegistry {
             // 3.1. Instance discovery: Based on the '..._ENABLE_INSTANCEID' format
             let enable_prefix = format!("{ENV_PREFIX}{NOTIFY_ROUTE_PREFIX}{target_type}_{ENABLE_KEY}_").to_uppercase();
             for (key, value) in &all_env {
-                if value.eq_ignore_ascii_case(ENABLE_ON)
-                    || value.eq_ignore_ascii_case("true")
-                    || value.eq_ignore_ascii_case("1")
-                    || value.eq_ignore_ascii_case("yes")
+                if value.eq_ignore_ascii_case(rustfs_config::EnableState::One.as_str())
+                    || value.eq_ignore_ascii_case(rustfs_config::EnableState::On.as_str())
+                    || value.eq_ignore_ascii_case(rustfs_config::EnableState::True.as_str())
+                    || value.eq_ignore_ascii_case(rustfs_config::EnableState::Yes.as_str())
                 {
                     if let Some(id) = key.strip_prefix(&enable_prefix) {
                         if !id.is_empty() {
@@ -202,10 +202,10 @@ impl TargetRegistry {
                 let enabled = merged_config
                     .lookup(ENABLE_KEY)
                     .map(|v| {
-                        v.eq_ignore_ascii_case(ENABLE_ON)
-                            || v.eq_ignore_ascii_case("true")
-                            || v.eq_ignore_ascii_case("1")
-                            || v.eq_ignore_ascii_case("yes")
+                        v.eq_ignore_ascii_case(rustfs_config::EnableState::One.as_str())
+                            || v.eq_ignore_ascii_case(rustfs_config::EnableState::On.as_str())
+                            || v.eq_ignore_ascii_case(rustfs_config::EnableState::True.as_str())
+                            || v.eq_ignore_ascii_case(rustfs_config::EnableState::Yes.as_str())
                     })
                     .unwrap_or(false);
 

crates/s3select-query/src/data_source/table_source.rs

@@ -103,7 +103,7 @@ impl TableSource for TableSourceAdapter {
     }
 
     /// Called by [`InlineTableScan`]
-    fn get_logical_plan(&self) -> Option<Cow<LogicalPlan>> {
+    fn get_logical_plan(&self) -> Option<Cow<'_, LogicalPlan>> {
         Some(Cow::Owned(self.plan.clone()))
     }
 }

crates/signer/Cargo.toml

@@ -31,14 +31,11 @@ bytes = { workspace = true }
 http.workspace = true
 time.workspace = true
 hyper.workspace = true
-serde.workspace = true
 serde_urlencoded.workspace = true
 rustfs-utils = { workspace = true, features = ["full"] }
 s3s.workspace = true
 
 [dev-dependencies]
-tempfile = { workspace = true }
-rand = { workspace = true }
 
 [lints]
 workspace = true

crates/utils/src/certs.rs

@@ -21,7 +21,7 @@ use std::collections::HashMap;
 use std::io::Error;
 use std::path::Path;
 use std::sync::Arc;
-use std::{fs, io};
+use std::{env, fs, io};
 use tracing::{debug, warn};
 
 /// Load public certificate from file.
@@ -194,6 +194,19 @@ pub fn create_multi_cert_resolver(
     })
 }
 
+/// Checks if TLS key logging is enabled.
+pub fn tls_key_log() -> bool {
+    env::var("RUSTFS_TLS_KEYLOG")
+        .map(|v| {
+            let v = v.trim();
+            v.eq_ignore_ascii_case("1")
+                || v.eq_ignore_ascii_case("on")
+                || v.eq_ignore_ascii_case("true")
+                || v.eq_ignore_ascii_case("yes")
+        })
+        .unwrap_or(false)
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;

crates/utils/src/lib.rs

@@ -22,6 +22,7 @@ pub mod net;
 #[cfg(feature = "net")]
 pub use net::*;
 
