ci: pin s3tests Python tools

2026-08-23 04:39:04 +00:00 · 2026-08-23 03:19:26 +08:00
32 changed files with 639 additions and 873 deletions
@@ -1,2 +1,2 @@
-sha256-darwin=e78234828a8893b8bb0b27f57135edaff5066123cd592c38256be117c350d9d0
-sha256-linux=8084e3e013286bdea17ee18fbd44f28ade8ad0fe2726fa644835278819a41082
+sha256-darwin=9f767b37ed8b1c82da62ea441462d75487785c8086e56f08fb6f6cd89c6e2e52
+sha256-linux=fbdaf42b220958d4b1e8880e0f8b5a7992d38e21051bb60596dd4538424757d6
@@ -399,7 +399,7 @@ path = "junit.xml"
 #     `e2e-smoke` (20 fast) and `e2e-repl-nightly` (55 slow) lanes and reserves
 #     it for those, so e2e-full does not double-run it.
 #   * #[ignore]d tests — nextest skips them by default (no --run-ignored); the
-#     manual-localhost:9000 reliant tests are ci-13's migration.
+#     manual-localhost:9000 reliant/policy tests are ci-13's migration.
 #
 # Each e2e test spawns its own single-node rustfs server on a random port with
 # an isolated temp dir (crates/e2e_test/src/common.rs), so the set is
@@ -178,9 +178,14 @@ jobs:

      - name: Install Python tools
        run: |
-          python3 -m pip install --user --upgrade pip awscurl tox
+          python3 -m pip install --user --upgrade pip "awscurl==0.44" "tox==4.60.0"
          echo "$HOME/.local/bin" >> "$GITHUB_PATH"

+      - name: Verify Python tools
+        run: |
+          test "$(python3 -c 'import importlib.metadata as m; print(m.version("awscurl"))')" = "0.44"
+          test "$(python3 -c 'import importlib.metadata as m; print(m.version("tox"))')" = "4.60.0"
+
      - name: Enable buildx
        uses: docker/setup-buildx-action@8d2750c68a42422c14e847fe6c8ac0403b4cbd6f # v3

@@ -30,8 +30,7 @@ make build-docker BUILD_OS=ubuntu22.04
 - Crate membership: `Cargo.toml` `[workspace].members`
 - Architecture, layering, crate map: [ARCHITECTURE.md](ARCHITECTURE.md)
 - Migration guardrails & readiness contracts: [docs/architecture/](docs/architecture/README.md)
- CI workflow steps: `.github/workflows/`; event, timeout, and required-status
-  matrix: [docs/testing/ci-gates.md](docs/testing/ci-gates.md)
+- CI gates: `.github/workflows/ci.yml` (source of truth; never copy its steps into docs)
 - Test-layer taxonomy, per-layer entry commands, serial/nextest rules, flake
  policy: [docs/testing/README.md](docs/testing/README.md)
 - Tier/ILM transition debugging (xl.meta inspection, versionId tracing):
@@ -70,8 +70,6 @@ make pre-pr

 > For the full test-layer taxonomy (unit / ecstore black-box / e2e / s3s-e2e / S3 compatibility / chaos / fuzz / bench), each layer's entry command, the naming conventions the migration gate depends on, and the serial/nextest rules, see [docs/testing/README.md](docs/testing/README.md).

-> For the event, timeout, required-status, and local reproduction matrix, see [docs/testing/ci-gates.md](docs/testing/ci-gates.md).
-
 ### 🔒 Automated Pre-commit Hooks
 #### What `make pre-commit` and `make pre-pr` actually run

@@ -1858,9 +1858,9 @@ dependencies = [

 [[package]]
 name = "cc"
-version = "1.4.4"
+version = "1.4.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0ad534f4357a5264cce5019c989cf66a4f0dc4e0d1b1d15f8aacec0ff7360273"
+checksum = "509591b7bcd67f4ef775afad7662703b4935daaa6ec0e5605cfb1090b32a2b6d"
 dependencies = [
 "find-msvc-tools",
 "jobserver",
@@ -2522,6 +2522,12 @@ dependencies = [
 "subtle",
 ]

+[[package]]
+name = "cty"
+version = "0.2.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b365fabc795046672053e29c954733ec3b05e4be654ab130fe8f1f94d7051f35"
+
 [[package]]
 name = "curve25519-dalek"
 version = "4.1.3"
@@ -5982,6 +5988,15 @@ version = "0.2.16"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b6d2cec3eae94f9f509c767b45932f1ada8350c4bdb85af2fcab4a3c14807981"

+[[package]]
+name = "libmimalloc-sys"
+version = "0.1.49"
+source = "git+https://github.com/xonatius/mimalloc_rust.git?rev=6d4c41bb10c6d9da1d1b6f07b38c4cc051667f11#6d4c41bb10c6d9da1d1b6f07b38c4cc051667f11"
+dependencies = [
+ "cc",
+ "cty",
+]
+
 [[package]]
 name = "libredox"
 version = "0.1.20"
@@ -6382,6 +6397,14 @@ dependencies = [
 "synstructure 0.13.2",
 ]

+[[package]]
+name = "mimalloc"
+version = "0.1.52"
+source = "git+https://github.com/xonatius/mimalloc_rust.git?rev=6d4c41bb10c6d9da1d1b6f07b38c4cc051667f11#6d4c41bb10c6d9da1d1b6f07b38c4cc051667f11"
+dependencies = [
+ "libmimalloc-sys",
+]
+
 [[package]]
 name = "mime"
 version = "0.3.17"
@@ -9139,11 +9162,13 @@ dependencies = [
 "insta",
 "jiff",
 "libc",
+ "libmimalloc-sys",
 "libsystemd",
 "matchit 0.9.2",
 "md-5 0.11.0",
 "metrics",
 "metrics-util",
+ "mimalloc",
 "mime_guess",
 "opentelemetry",
 "opentelemetry_sdk",
@@ -9179,8 +9204,6 @@ dependencies = [
 "rustfs-lock",
 "rustfs-log-analyzer",
 "rustfs-madmin",
- "rustfs-mimalloc",
- "rustfs-mimalloc-sys",
 "rustfs-notify",
 "rustfs-object-capacity",
 "rustfs-object-data-cache",
@@ -9852,24 +9875,6 @@ dependencies = [
 "tokio",
 ]

-[[package]]
-name = "rustfs-mimalloc"
-version = "0.5.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a406f4aa07084301d485beec873af6dccc8e3f8762da244743df92038b1db1a6"
-dependencies = [
- "rustfs-mimalloc-sys",
-]
-
-[[package]]
-name = "rustfs-mimalloc-sys"
-version = "0.5.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c3051b819175f58445d4c369a72f0ab88149f3885ba8bea2aff3be01f53fe7cd"
-dependencies = [
- "cc",
-]
-
 [[package]]
 name = "rustfs-notify"
 version = "1.0.0-rc.3"
@@ -12683,7 +12688,6 @@ dependencies = [
 "js-sys",
 "rand 0.10.2",
 "serde_core",
- "sha1_smol",
 "wasm-bindgen",
 ]

@@ -350,8 +350,8 @@ russh-sftp = "2.4.0"
 dav-server = "0.11.0"

 # Performance Analysis and Memory Profiling
-rustfs-mimalloc = { version = "0.5.0" }
-rustfs-mimalloc-sys = { version = "0.5.0" }
+mimalloc = { version = "0.1.52", git = "https://github.com/xonatius/mimalloc_rust.git", rev = "6d4c41bb10c6d9da1d1b6f07b38c4cc051667f11" }
+libmimalloc-sys = { version = "0.1.49", git = "https://github.com/xonatius/mimalloc_rust.git", rev = "6d4c41bb10c6d9da1d1b6f07b38c4cc051667f11", features = ["extended"] }
 hotpath = { version = "0.23.3", default-features = false }
 # Snapshot testing for output format regression detection
 insta = { version = "1.48" }
@@ -72,7 +72,7 @@ The reason string on each attribute is the classifier. Current classes:

 - **Needs a pre-started server** — `"requires running RustFS server at
  localhost:9000"` / `"Connects to existing rustfs server"`. These are the
-  `reliant/*` tests; start a server first (e.g.
+  `reliant/*` and `policy/test_runner` tests; start a server first (e.g.
  [`scripts/run_e2e_tests.sh`](../../scripts/run_e2e_tests.sh)) or use
  `--run-ignored`.
 - **Heavy / external tool** — `"Starts a rustfs server; enable when running
@@ -11,20 +11,29 @@ The tests cover the following AWS policy variable scenarios:
 3. **Variable concatenation** - Combining variables with static text like `prefix-${aws:username}-suffix`
 4. **Nested variables** - Complex nested variable patterns like `${${aws:username}-test}`
 5. **Deny scenarios** - Testing deny policies with variables
-6. **STS credentials** - Variable resolution inherited by temporary credentials