+#[cfg(all(feature = "net", feature = "io"))]
 pub mod retry;
 
 #[cfg(feature = "io")]

rustfs/Cargo.toml

@@ -54,8 +54,6 @@ rustfs-s3select-query = { workspace = true }
 atoi = { workspace = true }
 atomic_enum = { workspace = true }
 axum.workspace = true
-axum-extra = { workspace = true }
-axum-server = { workspace = true }
 async-trait = { workspace = true }
 bytes = { workspace = true }
 chrono = { workspace = true }

rustfs/src/admin/handlers.rs

@@ -53,7 +53,6 @@ use s3s::stream::{ByteStream, DynByteStream};
 use s3s::{Body, S3Error, S3Request, S3Response, S3Result, s3_error};
 use s3s::{S3ErrorCode, StdError};
 use serde::{Deserialize, Serialize};
-use tracing::debug;
 // use serde_json::to_vec;
 use std::collections::{HashMap, HashSet};
 use std::path::PathBuf;
@@ -65,6 +64,7 @@ use tokio::sync::mpsc::{self};
 use tokio::time::interval;
 use tokio::{select, spawn};
 use tokio_stream::wrappers::ReceiverStream;
+use tracing::debug;
 use tracing::{error, info, warn};
 // use url::UrlQuery;
 
@@ -81,6 +81,7 @@ pub mod trace;
 pub mod user;
 use urlencoding::decode;
 
+#[allow(dead_code)]
 #[derive(Debug, Serialize, Default)]
 #[serde(rename_all = "PascalCase", default)]
 pub struct AccountInfo {

rustfs/src/admin/handlers/event.rs

@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#![allow(dead_code)]
+
 use crate::admin::router::Operation;
 use crate::auth::{check_key_valid, get_session_token};
 use http::{HeaderMap, StatusCode};

rustfs/src/admin/handlers/tier.rs

@@ -341,6 +341,7 @@ impl Operation for RemoveTier {
     }
 }
 
+#[allow(dead_code)]
 pub struct VerifyTier {}
 #[async_trait::async_trait]
 impl Operation for VerifyTier {

rustfs/src/admin/handlers/trace.rs

@@ -22,6 +22,7 @@ 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();
     st_opts
@@ -31,6 +32,7 @@ fn extract_trace_options(uri: &Uri) -> S3Result<ServiceTraceOpts> {
     Ok(st_opts)
 }
 
+#[allow(dead_code)]
 pub struct Trace {}
 
 #[async_trait::async_trait]

rustfs/src/auth.rs

@@ -490,11 +490,9 @@ mod tests {
         let result = check_claims_from_token("", &cred);
 
         // This might fail due to global state dependencies, but should return error about global cred init
-        if result.is_ok() {
-            let claims = result.unwrap();
+        if let Ok(claims) = result {
             assert!(claims.is_empty());
-        } else {
-            let error = result.unwrap_err();
+        } else if let Err(error) = result {
             assert_eq!(error.code(), &S3ErrorCode::InternalError);
             assert!(error.message().unwrap_or("").contains("action cred not init"));
         }

rustfs/src/main.rs

@@ -183,10 +183,6 @@ async fn run(opt: config::Opt) -> Result<()> {
         Error::other(err)
     })?;
 
-    // init scanner and auto heal with unified cancellation token
-    // let _background_services_cancel_token = create_background_services_cancel_token();
-    // init_data_scanner().await;
-    // init_auto_heal().await;
     let _ = create_ahm_services_cancel_token();
 
     // Initialize heal manager with channel processor

rustfs/src/server/http.rs

@@ -14,7 +14,6 @@
 
 // Ensure the correct path for parse_license is imported
 use crate::admin;
-// use crate::admin::console::{CONSOLE_CONFIG, init_console_cfg};
 use crate::auth::IAMAuth;
 use crate::config;
 use crate::server::hybrid::hybrid;
@@ -43,8 +42,6 @@ use std::net::SocketAddr;
 use std::sync::Arc;
 use std::time::Duration;
 use tokio::net::{TcpListener, TcpStream};
-#[cfg(unix)]
-use tokio::signal::unix::{SignalKind, signal};
 use tokio_rustls::TlsAcceptor;
 use tonic::{Request, Status, metadata::MetadataValue};
 use tower::ServiceBuilder;
@@ -63,9 +60,6 @@ pub async fn start_http_server(
     let server_port = server_addr.port();
     let server_address = server_addr.to_string();
 
-    // The listening address and port are obtained from the parameters
-    // let listener = TcpListener::bind(server_address.clone()).await?;
-
     // The listening address and port are obtained from the parameters
     let listener = {
         let mut server_addr = server_addr;
@@ -172,6 +166,7 @@ pub async fn start_http_server(
     tokio::spawn(async move {
         #[cfg(unix)]
         let (mut sigterm_inner, mut sigint_inner) = {
+            use tokio::signal::unix::{SignalKind, signal};
             // Unix platform specific code
             let sigterm_inner = signal(SignalKind::terminate()).expect("Failed to create SIGTERM signal handler");
             let sigint_inner = signal(SignalKind::interrupt()).expect("Failed to create SIGINT signal handler");
@@ -292,32 +287,55 @@ async fn setup_tls_acceptor(tls_path: &str) -> Result<Option<TlsAcceptor>> {
 
     debug!("Found TLS directory, checking for certificates");
 
-    // 1. Try to load all certificates from the directory (multi-cert support)
+    // Make sure to use a modern encryption suite
+    let _ = rustls::crypto::aws_lc_rs::default_provider().install_default();
+
+    // 1. Attempt to load all certificates in the directory (multi-certificate support, for SNI)
     if let Ok(cert_key_pairs) = rustfs_utils::load_all_certs_from_directory(tls_path) {
         if !cert_key_pairs.is_empty() {
-            debug!("Found {} certificates, creating multi-cert resolver", cert_key_pairs.len());
-            let _ = rustls::crypto::aws_lc_rs::default_provider().install_default();
+            debug!("Found {} certificates, creating SNI-aware multi-cert resolver", cert_key_pairs.len());
+
+            // Create an SNI-enabled certificate resolver
+            let resolver = rustfs_utils::create_multi_cert_resolver(cert_key_pairs)?;
+
+            // Configure the server to enable SNI support
             let mut server_config = ServerConfig::builder()
                 .with_no_client_auth()
-                .with_cert_resolver(Arc::new(rustfs_utils::create_multi_cert_resolver(cert_key_pairs)?));
+                .with_cert_resolver(Arc::new(resolver));
+
+            // Configure ALPN protocol priority
             server_config.alpn_protocols = vec![b"h2".to_vec(), b"http/1.1".to_vec(), b"http/1.0".to_vec()];
+
+            // Log SNI requests
+            if rustfs_utils::tls_key_log() {
+                server_config.key_log = Arc::new(rustls::KeyLogFile::new());
+            }
+
             return Ok(Some(TlsAcceptor::from(Arc::new(server_config))));
         }
     }
 
-    // 2. Fallback to legacy single certificate mode
+    // 2. Revert to the traditional single-certificate mode
     let key_path = format!("{tls_path}/{RUSTFS_TLS_KEY}");
     let cert_path = format!("{tls_path}/{RUSTFS_TLS_CERT}");
     if tokio::try_join!(tokio::fs::metadata(&key_path), tokio::fs::metadata(&cert_path)).is_ok() {
         debug!("Found legacy single TLS certificate, starting with HTTPS");
-        let _ = rustls::crypto::aws_lc_rs::default_provider().install_default();
         let certs = rustfs_utils::load_certs(&cert_path).map_err(|e| rustfs_utils::certs_error(e.to_string()))?;
         let key = rustfs_utils::load_private_key(&key_path).map_err(|e| rustfs_utils::certs_error(e.to_string()))?;
+
         let mut server_config = ServerConfig::builder()
             .with_no_client_auth()
             .with_single_cert(certs, key)
             .map_err(|e| rustfs_utils::certs_error(e.to_string()))?;
+
+        // Configure ALPN protocol priority
         server_config.alpn_protocols = vec![b"h2".to_vec(), b"http/1.1".to_vec(), b"http/1.0".to_vec()];
+
+        // Log SNI requests
+        if rustfs_utils::tls_key_log() {
+            server_config.key_log = Arc::new(rustls::KeyLogFile::new());
+        }
+
         return Ok(Some(TlsAcceptor::from(Arc::new(server_config))));
     }
 