 ## Prerequisites

+- RustFS server binary
 - `awscurl` utility for admin API calls
 - AWS SDK for Rust (included in the project)

 ## Running Tests

+### Run All Policy Tests Using Unified Test Runner
+
+```bash
+# Run all policy tests with comprehensive reporting
+# Note: Requires a RustFS server running on localhost:9000
+cargo test -p e2e_test policy::test_runner::test_policy_full_suite -- --nocapture --ignored  --test-threads=1
+
+# Run only critical policy tests
+cargo test -p e2e_test policy::test_runner::test_policy_critical_suite -- --nocapture --ignored --test-threads=1
+```
+
 ### Run All Policy Tests

 ```bash
 # From the project root directory
-cargo test -p e2e_test policy:: -- --nocapture
+cargo test -p e2e_test policy:: -- --nocapture --ignored --test-threads=1
 ```
-
-Each test starts an isolated RustFS server on a dynamically allocated local port and cleans it up afterward.
@@ -18,3 +18,5 @@
 //! including single-value, multi-value, and nested variable scenarios.

 mod policy_variables_test;
+mod test_env;
+mod test_runner;
@@ -14,17 +14,14 @@

 //! Tests for AWS IAM policy variables with single-value, multi-value, and nested scenarios

-use crate::common::{
-    RustFSTestEnvironment, awscurl_delete, awscurl_put, build_test_s3_config, build_test_sts_client, init_logging,
-};
-use aws_sdk_s3::Client;
-use aws_sdk_s3::error::ProvideErrorMetadata;
+use crate::common::{awscurl_delete, awscurl_put, init_logging};
+use crate::policy::test_env::PolicyTestEnvironment;
 use aws_sdk_s3::primitives::ByteStream;
 use tracing::info;

 /// Helper function to create a regular user with given credentials
 async fn create_user(
-    env: &RustFSTestEnvironment,
+    env: &PolicyTestEnvironment,
    username: &str,
    password: &str,
 ) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
@@ -39,9 +36,20 @@ async fn create_user(
    Ok(())
 }

+/// Helper function to create an STS user with given credentials
+async fn create_sts_user(
+    env: &PolicyTestEnvironment,
+    username: &str,
+    password: &str,
+) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
+    // For STS, we create a regular user first, then use it to assume roles
+    create_user(env, username, password).await?;
+    Ok(())
+}
+
 /// Helper function to create and attach a policy
 async fn create_and_attach_policy(
-    env: &RustFSTestEnvironment,
+    env: &PolicyTestEnvironment,
    policy_name: &str,
    username: &str,
    policy_document: serde_json::Value,
@@ -62,9 +70,9 @@ async fn create_and_attach_policy(
 }

 /// Helper function to clean up test resources
-async fn cleanup_user_and_policy(env: &RustFSTestEnvironment, username: &str, policy_name: &str) {
+async fn cleanup_user_and_policy(env: &PolicyTestEnvironment, username: &str, policy_name: &str) {
    // Create admin client for cleanup
-    let admin_client = env.create_s3_client();
+    let admin_client = env.create_s3_client(&env.access_key, &env.secret_key);

    // Delete buckets that might have been created by this user
    let bucket_patterns = [
@@ -76,7 +84,7 @@ async fn cleanup_user_and_policy(env: &RustFSTestEnvironment, username: &str, po
        format!("{username}-test"),
        format!("{username}-sts-bucket"),
        format!("{username}-service-bucket"),
-        format!("{username}-private-bucket"),
+        "private-test-bucket".to_string(), // For deny test
    ];

    // Try to delete objects and buckets
@@ -113,18 +121,24 @@ async fn cleanup_user_and_policy(env: &RustFSTestEnvironment, username: &str, po

 /// Test AWS policy variables with single-value scenarios
 #[tokio::test(flavor = "multi_thread")]
+#[ignore = "Starts a rustfs server; enable when running full E2E"]
 pub async fn test_aws_policy_variables_single_value() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
+    test_aws_policy_variables_single_value_impl().await
+}
+
+/// Implementation function for single-value policy variables test
+pub async fn test_aws_policy_variables_single_value_impl() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
    info!("Starting AWS policy variables single-value test");

-    let mut env = RustFSTestEnvironment::new().await?;
-    env.start_rustfs_server(vec![]).await?;
+    let env = PolicyTestEnvironment::with_address("127.0.0.1:9000").await?;

    test_aws_policy_variables_single_value_impl_with_env(&env).await
 }

-async fn test_aws_policy_variables_single_value_impl_with_env(
-    env: &RustFSTestEnvironment,
+/// Implementation function for single-value policy variables test with shared environment
+pub async fn test_aws_policy_variables_single_value_impl_with_env(
+    env: &PolicyTestEnvironment,
 ) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    // Create test user
    let test_user = "testuser1";
@@ -184,7 +198,9 @@ async fn test_aws_policy_variables_single_value_impl_with_env(
    awscurl_put(&attach_policy_url, "", &env.access_key, &env.secret_key).await?;

    // Create S3 client for test user
-    let test_client = env.create_s3_client_with_credentials(test_user, test_password);
+    let test_client = env.create_s3_client(test_user, test_password);
+
+    tokio::time::sleep(std::time::Duration::from_millis(500)).await;

    // Test 1: User should be able to list buckets (allowed by policy)
    info!("Test 1: User listing buckets");
@@ -241,13 +257,11 @@ async fn test_aws_policy_variables_single_value_impl_with_env(
    // Test 6: User should NOT be able to create bucket NOT matching username pattern
    info!("Test 6: User attempting to create bucket NOT matching pattern");
    let other_bucket_name = "other-user-bucket";
-    let denied = test_client
-        .create_bucket()
-        .bucket(other_bucket_name)
-        .send()
-        .await
-        .expect_err("a bucket outside the username pattern must be denied");
-    assert_eq!(denied.as_service_error().and_then(ProvideErrorMetadata::code), Some("AccessDenied"));
+    let create_other_result = test_client.create_bucket().bucket(other_bucket_name).send().await;
+    if create_other_result.is_ok() {
+        cleanup().await;
+        return Err("User should NOT be able to create bucket NOT matching username pattern".into());
+    }

    // Cleanup
    info!("Cleaning up test resources");
@@ -259,18 +273,24 @@ async fn test_aws_policy_variables_single_value_impl_with_env(

 /// Test AWS policy variables with multi-value scenarios
 #[tokio::test(flavor = "multi_thread")]
+#[ignore = "Starts a rustfs server; enable when running full E2E"]
 pub async fn test_aws_policy_variables_multi_value() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
+    test_aws_policy_variables_multi_value_impl().await
+}
+
+/// Implementation function for multi-value policy variables test
+pub async fn test_aws_policy_variables_multi_value_impl() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
    info!("Starting AWS policy variables multi-value test");

-    let mut env = RustFSTestEnvironment::new().await?;
-    env.start_rustfs_server(vec![]).await?;
+    let env = PolicyTestEnvironment::with_address("127.0.0.1:9000").await?;

    test_aws_policy_variables_multi_value_impl_with_env(&env).await
 }

-async fn test_aws_policy_variables_multi_value_impl_with_env(
-    env: &RustFSTestEnvironment,
+/// Implementation function for multi-value policy variables test with shared environment
+pub async fn test_aws_policy_variables_multi_value_impl_with_env(
+    env: &PolicyTestEnvironment,
 ) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    // Create test user
    let test_user = "testuser2";
@@ -318,7 +338,7 @@ async fn test_aws_policy_variables_multi_value_impl_with_env(
    create_and_attach_policy(env, policy_name, test_user, policy_document).await?;

    // Create S3 client for test user
-    let test_client = env.create_s3_client_with_credentials(test_user, test_password);
+    let test_client = env.create_s3_client(test_user, test_password);

    // Test 1: User should be able to create buckets matching any of the multi-value patterns
    info!("Test 1: User creating first bucket matching multi-value pattern");
@@ -348,13 +368,11 @@ async fn test_aws_policy_variables_multi_value_impl_with_env(
    // Test 4: User should NOT be able to create bucket NOT matching any multi-value pattern
    info!("Test 4: User attempting to create bucket NOT matching any pattern");
    let other_bucket_name = format!("{test_user}-other-bucket");
-    let denied = test_client
-        .create_bucket()
-        .bucket(&other_bucket_name)
-        .send()
-        .await
-        .expect_err("a bucket outside all allowed patterns must be denied");
-    assert_eq!(denied.as_service_error().and_then(ProvideErrorMetadata::code), Some("AccessDenied"));
+    let create_other_result = test_client.create_bucket().bucket(&other_bucket_name).send().await;
+    if create_other_result.is_ok() {
+        cleanup().await;
+        return Err("User should NOT be able to create bucket NOT matching any multi-value pattern".into());
+    }