@@ -398,6 +416,10 @@ fn process_connection(
         // Decide whether to handle HTTPS or HTTP connections based on the existence of TLS Acceptor
         if let Some(acceptor) = tls_acceptor {
             debug!("TLS handshake start");
+            let peer_addr = socket
+                .peer_addr()
+                .ok()
+                .map_or_else(|| "unknown".to_string(), |addr| addr.to_string());
             match acceptor.accept(socket).await {
                 Ok(tls_socket) => {
                     debug!("TLS handshake successful");
@@ -408,8 +430,44 @@ fn process_connection(
                     }
                 }
                 Err(err) => {
-                    error!(?err, "TLS handshake failed");
-                    return; // Failed to end the task directly
+                    // Detailed analysis of the reasons why the TLS handshake fails
+                    let err_str = err.to_string();
+                    let mut key_failure_type_str: &str = "UNKNOWN";
+                    if err_str.contains("unexpected EOF") || err_str.contains("handshake eof") {
+                        warn!(peer_addr = %peer_addr, "TLS handshake failed. If this client needs HTTP, it should connect to the HTTP port instead");
+                        key_failure_type_str = "UNEXPECTED_EOF";
+                    } else if err_str.contains("protocol version") {
+                        error!(
+                            peer_addr = %peer_addr,
+                            "TLS handshake failed due to protocol version mismatch: {}", err
+                        );
+                        key_failure_type_str = "PROTOCOL_VERSION";
+                    } else if err_str.contains("certificate") {
+                        error!(
+                            peer_addr = %peer_addr,
+                            "TLS handshake failed due to certificate issues: {}", err
+                        );
+                        key_failure_type_str = "CERTIFICATE";
+                    } else {
+                        error!(
+                            peer_addr = %peer_addr,
+                            "TLS handshake failed: {}", err
+                        );
+                    }
+                    info!(
+                        counter.rustfs_tls_handshake_failures = 1_u64,
+                        key_failure_type = key_failure_type_str,
+                        "TLS handshake failure metric"
+                    );
+                    // Record detailed diagnostic information
+                    debug!(
+                        peer_addr = %peer_addr,
+                        error_type = %std::any::type_name_of_val(&err),
+                        error_details = %err,
+                        "TLS handshake failure details"
+                    );
+
+                    return;
                 }
             }
             debug!("TLS handshake success");

rustfs/src/storage/ecfs.rs

@@ -21,7 +21,7 @@ use crate::error::ApiError;
 use crate::storage::access::ReqInfo;
 use crate::storage::options::copy_dst_opts;
 use crate::storage::options::copy_src_opts;
-use crate::storage::options::{extract_metadata_from_mime, get_opts};
+use crate::storage::options::{extract_metadata_from_mime_with_object_name, get_opts};
 use bytes::Bytes;
 use chrono::DateTime;
 use chrono::Utc;
@@ -1318,7 +1318,7 @@ impl S3 for FS {
         let objects: Vec<ObjectVersion> = object_infos
             .objects
             .iter()
-            .filter(|v| !v.name.is_empty())
+            .filter(|v| !v.name.is_empty() && !v.delete_marker)
             .map(|v| {
                 ObjectVersion {
                     key: Some(v.name.to_owned()),
@@ -1340,6 +1340,19 @@ impl S3 for FS {
             .map(|v| CommonPrefix { prefix: Some(v) })
             .collect();
 
+        let delete_markers = object_infos
+            .objects
+            .iter()
+            .filter(|o| o.delete_marker)
+            .map(|o| DeleteMarkerEntry {
+                key: Some(o.name.clone()),
+                version_id: o.version_id.map(|v| v.to_string()),
+                is_latest: Some(o.is_latest),
+                last_modified: o.mod_time.map(Timestamp::from),
+                ..Default::default()
+            })
+            .collect::<Vec<_>>();
+
         let output = ListObjectVersionsOutput {
             // is_truncated: Some(object_infos.is_truncated),
             max_keys: Some(key_count),
@@ -1348,6 +1361,7 @@ impl S3 for FS {
             prefix: Some(prefix),
             common_prefixes: Some(common_prefixes),
             versions: Some(objects),
+            delete_markers: Some(delete_markers),
             ..Default::default()
         };
 
@@ -1412,7 +1426,7 @@ impl S3 for FS {
 
         let mut metadata = metadata.unwrap_or_default();
 
-        extract_metadata_from_mime(&req.headers, &mut metadata);
+        extract_metadata_from_mime_with_object_name(&req.headers, &mut metadata, Some(&key));
 
         if let Some(tags) = tagging {
             metadata.insert(AMZ_OBJECT_TAGGING.to_owned(), tags);

rustfs/src/storage/options.rs

@@ -186,6 +186,15 @@ pub fn extract_metadata(headers: &HeaderMap<HeaderValue>) -> HashMap<String, Str
 
 /// Extracts metadata from headers and returns it as a HashMap.
 pub fn extract_metadata_from_mime(headers: &HeaderMap<HeaderValue>, metadata: &mut HashMap<String, String>) {
+    extract_metadata_from_mime_with_object_name(headers, metadata, None);
+}
+
+/// Extracts metadata from headers and returns it as a HashMap with object name for MIME type detection.
+pub fn extract_metadata_from_mime_with_object_name(
+    headers: &HeaderMap<HeaderValue>,
+    metadata: &mut HashMap<String, String>,
+    object_name: Option<&str>,
+) {
     for (k, v) in headers.iter() {
         if let Some(key) = k.as_str().strip_prefix("x-amz-meta-") {
             if key.is_empty() {
@@ -210,10 +219,42 @@ pub fn extract_metadata_from_mime(headers: &HeaderMap<HeaderValue>, metadata: &m
     }
 
     if !metadata.contains_key("content-type") {
-        metadata.insert("content-type".to_owned(), "binary/octet-stream".to_owned());
+        let default_content_type = if let Some(obj_name) = object_name {
+            detect_content_type_from_object_name(obj_name)
+        } else {
+            "binary/octet-stream".to_owned()
+        };
+        metadata.insert("content-type".to_owned(), default_content_type);
     }
 }
 