    // Test 5: User should be able to list objects in their allowed buckets
    info!("Test 5: User listing objects in allowed buckets");
@@ -380,18 +398,24 @@ async fn test_aws_policy_variables_multi_value_impl_with_env(

 /// Test AWS policy variables with variable concatenation
 #[tokio::test(flavor = "multi_thread")]
+#[ignore = "Starts a rustfs server; enable when running full E2E"]
 pub async fn test_aws_policy_variables_concatenation() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
+    test_aws_policy_variables_concatenation_impl().await
+}
+
+/// Implementation function for concatenation policy variables test
+pub async fn test_aws_policy_variables_concatenation_impl() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
    info!("Starting AWS policy variables concatenation test");

-    let mut env = RustFSTestEnvironment::new().await?;
-    env.start_rustfs_server(vec![]).await?;
+    let env = PolicyTestEnvironment::with_address("127.0.0.1:9000").await?;

    test_aws_policy_variables_concatenation_impl_with_env(&env).await
 }

-async fn test_aws_policy_variables_concatenation_impl_with_env(
-    env: &RustFSTestEnvironment,
+/// Implementation function for concatenation policy variables test with shared environment
+pub async fn test_aws_policy_variables_concatenation_impl_with_env(
+    env: &PolicyTestEnvironment,
 ) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    // Create test user
    let test_user = "testuser3";
@@ -431,7 +455,10 @@ async fn test_aws_policy_variables_concatenation_impl_with_env(
    create_and_attach_policy(env, policy_name, test_user, policy_document).await?;

    // Create S3 client for test user
-    let test_client = env.create_s3_client_with_credentials(test_user, test_password);
+    let test_client = env.create_s3_client(test_user, test_password);
+
+    // Add a small delay to allow policy to propagate
+    tokio::time::sleep(std::time::Duration::from_millis(500)).await;

    // Test: User should be able to create bucket matching concatenated pattern
    info!("Test: User creating bucket matching concatenated pattern");
@@ -460,30 +487,41 @@ async fn test_aws_policy_variables_concatenation_impl_with_env(

 /// Test AWS policy variables with nested scenarios
 #[tokio::test(flavor = "multi_thread")]
+#[ignore = "Starts a rustfs server; enable when running full E2E"]
 pub async fn test_aws_policy_variables_nested() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
+    test_aws_policy_variables_nested_impl().await
+}
+
+/// Implementation function for nested policy variables test
+pub async fn test_aws_policy_variables_nested_impl() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
    info!("Starting AWS policy variables nested test");

-    let mut env = RustFSTestEnvironment::new().await?;
-    env.start_rustfs_server(vec![]).await?;
+    let env = PolicyTestEnvironment::with_address("127.0.0.1:9000").await?;

    test_aws_policy_variables_nested_impl_with_env(&env).await
 }

 /// Test AWS policy variables with STS temporary credentials
 #[tokio::test(flavor = "multi_thread")]
+#[ignore = "Starts a rustfs server; enable when running full E2E"]
 pub async fn test_aws_policy_variables_sts() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
+    test_aws_policy_variables_sts_impl().await
+}
+
+/// Implementation function for STS policy variables test
+pub async fn test_aws_policy_variables_sts_impl() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
    info!("Starting AWS policy variables STS test");

-    let mut env = RustFSTestEnvironment::new().await?;
-    env.start_rustfs_server(vec![]).await?;
+    let env = PolicyTestEnvironment::with_address("127.0.0.1:9000").await?;

    test_aws_policy_variables_sts_impl_with_env(&env).await
 }

-async fn test_aws_policy_variables_nested_impl_with_env(
-    env: &RustFSTestEnvironment,
+/// Implementation function for nested policy variables test with shared environment
+pub async fn test_aws_policy_variables_nested_impl_with_env(
+    env: &PolicyTestEnvironment,
 ) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    // Create test user
    let test_user = "testuser4";
@@ -523,7 +561,10 @@ async fn test_aws_policy_variables_nested_impl_with_env(
    create_and_attach_policy(env, policy_name, test_user, policy_document).await?;

    // Create S3 client for test user
-    let test_client = env.create_s3_client_with_credentials(test_user, test_password);
+    let test_client = env.create_s3_client(test_user, test_password);
+
+    // Add a small delay to allow policy to propagate
+    tokio::time::sleep(std::time::Duration::from_millis(500)).await;

    // Test nested variable resolution
    info!("Test: Nested variable resolution");
@@ -540,14 +581,14 @@ async fn test_aws_policy_variables_nested_impl_with_env(
        return Err(format!("User should be able to create bucket with nested variable: {e}").into());
    }

-    // Verify a valid bucket name outside the resolved resource is denied.
-    let denied = test_client
-        .create_bucket()
-        .bucket("other-user-test")
-        .send()
-        .await
-        .expect_err("a bucket outside the resolved nested variable must be denied");
-    assert_eq!(denied.as_service_error().and_then(ProvideErrorMetadata::code), Some("AccessDenied"));
+    // Verify bucket creation fails with unresolved variable
+    let unresolved_bucket = format!("${{}}-test {test_user}");
+    let create_unresolved = test_client.create_bucket().bucket(&unresolved_bucket).send().await;
+
+    if create_unresolved.is_ok() {
+        cleanup().await;
+        return Err("User should NOT be able to create bucket with unresolved variable".into());
+    }

    // Cleanup
    info!("Cleaning up test resources");
@@ -557,8 +598,9 @@ async fn test_aws_policy_variables_nested_impl_with_env(
    Ok(())
 }

-async fn test_aws_policy_variables_sts_impl_with_env(
-    env: &RustFSTestEnvironment,
+/// Implementation function for STS policy variables test with shared environment
+pub async fn test_aws_policy_variables_sts_impl_with_env(
+    env: &PolicyTestEnvironment,
 ) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    // Create test user for STS
    let test_user = "testuser-sts";
@@ -570,7 +612,8 @@ async fn test_aws_policy_variables_sts_impl_with_env(
        cleanup_user_and_policy(env, test_user, policy_name).await;
    };

-    create_user(env, test_user, test_password).await?;
+    // Create STS user
+    create_sts_user(env, test_user, test_password).await?;

    // Create policy with STS-compatible variables
    let policy_document = serde_json::json!({
@@ -581,11 +624,6 @@ async fn test_aws_policy_variables_sts_impl_with_env(
                "Action": ["s3:ListAllMyBuckets"],
                "Resource": ["arn:aws:s3:::*"]
            },
-            {
-                "Effect": "Allow",
-                "Action": ["sts:AssumeRole"],
-                "Resource": ["arn:aws:s3:::*"]
-            },
            {
                "Effect": "Allow",
                "Action": ["s3:CreateBucket"],
@@ -593,12 +631,7 @@ async fn test_aws_policy_variables_sts_impl_with_env(
            },
            {
                "Effect": "Allow",
-                "Action": ["s3:ListBucket"],
-                "Resource": [format!("arn:aws:s3:::{}-sts-bucket", "${aws:username}")]
-            },
-            {
-                "Effect": "Allow",
-                "Action": ["s3:PutObject", "s3:GetObject"],
+                "Action": ["s3:ListBucket", "s3:PutObject", "s3:GetObject"],
                "Resource": [format!("arn:aws:s3:::{}-sts-bucket/*", "${aws:username}")]
            }
        ]
@@ -606,22 +639,11 @@ async fn test_aws_policy_variables_sts_impl_with_env(

    create_and_attach_policy(env, policy_name, test_user, policy_document).await?;

-    let assumed = build_test_sts_client(&env.url, test_user, test_password, None, "policy-variable-sts")
-        .assume_role()
-        .role_arn("arn:aws:iam::123456789012:role/policy-variable")
-        .role_session_name("policy-variable-e2e")
-        .send()
-        .await?;
-    let credentials = assumed
-        .credentials()
-        .ok_or("AssumeRole response should contain temporary credentials")?;
-    let test_client = Client::from_conf(build_test_s3_config(
-        &env.url,
-        credentials.access_key_id(),
-        credentials.secret_access_key(),
-        Some(credentials.session_token()),
-        "policy-variable-sts-session",
-    ));
+    // Create S3 client for test user
+    let test_client = env.create_s3_client(test_user, test_password);
+
+    // Add a small delay to allow policy to propagate
+    tokio::time::sleep(std::time::Duration::from_millis(500)).await;

    // Test: User should be able to create bucket matching STS pattern
    info!("Test: User creating bucket matching STS pattern");
@@ -677,18 +699,24 @@ async fn test_aws_policy_variables_sts_impl_with_env(