+/// Detects content type from object name based on file extension.
+pub(crate) fn detect_content_type_from_object_name(object_name: &str) -> String {
+    let lower_name = object_name.to_lowercase();
+
+    // Check for Parquet files specifically
+    if lower_name.ends_with(".parquet") {
+        return "application/vnd.apache.parquet".to_owned();
+    }
+
+    // Special handling for other data formats that mime_guess doesn't know
+    if lower_name.ends_with(".avro") {
+        return "application/avro".to_owned();
+    }
+    if lower_name.ends_with(".orc") {
+        return "application/orc".to_owned();
+    }
+    if lower_name.ends_with(".feather") {
+        return "application/feather".to_owned();
+    }
+    if lower_name.ends_with(".arrow") {
+        return "application/arrow".to_owned();
+    }
+
+    // Use mime_guess for standard file types
+    mime_guess::from_path(object_name).first_or_octet_stream().to_string()
+}
+
 /// List of supported headers.
 static SUPPORTED_HEADERS: LazyLock<Vec<&'static str>> = LazyLock::new(|| {
     vec![
@@ -646,4 +687,79 @@ mod tests {
         assert_eq!(metadata.get("cache-control"), Some(&"public".to_string()));
         assert!(!metadata.contains_key("authorization"));
     }
+
+    #[test]
+    fn test_extract_metadata_from_mime_with_parquet_object_name() {
+        let headers = HeaderMap::new();
+        let mut metadata = HashMap::new();
+
+        extract_metadata_from_mime_with_object_name(&headers, &mut metadata, Some("data/test.parquet"));
+
+        assert_eq!(metadata.get("content-type"), Some(&"application/vnd.apache.parquet".to_string()));
+    }
+
+    #[test]
+    fn test_extract_metadata_from_mime_with_various_data_formats() {
+        let test_cases = vec![
+            ("data.parquet", "application/vnd.apache.parquet"),
+            ("data.PARQUET", "application/vnd.apache.parquet"), // 测试大小写不敏感
+            ("file.avro", "application/avro"),
+            ("file.orc", "application/orc"),
+            ("file.feather", "application/feather"),
+            ("file.arrow", "application/arrow"),
+            ("file.json", "application/json"),
+            ("file.csv", "text/csv"),
+            ("file.txt", "text/plain"),
+            ("file.unknownext", "application/octet-stream"), // 使用真正未知的扩展名
+        ];
+
+        for (filename, expected_content_type) in test_cases {
+            let headers = HeaderMap::new();
+            let mut metadata = HashMap::new();
+
+            extract_metadata_from_mime_with_object_name(&headers, &mut metadata, Some(filename));
+
+            assert_eq!(
+                metadata.get("content-type"),
+                Some(&expected_content_type.to_string()),
+                "Failed for filename: {filename}"
+            );
+        }
+    }
+
+    #[test]
+    fn test_extract_metadata_from_mime_with_existing_content_type() {
+        let mut headers = HeaderMap::new();
+        headers.insert("content-type", HeaderValue::from_static("custom/type"));
+
+        let mut metadata = HashMap::new();
+        extract_metadata_from_mime_with_object_name(&headers, &mut metadata, Some("test.parquet"));
+
+        // 应该保留现有的 content-type，不被覆盖
+        assert_eq!(metadata.get("content-type"), Some(&"custom/type".to_string()));
+    }
+
+    #[test]
+    fn test_detect_content_type_from_object_name() {
+        // 测试 Parquet 文件（我们的自定义处理）
+        assert_eq!(detect_content_type_from_object_name("test.parquet"), "application/vnd.apache.parquet");
+        assert_eq!(detect_content_type_from_object_name("TEST.PARQUET"), "application/vnd.apache.parquet");
+
+        // 测试其他自定义数据格式
+        assert_eq!(detect_content_type_from_object_name("data.avro"), "application/avro");
+        assert_eq!(detect_content_type_from_object_name("data.orc"), "application/orc");
+        assert_eq!(detect_content_type_from_object_name("data.feather"), "application/feather");
+        assert_eq!(detect_content_type_from_object_name("data.arrow"), "application/arrow");
+
+        // 测试标准格式（mime_guess 处理）
+        assert_eq!(detect_content_type_from_object_name("data.json"), "application/json");
+        assert_eq!(detect_content_type_from_object_name("data.csv"), "text/csv");
+        assert_eq!(detect_content_type_from_object_name("data.txt"), "text/plain");
+
+        // 测试真正未知的格式（使用一个 mime_guess 不认识的扩展名）
+        assert_eq!(detect_content_type_from_object_name("unknown.unknownext"), "application/octet-stream");
+
+        // 测试没有扩展名的文件
+        assert_eq!(detect_content_type_from_object_name("noextension"), "application/octet-stream");
+    }
 }