 /// Test AWS policy variables with deny scenarios
 #[tokio::test(flavor = "multi_thread")]
+#[ignore = "Starts a rustfs server; enable when running full E2E"]
 pub async fn test_aws_policy_variables_deny() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
+    test_aws_policy_variables_deny_impl().await
+}
+
+/// Implementation function for deny policy variables test
+pub async fn test_aws_policy_variables_deny_impl() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
    info!("Starting AWS policy variables deny test");

-    let mut env = RustFSTestEnvironment::new().await?;
-    env.start_rustfs_server(vec![]).await?;
+    let env = PolicyTestEnvironment::with_address("127.0.0.1:9000").await?;

    test_aws_policy_variables_deny_impl_with_env(&env).await
 }

-async fn test_aws_policy_variables_deny_impl_with_env(
-    env: &RustFSTestEnvironment,
+/// Implementation function for deny policy variables test with shared environment
+pub async fn test_aws_policy_variables_deny_impl_with_env(
+    env: &PolicyTestEnvironment,
 ) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    // Create test user
    let test_user = "testuser5";
@@ -731,7 +759,10 @@ async fn test_aws_policy_variables_deny_impl_with_env(
    create_and_attach_policy(env, policy_name, test_user, policy_document).await?;

    // Create S3 client for test user
-    let test_client = env.create_s3_client_with_credentials(test_user, test_password);
+    let test_client = env.create_s3_client(test_user, test_password);
+
+    // Add a small delay to allow policy to propagate
+    tokio::time::sleep(std::time::Duration::from_millis(500)).await;

    // Test 1: User should be able to create bucket matching username pattern
    info!("Test 1: User creating bucket matching username pattern");
@@ -744,14 +775,12 @@ async fn test_aws_policy_variables_deny_impl_with_env(

    // Test 2: User should NOT be able to create bucket with "private" in the name (deny rule)
    info!("Test 2: User attempting to create bucket with 'private' in name (should be denied)");
-    let private_bucket_name = format!("{test_user}-private-bucket");
-    let denied = test_client
-        .create_bucket()
-        .bucket(&private_bucket_name)
-        .send()
-        .await
-        .expect_err("the explicit deny must reject a matching bucket name");
-    assert_eq!(denied.as_service_error().and_then(ProvideErrorMetadata::code), Some("AccessDenied"));
+    let private_bucket_name = "private-test-bucket";
+    let create_private_result = test_client.create_bucket().bucket(private_bucket_name).send().await;
+    if create_private_result.is_ok() {
+        cleanup().await;
+        return Err("User should NOT be able to create bucket with 'private' in name due to deny rule".into());
+    }

    // Cleanup
    info!("Cleaning up test resources");
@@ -0,0 +1,100 @@
+// 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.
+
+//! Custom test environment for policy variables tests
+//!
+//! This module provides a custom test environment that doesn't automatically
+//! stop servers when destroyed, addressing the server stopping issue.
+
+use aws_sdk_s3::Client;
+use aws_sdk_s3::config::{Config, Credentials, Region};
+use std::net::TcpStream;
+use std::time::Duration;
+use tokio::time::sleep;
+use tracing::{info, warn};
+
+// Default credentials
+const DEFAULT_ACCESS_KEY: &str = "rustfsadmin";
+const DEFAULT_SECRET_KEY: &str = "rustfsadmin";
+
+/// Custom test environment that doesn't automatically stop servers
+pub struct PolicyTestEnvironment {
+    pub temp_dir: String,
+    pub address: String,
+    pub url: String,
+    pub access_key: String,
+    pub secret_key: String,
+}
+
+impl PolicyTestEnvironment {
+    /// Create a new test environment with specific address
+    /// This environment won't stop any server when dropped
+    pub async fn with_address(address: &str) -> Result<Self, Box<dyn std::error::Error + Send + Sync>> {
+        let temp_dir = format!("/tmp/rustfs_policy_test_{}", uuid::Uuid::new_v4());
+        tokio::fs::create_dir_all(&temp_dir).await?;
+
+        let url = format!("http://{address}");
+
+        Ok(Self {
+            temp_dir,
+            address: address.to_string(),
+            url,
+            access_key: DEFAULT_ACCESS_KEY.to_string(),
+            secret_key: DEFAULT_SECRET_KEY.to_string(),
+        })
+    }
+
+    /// Create an AWS S3 client configured for this RustFS instance
+    pub fn create_s3_client(&self, access_key: &str, secret_key: &str) -> Client {
+        let credentials = Credentials::new(access_key, secret_key, None, None, "policy-test");
+        let config = Config::builder()
+            .credentials_provider(credentials)
+            .region(Region::new("us-east-1"))
+            .endpoint_url(&self.url)
+            .force_path_style(true)
+            .behavior_version_latest()
+            .build();
+        Client::from_conf(config)
+    }
+
+    /// Wait for RustFS server to be ready by checking TCP connectivity
+    pub async fn wait_for_server_ready(&self) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
+        info!("Waiting for RustFS server to be ready on {}", self.address);
+
+        for i in 0..30 {
+            if TcpStream::connect(&self.address).is_ok() {
+                info!("✅ RustFS server is ready after {} attempts", i + 1);
+                return Ok(());
+            }
+
+            if i == 29 {
+                return Err("RustFS server failed to become ready within 30 seconds".into());
+            }
+
+            sleep(Duration::from_secs(1)).await;
+        }
+
+        Ok(())
+    }
+}
+
+// Implement Drop trait that doesn't stop servers
+impl Drop for PolicyTestEnvironment {
+    fn drop(&mut self) {
+        // Clean up temp directory only, don't stop any server
+        if let Err(e) = std::fs::remove_dir_all(&self.temp_dir) {
+            warn!("Failed to clean up temp directory {}: {}", self.temp_dir, e);
+        }
+    }
+}
@@ -0,0 +1,230 @@
+// Copyright 2024 RustFS Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use crate::common::init_logging;
+use crate::policy::test_env::PolicyTestEnvironment;
+use std::time::Instant;
+use tokio::time::{Duration, sleep};
+use tracing::{error, info};
+
+/// Test case definition
+#[derive(Debug, Clone)]
+pub struct TestDefinition {
+    pub name: String,
+    pub is_critical: bool,
+}
+
+impl TestDefinition {
+    pub fn new(name: impl Into<String>, is_critical: bool) -> Self {
+        Self {
+            name: name.into(),
+            is_critical,
+        }
+    }
+}
+
+/// Test result
+#[derive(Debug, Clone)]
+pub struct TestResult {
+    pub test_name: String,
+    pub success: bool,
+    pub error_message: Option<String>,
+}
+
+impl TestResult {
+    pub fn success(test_name: String) -> Self {
+        Self {
+            test_name,
+            success: true,
+            error_message: None,
+        }
+    }
+
+    pub fn failure(test_name: String, error: String) -> Self {
+        Self {
+            test_name,
+            success: false,
+            error_message: Some(error),
+        }
+    }
+}
+
+/// Test suite configuration
+#[derive(Debug, Clone, Default)]
+pub struct TestSuiteConfig {
+    pub include_critical_only: bool,
+}
+
+/// Policy test suite
+pub struct PolicyTestSuite {
+    tests: Vec<TestDefinition>,
+    config: TestSuiteConfig,
+}
+
+impl PolicyTestSuite {
+    /// Create default test suite
+    pub fn new() -> Self {
+        let tests = vec![
+            TestDefinition::new("test_aws_policy_variables_single_value", true),
+            TestDefinition::new("test_aws_policy_variables_multi_value", true),
+            TestDefinition::new("test_aws_policy_variables_concatenation", true),
+            TestDefinition::new("test_aws_policy_variables_nested", true),
+            TestDefinition::new("test_aws_policy_variables_deny", true),
+            TestDefinition::new("test_aws_policy_variables_sts", true),
+        ];
+
+        Self {
+            tests,
+            config: TestSuiteConfig::default(),
+        }
+    }
+
+    /// Configure test suite
+    pub fn with_config(mut self, config: TestSuiteConfig) -> Self {
+        self.config = config;
+        self
+    }
+
+    /// Run test suite
+    pub async fn run_test_suite(&self) -> Vec<TestResult> {
+        init_logging();
+        info!("Starting Policy Variables test suite");
+
+        let start_time = Instant::now();
+        let mut results = Vec::new();
+
+        // Create test environment
+        let env = match PolicyTestEnvironment::with_address("127.0.0.1:9000").await {
+            Ok(env) => env,
+            Err(e) => {
+                error!("Failed to create test environment: {}", e);
+                return vec![TestResult::failure("env_creation".into(), e.to_string())];
+            }
+        };
+
+        // Wait for server to be ready
+        if env.wait_for_server_ready().await.is_err() {
+            error!("Server is not ready");
+            return vec![TestResult::failure("server_check".into(), "Server not ready".into())];
+        }
+
+        // Filter tests
+        let tests_to_run: Vec<&TestDefinition> = self
+            .tests
+            .iter()
+            .filter(|test| !self.config.include_critical_only || test.is_critical)
+            .collect();
+
+        info!("Scheduled {} tests", tests_to_run.len());
+
+        // Run tests
+        for (i, test_def) in tests_to_run.iter().enumerate() {
+            info!("Running test {}/{}: {}", i + 1, tests_to_run.len(), test_def.name);
+            let test_start = Instant::now();
+
+            let result = self.run_single_test(test_def, &env).await;
+            let test_duration = test_start.elapsed();
+
+            match result {
+                Ok(_) => {
+                    info!("Test passed: {} ({:.2}s)", test_def.name, test_duration.as_secs_f64());
+                    results.push(TestResult::success(test_def.name.clone()));
+                }
+                Err(e) => {
+                    error!("Test failed: {} ({:.2}s): {}", test_def.name, test_duration.as_secs_f64(), e);
+                    results.push(TestResult::failure(test_def.name.clone(), e.to_string()));
+                }
+            }
+
+            // Delay between tests to avoid resource conflicts
+            if i < tests_to_run.len() - 1 {
+                sleep(Duration::from_secs(2)).await;
+            }
+        }
+
+        // Print summary
+        self.print_summary(&results, start_time.elapsed());
+
+        results
+    }
+
+    /// Run a single test
+    async fn run_single_test(
+        &self,
+        test_def: &TestDefinition,
+        env: &PolicyTestEnvironment,
+    ) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
+        match test_def.name.as_str() {
+            "test_aws_policy_variables_single_value" => {
+                super::policy_variables_test::test_aws_policy_variables_single_value_impl_with_env(env).await
+            }
+            "test_aws_policy_variables_multi_value" => {
+                super::policy_variables_test::test_aws_policy_variables_multi_value_impl_with_env(env).await
+            }
+            "test_aws_policy_variables_concatenation" => {
+                super::policy_variables_test::test_aws_policy_variables_concatenation_impl_with_env(env).await
+            }
+            "test_aws_policy_variables_nested" => {
+                super::policy_variables_test::test_aws_policy_variables_nested_impl_with_env(env).await
+            }
+            "test_aws_policy_variables_deny" => {
+                super::policy_variables_test::test_aws_policy_variables_deny_impl_with_env(env).await
+            }
+            "test_aws_policy_variables_sts" => {
+                super::policy_variables_test::test_aws_policy_variables_sts_impl_with_env(env).await
+            }
+            _ => Err(format!("Test {} not implemented", test_def.name).into()),
+        }
+    }
+
+    /// Print test summary
+    fn print_summary(&self, results: &[TestResult], total_duration: Duration) {
+        info!("=== Test Suite Summary ===");
+        info!("Total duration: {:.2}s", total_duration.as_secs_f64());
+        info!("Total tests: {}", results.len());
+
+        let passed = results.iter().filter(|r| r.success).count();
+        let failed = results.len() - passed;
+        let success_rate = (passed as f64 / results.len() as f64) * 100.0;
+
+        info!("Passed: {} | Failed: {}", passed, failed);
+        info!("Success rate: {:.1}%", success_rate);
+
+        if failed > 0 {
+            error!("Failed tests:");
+            for result in results.iter().filter(|r| !r.success) {
+                error!("  - {}: {}", result.test_name, result.error_message.as_ref().unwrap());
+            }
+        }
+    }
+}
+
+/// Test suite
+#[tokio::test]
+#[ignore = "Connects to existing rustfs server"]
+async fn test_policy_critical_suite() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
+    let config = TestSuiteConfig {
+        include_critical_only: true,
+    };
+    let suite = PolicyTestSuite::new().with_config(config);
+    let results = suite.run_test_suite().await;
+
+    let failed = results.iter().filter(|r| !r.success).count();
+    if failed > 0 {
+        return Err(format!("Critical tests failed: {failed} failures").into());
+    }
+
+    info!("All critical tests passed");
+    Ok(())
+}
@@ -2026,7 +2026,7 @@ impl PoolMeta {
        self.load_no_lock(pool).await
    }

-    pub(crate) async fn load_no_lock<S>(&mut self, pool: Arc<S>) -> Result<()>
+    async fn load_no_lock<S>(&mut self, pool: Arc<S>) -> Result<()>
    where
        S: EcstoreObjectIO,
    {
@@ -988,11 +988,14 @@ impl crate::storage_api_contracts::multipart::MultipartOperations for Sets {
    }
 }

-impl Sets {
-    pub(crate) async fn heal_format_with_fence<F>(&self, dry_run: bool, fence_lost: F) -> Result<(HealResultItem, Option<Error>)>
-    where
-        F: Fn() -> bool + Send + Sync,
-    {
+#[async_trait::async_trait]
+impl crate::storage_api_contracts::heal::HealOperations for Sets {
+    type Error = Error;
+    type HealResultItem = HealResultItem;
+    type HealOptions = HealOpts;
+
+    #[tracing::instrument(skip(self))]
+    async fn heal_format(&self, dry_run: bool) -> Result<(HealResultItem, Option<Error>)> {
        let (disks, init_errs) = init_storage_disks_with_errors(
            &self.endpoints.endpoints,
            &DiskOption {
@@ -1065,9 +1068,6 @@ impl Sets {
            // Save new formats `format.json` on unformatted disks.
            for (index, (fm, disk)) in tmp_new_formats.iter_mut().zip(disks.iter()).enumerate() {
                if fm.is_some() && disk.is_some() {
-                    if fence_lost() {
-                        return Ok((res, Some(StorageError::SlowDown)));
-                    }
                    if let Err(err) = save_format_file(disk, fm).await {
                        if let Some(disk) = disk.as_ref() {
                            let _ = disk.close().await;
@@ -1101,18 +1101,6 @@ impl Sets {
        }
        Ok((res, None))
    }
-}
-
-#[async_trait::async_trait]
-impl crate::storage_api_contracts::heal::HealOperations for Sets {
-    type Error = Error;
-    type HealResultItem = HealResultItem;
-    type HealOptions = HealOpts;
-
-    #[tracing::instrument(skip(self))]
-    async fn heal_format(&self, dry_run: bool) -> Result<(HealResultItem, Option<Error>)> {
-        self.heal_format_with_fence(dry_run, || false).await
-    }
    #[tracing::instrument(skip(self))]
    async fn heal_bucket(&self, bucket: &str, opts: &HealOpts) -> Result<HealResultItem> {
        let mut result = HealResultItem {
@@ -2124,13 +2124,26 @@ impl SetDisks {

            let put_object_size = known_put_object_storage_size(data.size());
            let shard_file_size_raw = erasure.shard_file_size(put_object_size);
-            let is_inline_buffer = storage_class_config.should_inline(shard_file_size_raw, erasure.data_shards, opts.versioned);
+            let is_inline_buffer =
+                storage_class_config.should_inline(shard_file_size_raw, erasure.data_shards, opts.versioned);

            let collect_stage_timing = rustfs_io_metrics::put_stage_metrics_enabled() || issue3031_diag_enabled();
            let shard_file_size = shard_file_size_raw;
            let shard_size = erasure.shard_size();
            let write_path = classify_put_write_path(is_inline_buffer, put_object_size, fi.erasure.block_size);
            let direct_inline_commit = matches!(write_path, SmallWritePath::Inline);
+            {
+                use std::io::Write;
+                let msg = format!(
+                    "INLINE_DEBUG: bucket={} obj={} size={} shard_fs={} ds={} bs={} inline={} direct={} path={} iblock={} ver={}\n",
+                    bucket, object, put_object_size, shard_file_size_raw, erasure.data_shards, fi.erasure.block_size,
+                    is_inline_buffer, direct_inline_commit, write_path.metric_label(), storage_class_config.inline_block(), opts.versioned
+                );
+                if let Ok(mut f) = std::fs::OpenOptions::new().create(true).append(true).open("/tmp/rustfs_inline_debug.log") {
+                    let _ = f.write_all(msg.as_bytes());
+                }
+                let _ = std::io::stderr().write_all(msg.as_bytes());
+            }
            rustfs_io_metrics::record_put_object_path(write_path.metric_label());
            let writer_setup_stage_start = collect_stage_timing.then(Instant::now);
            let (mut writers, errors) = if direct_inline_commit {
@@ -13,12 +13,7 @@
 // limitations under the License.

 use super::*;
-use crate::core::pools::POOL_META_NAME;
-use crate::services::rebalance::{REBAL_META_NAME, RebalStatus};
-use crate::set_disk::get_lock_acquire_timeout;
 use crate::storage_api_contracts::heal::HealOperations as _;
-use crate::storage_api_contracts::namespace::NamespaceLocking as _;
-use rustfs_lock::NamespaceLockGuard;
 use tracing::trace;

 const LOG_COMPONENT_ECSTORE: &str = "ecstore";
@@ -35,119 +30,7 @@ fn invalid_heal_pool_index(pool_idx: usize, pool_count: usize) -> Error {
    )
 }

-#[derive(Debug, Clone, Copy)]
-enum HealFormatPoolSkip {
-    Completed,
-    Retryable,
-}
-
-fn classify_heal_format_pool(
-    pool_idx: usize,
-    pool_cmd_line: &str,
-    pool_meta: &PoolMeta,
-    rebalance_meta: Option<&RebalanceMeta>,
-) -> Option<HealFormatPoolSkip> {
-    let Some(pool) = pool_meta.pools.get(pool_idx) else {
-        return Some(HealFormatPoolSkip::Retryable);
-    };
-
-    if pool.id != pool_idx || pool_cmd_line.is_empty() || pool.cmd_line.is_empty() || pool.cmd_line != pool_cmd_line {
-        return Some(HealFormatPoolSkip::Retryable);
-    }
-
-    if let Some(decommission) = pool.decommission.as_ref() {
-        if decommission.complete {
-            return Some(HealFormatPoolSkip::Completed);
-        }
-        if decommission.failed || decommission.canceled || decommission.queued || pool_meta.is_suspended(pool_idx) {
-            return Some(HealFormatPoolSkip::Retryable);
-        }
-    }
-
-    if let Some(meta) = rebalance_meta {
-        let Some(pool_stats) = meta.pool_stats.get(pool_idx) else {
-            return Some(HealFormatPoolSkip::Retryable);
-        };
-        if pool_stats.info.stopping || (pool_stats.participating && pool_stats.info.status == RebalStatus::Started) {
-            return Some(HealFormatPoolSkip::Retryable);
-        }
-    }
-
-    None
-}
-
-fn heal_format_pool_skip_error(skip: HealFormatPoolSkip) -> Error {
-    match skip {
-        HealFormatPoolSkip::Completed => StorageError::NoHealRequired,
-        HealFormatPoolSkip::Retryable => StorageError::SlowDown,
-    }
-}
-
-fn heal_format_fence_lost_error() -> Error {
-    StorageError::SlowDown
-}
-
 impl ECStore {
-    async fn acquire_heal_format_fence(
-        &self,
-    ) -> Result<(NamespaceLockGuard, NamespaceLockGuard, PoolMeta, Option<RebalanceMeta>)> {
-        let metadata_pool = self
-            .pools
-            .first()
-            .cloned()
-            .ok_or_else(|| Error::other("heal format requires at least one storage pool"))?;
-
-        // Metadata fence order is part of the decommission/rebalance protocol:
-        // pool.bin must always be acquired before rebalance.bin.
-        let pool_lock = metadata_pool.new_ns_lock(RUSTFS_META_BUCKET, POOL_META_NAME).await?;
-        let pool_guard = pool_lock.get_write_lock(get_lock_acquire_timeout()).await?;
-        let rebalance_lock = metadata_pool.new_ns_lock(RUSTFS_META_BUCKET, REBAL_META_NAME).await?;
-        let rebalance_guard = rebalance_lock.get_write_lock(get_lock_acquire_timeout()).await?;
-
-        if pool_guard.is_lock_lost() || rebalance_guard.is_lock_lost() {
-            return Err(heal_format_fence_lost_error());
-        }
-
-        let mut pool_meta = PoolMeta::default();
-        pool_meta.load_no_lock(metadata_pool.clone()).await?;
-        if pool_meta.pools.len() != self.pools.len()
-            || pool_meta.pools.iter().enumerate().any(|(pool_idx, pool)| {
-                pool.id != pool_idx || pool.cmd_line.is_empty() || pool.cmd_line != self.pools[pool_idx].endpoints.cmd_line
-            })
-        {
-            return Err(heal_format_fence_lost_error());
-        }
-
-        let mut rebalance_meta = RebalanceMeta::new();
-        let rebalance_meta = match rebalance_meta
-            .load_with_opts(
-                metadata_pool,
-                ObjectOptions {
-                    no_lock: true,
-                    ..Default::default()
-                },
-            )
-            .await
-        {
-            Ok(()) => Some(rebalance_meta),
-            Err(Error::ConfigNotFound) => None,
-            Err(err) => return Err(err),
-        };
-
-        if rebalance_meta
-            .as_ref()
-            .is_some_and(|meta| meta.pool_stats.len() != self.pools.len())
-        {
-            return Err(heal_format_fence_lost_error());
-        }
-
-        if pool_guard.is_lock_lost() || rebalance_guard.is_lock_lost() {
-            return Err(heal_format_fence_lost_error());
-        }
-
-        Ok((pool_guard, rebalance_guard, pool_meta, rebalance_meta))
-    }
-
    fn get_pools_for_heal_object(&self, opts: &HealOpts) -> Result<Vec<Arc<Sets>>> {
        match opts.pool {
            Some(pool_idx) => Ok(vec![
@@ -169,26 +52,9 @@ impl ECStore {
        };

        let mut count_no_heal = 0;
-        let mut count_completed = 0;
        let mut first_error = None;
-        for (pool_idx, pool) in self.pools.iter().enumerate() {
-            let (pool_guard, rebalance_guard, pool_meta, rebalance_meta) = self.acquire_heal_format_fence().await?;
-            if pool_guard.is_lock_lost() || rebalance_guard.is_lock_lost() {
-                first_error.get_or_insert(heal_format_fence_lost_error());
-                break;
-            }
-            if let Some(skip) = classify_heal_format_pool(pool_idx, &pool.endpoints.cmd_line, &pool_meta, rebalance_meta.as_ref())
-            {
-                if matches!(skip, HealFormatPoolSkip::Completed) {
-                    count_completed += 1;
-                } else {
-                    first_error.get_or_insert(heal_format_pool_skip_error(skip));
-                }
-                continue;
-            }
-
-            let fence_lost = || pool_guard.is_lock_lost() || rebalance_guard.is_lock_lost();
-            let (mut result, err) = pool.heal_format_with_fence(dry_run, fence_lost).await?;
+        for pool in self.pools.iter() {
+            let (mut result, err) = pool.heal_format(dry_run).await?;
            if let Some(err) = err {
                match err {
                    StorageError::NoHealRequired => {
@@ -203,18 +69,11 @@ impl ECStore {
            r.set_count += result.set_count;
            r.before.drives.append(&mut result.before.drives);
            r.after.drives.append(&mut result.after.drives);
-
-            // A lease can be lost after the final write; fail closed before
-            // reporting the pool as successfully healed.
-            if pool_guard.is_lock_lost() || rebalance_guard.is_lock_lost() {
-                first_error.get_or_insert(heal_format_fence_lost_error());
-                break;
-            }
        }
        if let Some(err) = first_error {
            return Ok((r, Some(err)));
        }
-        if count_no_heal + count_completed == self.pools.len() {
+        if count_no_heal == self.pools.len() {
            info!(
                event = EVENT_HEAL_FORMAT_COMPLETED,
                component = LOG_COMPONENT_ECSTORE,
@@ -443,7 +302,6 @@ mod tests {
    use crate::disk::{DeleteOptions, DiskOption, format::FormatV3, new_disk};
    use crate::layout::endpoints::{EndpointServerPools, Endpoints, PoolEndpoints};
    use crate::runtime::instance::InstanceContext;
-    use crate::services::rebalance::{RebalanceInfo, RebalanceStats};
    use crate::storage_api_contracts::bucket::{BucketOperations, MakeBucketOptions};
    use crate::storage_api_contracts::object::{ObjectIO as _, ObjectOperations};
    use crate::store::init_format::{load_format_erasure, save_format_file};
@@ -495,164 +353,6 @@ mod tests {
        }
    }

-    fn pool_meta_with_decommission(info: PoolDecommissionInfo) -> PoolMeta {
-        PoolMeta {
-            pools: vec![PoolStatus {
-                id: 0,
-                cmd_line: "pool-0".to_string(),
-                last_update: OffsetDateTime::UNIX_EPOCH,
-                decommission: Some(info),
-            }],
-            ..Default::default()
-        }
-    }
-
-    #[test]
-    fn heal_format_pool_state_barriers_are_classified() {
-        let active = pool_meta_with_decommission(PoolDecommissionInfo {
-            start_time: Some(OffsetDateTime::UNIX_EPOCH),
-            ..Default::default()
-        });
-        assert!(matches!(
-            classify_heal_format_pool(0, "pool-0", &active, None),
-            Some(HealFormatPoolSkip::Retryable)
-        ));
-
-        for info in [
-            PoolDecommissionInfo {
-                failed: true,
-                ..Default::default()
-            },
-            PoolDecommissionInfo {
-                canceled: true,
-                ..Default::default()
-            },
-        ] {
-            assert!(matches!(
-                classify_heal_format_pool(0, "pool-0", &pool_meta_with_decommission(info), None),
-                Some(HealFormatPoolSkip::Retryable)
-            ));
-        }
-
-        let completed = pool_meta_with_decommission(PoolDecommissionInfo {
-            complete: true,
-            ..Default::default()
-        });
-        assert!(matches!(
-            classify_heal_format_pool(0, "pool-0", &completed, None),
-            Some(HealFormatPoolSkip::Completed)
-        ));
-    }
-
-    #[test]
-    fn heal_format_pool_rebalance_barriers_and_identity_are_fail_closed() {
-        let identity_meta = pool_meta_with_decommission(PoolDecommissionInfo::default());
-        let rebalance = RebalanceMeta {
-            pool_stats: vec![RebalanceStats {
-                participating: true,
-                info: RebalanceInfo {
-                    status: RebalStatus::Started,
-                    ..Default::default()
-                },
-                ..Default::default()
-            }],
-            ..Default::default()
-        };
-        assert!(matches!(
-            classify_heal_format_pool(0, "pool-0", &identity_meta, Some(&rebalance)),
-            Some(HealFormatPoolSkip::Retryable)
-        ));
-
-        let stopping = RebalanceMeta {
-            pool_stats: vec![RebalanceStats {
-                info: RebalanceInfo {
-                    stopping: true,
-                    ..Default::default()
-                },
-                ..Default::default()
-            }],
-            ..Default::default()
-        };
-        assert!(matches!(
-            classify_heal_format_pool(0, "pool-0", &identity_meta, Some(&stopping)),
-            Some(HealFormatPoolSkip::Retryable)
-        ));
-
-        let identity = pool_meta_with_decommission(PoolDecommissionInfo::default());
-        assert!(matches!(
-            classify_heal_format_pool(0, "pool-new", &identity, None),
-            Some(HealFormatPoolSkip::Retryable)
-        ));
-
-        let identity_without_decommission = PoolMeta {
-            pools: vec![PoolStatus {
-                id: 0,
-                cmd_line: "pool-0".to_string(),
-                last_update: OffsetDateTime::UNIX_EPOCH,
-                decommission: None,
-            }],
-            ..Default::default()
-        };
-        assert!(matches!(
-            classify_heal_format_pool(0, "pool-new", &identity_without_decommission, None),
-            Some(HealFormatPoolSkip::Retryable)
-        ));
-
-        assert!(matches!(
-            classify_heal_format_pool(0, "", &identity_meta, None),
-            Some(HealFormatPoolSkip::Retryable)
-        ));
-
-        assert!(matches!(
-            classify_heal_format_pool(0, "pool-0", &PoolMeta::default(), None),
-            Some(HealFormatPoolSkip::Retryable)
-        ));
-
-        let stopped = RebalanceMeta {
-            stopped_at: Some(OffsetDateTime::UNIX_EPOCH),
-            pool_stats: vec![RebalanceStats {
-                participating: true,
-                info: RebalanceInfo {
-                    status: RebalStatus::Stopped,
-                    ..Default::default()
-                },
-                ..Default::default()
-            }],
-            ..Default::default()
-        };
-        assert!(classify_heal_format_pool(0, "pool-0", &identity_meta, Some(&stopped)).is_none());
-
-        let stopping_after_stop = RebalanceMeta {
-            stopped_at: Some(OffsetDateTime::UNIX_EPOCH),
-            pool_stats: vec![RebalanceStats {
-                participating: true,
-                info: RebalanceInfo {
-                    status: RebalStatus::Started,
-                    stopping: true,
-                    ..Default::default()
-                },
-                ..Default::default()
-            }],
-            ..Default::default()
-        };
-        assert!(matches!(
-            classify_heal_format_pool(0, "pool-0", &identity_meta, Some(&stopping_after_stop)),
-            Some(HealFormatPoolSkip::Retryable)
-        ));
-    }
-
-    #[test]
-    fn skipped_heal_format_pool_is_never_reported_as_success() {
-        assert!(matches!(
-            heal_format_pool_skip_error(HealFormatPoolSkip::Retryable),
-            StorageError::SlowDown
-        ));
-        assert!(matches!(
-            heal_format_pool_skip_error(HealFormatPoolSkip::Completed),
-            StorageError::NoHealRequired
-        ));
-    }
-
    async fn multi_pool_heal_store() -> (tempfile::TempDir, Arc<ECStore>, CancellationToken) {
        let temp_dir = tempfile::tempdir().expect("multi-pool heal test directory should be created");
        let mut pool_endpoints = Vec::new();
@@ -1189,18 +889,6 @@ mod tests {
            bucket_fence_registry: std::sync::Arc::default(),
        };

-        let err = store
-            .handle_heal_format(false)
-            .await
-            .expect_err("missing pool metadata must fail closed before format writes");
-        assert!(matches!(err, StorageError::SlowDown));
-
-        let pool_meta = PoolMeta::new(&store.pools, &PoolMeta::default());
-        pool_meta
-            .save(store.pools.clone())
-            .await
-            .expect("pool metadata should be persisted before format heal");
-
        let (result, err) = store
            .handle_heal_format(false)
            .await
@@ -1214,22 +902,5 @@ mod tests {
            .await
            .expect("the later pool should be healed despite the first pool error");
        assert_eq!(healed.erasure.this, recoverable_format.erasure.sets[0][2]);
-
-        let mut completed_meta = PoolMeta::new(&store.pools, &PoolMeta::default());
-        for status in &mut completed_meta.pools {
-            status.decommission = Some(PoolDecommissionInfo {
-                complete: true,
-                ..Default::default()
-            });
-        }
-        completed_meta
-            .save(store.pools.clone())
-            .await
-            .expect("completed pool metadata should be persisted");
-        let (_, err) = store
-            .handle_heal_format(false)
-            .await
-            .expect("completed pools should be reported as a no-op");
-        assert!(matches!(err, Some(StorageError::NoHealRequired)));
    }
 }
@@ -3194,7 +3194,7 @@ impl ECStore {

        // Default return value
        let mut del_objects = vec![DeletedObject::default(); objects.len()];
-        let accounting = vec![None; objects.len()];
+        let mut accounting = vec![None; objects.len()];

        let mut del_errs = Vec::with_capacity(objects.len());
        for _ in 0..objects.len() {
@@ -271,7 +271,7 @@ pub(super) fn resolve_latest_object_info_candidates(
            .filter(|candidate| latest_candidate_mod_time(candidate) == Some(latest_mod_time))
            .collect::<Vec<_>>();

-        latest_candidates.sort_by_key(|candidate| std::cmp::Reverse(candidate.idx));
+        latest_candidates.sort_by(|left, right| right.idx.cmp(&left.idx));

        let Some(winner) = latest_candidates.first() else {
            return Err(Error::ErasureReadQuorum);
@@ -231,10 +231,6 @@ impl HealTask {
                            "Heal erasure set format repair skipped because no format heal was required"
                        );
                    } else {
-                        let error = e;
-                        if error.is_recoverable_heal() {
-                            return Err(error);
-                        }
                        error!(
                            target: "rustfs::heal::task",
                            event = EVENT_HEAL_ERASURE_SET_RESULT,
@@ -243,7 +239,7 @@ impl HealTask {
                            task_id = %self.id,
                            set_disk_id,
                            result = "format_failed",
-                            error = %error,
+                            error = %e,
                            "Heal erasure set failed"
                        );
                        {
@@ -251,7 +247,7 @@ impl HealTask {
                            progress.update_progress(4, 4, 0, 0);
                        }
                        return Err(Error::TaskExecutionFailed {
-                            message: format!("Failed to heal disk format for {set_disk_id}: {error}"),
+                            message: format!("Failed to heal disk format for {set_disk_id}: {e}"),
                        });
                    }
                } else {
@@ -288,9 +284,6 @@ impl HealTask {
            Err(Error::TaskCancelled) => return Err(Error::TaskCancelled),
            Err(Error::TaskTimeout) => return Err(Error::TaskTimeout),
            Err(e) => {
-                if e.is_recoverable_heal() {
-                    return Err(e);
-                }
                error!(
                    target: "rustfs::heal::task",
                    event = EVENT_HEAL_ERASURE_SET_RESULT,
@@ -547,7 +547,6 @@ struct MockStorage {
    heal_object_outcome: Mutex<Option<MockHealObjectOutcome>>,
    heal_object_outcomes: Mutex<HashMap<String, VecDeque<MockHealObjectOutcome>>>,
    format_no_heal_required: Mutex<bool>,
-    format_error: Mutex<Option<Error>>,
    global_format_calls: Mutex<u32>,
    replacement_format_calls: Mutex<Vec<(usize, usize, Vec<String>)>>,
    replacement_targets_ready: Mutex<bool>,
@@ -868,9 +867,6 @@ impl HealStorageAPI for MockStorage {

    async fn heal_format(&self, _dry_run: bool) -> Result<(HealResultItem, Option<Error>)> {
        *self.global_format_calls.lock().unwrap() += 1;
-        if let Some(error) = self.format_error.lock().unwrap().take() {
-            return Err(error);
-        }
        let no_heal_required = *self.format_no_heal_required.lock().unwrap();
        if no_heal_required {
            Ok((HealResultItem::default(), Some(Error::Storage(EcstoreError::NoHealRequired))))
@@ -2056,30 +2052,6 @@ async fn test_erasure_set_heal_continues_after_format_no_heal_required() {
    );
 }

-#[tokio::test]
-async fn erasure_set_format_slowdown_is_propagated() {
-    let storage = Arc::new(MockStorage {
-        format_error: Mutex::new(Some(Error::Storage(EcstoreError::SlowDown))),
-        ..Default::default()
-    });
-    let request = HealRequest::new(
-        HealType::ErasureSet {
-            buckets: Vec::new(),
-            set_disk_id: "pool_0_set_0".to_string(),
-        },
-        HealOptions::default(),
-        HealPriority::Normal,
-    );
-    let task = HealTask::from_request(request, storage);
-
-    let error = task
-        .execute()
-        .await
-        .expect_err("format SlowDown must remain recoverable for the task manager");
-
-    assert!(matches!(error, Error::Storage(EcstoreError::SlowDown)));
-}
-
 #[tokio::test]
 async fn erasure_set_bucket_prepass_failure_stops_before_object_heal() {
    let temp = TempDir::new().expect("temporary directory should be created");
@@ -245,18 +245,6 @@ impl TestECStoreEnvBuilder {
            .await
            .expect("build test ECStore");

-        // The production bootstrap only persists pool.bin from the elected
-        // first cluster node.  Test stores intentionally have no cluster
-        // election, but heal-format still requires that durable fence before
-        // it can write any disk format.  Materialize the validated topology
-        // here so the shared fixture models a ready single-node store.
-        let mut pool_meta = ecstore.pool_meta.read().await.clone();
-        pool_meta.dont_save = false;
-        pool_meta
-            .save(ecstore.pools.clone())
-            .await
-            .expect("persist test pool metadata");
-
        if self.init_bucket_metadata {
            let buckets_list = ecstore
                .list_bucket(&BucketOptions {
@@ -43,6 +43,9 @@ allow-git = [
    # RustFS fork carrying presigned expiry and constant-time authentication fixes.
    # owner: rustfs-maintainers  review: 2026-10
    "https://github.com/rustfs/s3s.git",
+    # MiMalloc fork pinned for hotpath allocation counting support.
+    # owner: houseme  review: 2026-10
+    "https://github.com/xonatius/mimalloc_rust.git",
 ]

 [bans]
@@ -1,149 +0,0 @@
-# CI gate matrix
-
-This file is the source of truth for which validation runs on each event, its
-configured wall-clock budget, and whether it can block a merge. Test taxonomy,
-naming, and nextest serialization rules remain in [README.md](README.md); e2e
-membership and counts remain in
-[e2e-suite-inventory.md](e2e-suite-inventory.md).
-
-The distinction between **required** and **report-only** is load-bearing:
-a failing job blocks a merge only when its exact check name is present in the
-live `main` ruleset. A workflow name, a `merge_group` trigger, or a red PR check
-does not make a job required by itself.
-
-## Required merge checks
-
-The live `main` ruleset (`6436880`) currently requires exactly these contexts:
-
-| Required context | Producer | Validation |
-|---|---|---|
-| `CLA Check` | `.github/workflows/cla.yml` | Contributor agreement |
-| `Quick Checks` | `.github/workflows/ci.yml` | Formatting and repository guard scripts |
-| `Test and Lint` | `.github/workflows/ci.yml` | Clippy, workspace nextest excluding `e2e_test`, doctests, and migration proofs |
-
-For pull requests limited to the paths excluded by the main CI workflow,
-`.github/workflows/ci-docs-only.yml` reports `Quick Checks` and
-`Test and Lint` under the same names. It runs the real quick checks and the
-planning-document guard; it does not claim that Rust compilation or runtime
-tests ran. Despite the workflow name, these paths also include selected deploy,
-workflow, and lock files.
-
-Verify the live rule rather than trusting this snapshot before changing merge
-policy:
-
-```bash
-gh api repos/rustfs/rustfs/rulesets/6436880 \
-  --jq '.rules[] | select(.type == "required_status_checks") | .parameters'
-```
-
-The ruleset currently has `strict_required_status_checks_policy=false`.
-`Continuous Integration` accepts `merge_group` events and runs `e2e-full` for
-them, but `End-to-End Tests (full merge gate)` is not currently a required
-context. Therefore the repository is prepared to test a merge-queue SHA, but
-the workflow alone does not prove that every merge passed that lane.
-
-## Pull request and merge matrix
-
-Budgets below are job `timeout-minutes`, not typical runtimes. “Report-only”
-means the result is visible and actionable but is not in the live required
-context list.
-
-| Event | Validation | Budget | Merge status | Reproduction |
-|---|---|---:|---|---|
-| PR, non-doc change | `Quick Checks` | 10 min | Required | `make pre-commit` (broader local umbrella) |
-| PR, non-doc change | `Test and Lint` | 90 min | Required | `cargo nextest run --profile ci --all --exclude e2e_test` |
-| PR, non-doc change | `Typos` | 10 min | Report-only | `typos` |
-| PR, non-doc change | `ILM Integration (serial)` | 90 min | Report-only | Use the exact command in `.github/workflows/ci.yml` |
-| PR, non-doc change | rio-v2 / swift / sftp test-and-lint variants | 90 min each | Report-only | `cargo nextest run` with the workflow's feature set |
-| PR, non-doc change | `Build RustFS Debug Binary` | 30 min | Report-only; prerequisite for black-box lanes | `cargo build -p rustfs --bins` |
-| PR, non-doc change | `io_uring Integration (real)` | 30 min | Report-only | `cargo test -p rustfs-ecstore --lib uring_ -- --test-threads=1 --nocapture` |
-| PR, non-doc change | `End-to-End Tests` (`e2e-smoke` plus `s3s-e2e`) | 30 min | Report-only | `cargo nextest run --profile e2e-smoke -p e2e_test`; then `./scripts/e2e-run.sh ./target/debug/rustfs <data-dir>` |
-| PR, non-doc change | `S3 Implemented Tests` | 60 min | Report-only | Build `rustfs`, then run `scripts/s3-tests/run.sh` with `DEPLOY_MODE=binary`, `TEST_MODE=single`, and `MAXFAIL=0` |
-| PR, non-doc change | `S3 Lifecycle Behavior Tests` | 30 min | Report-only | Use the accelerated scanner environment in `.github/workflows/ci.yml` with `scripts/s3-tests/run.sh` |
-| PR touching dependency or workflow inputs | Cargo Deny / Workflow Pin Report / Dependency Review | 20 / 5 / 30 min | Report-only | `cargo deny check`; `scripts/security/check_workflow_pins.sh` |
-| PR touching architecture rules or architecture docs | `Architecture Migration Rules` | 10 min | Report-only | `scripts/check_architecture_migration_rules.sh` |
-| PR touching Nix or workspace manifests | `Nix Build & Check` | 60 min | Report-only | `nix flake check` |
-| PR limited to main-CI-excluded paths | companion `Quick Checks` and `Test and Lint` | 10 min each | Required | `git diff --check`; `make doc-paths-check` when documentation paths changed |
-| `merge_group` | Standard CI plus `e2e-full` | 55 min for `e2e-full` | Standard required contexts only; `e2e-full` report-only | `cargo nextest run --profile e2e-full -p e2e_test` |
-| Push to `main` | Standard CI plus `e2e-full` | 55 min for `e2e-full` | Post-merge detection | Same as `merge_group` |
-| PR touching fuzz inputs or harness paths | Build plus five 60-second fuzz smoke targets | 60 min build; 30 min per target | Report-only | `MAX_TOTAL_TIME=60 ./scripts/fuzz/run.sh` |
-| PR touching selected ecstore disk/format paths | `Rename Safety` on Windows | 60 min | Report-only | Run the four `cargo test -p rustfs-ecstore --lib <filter>` commands in `windows-filesystem.yml` on Windows |
-
-The authoritative e2e filters live in `.config/nextest.toml`; extend a profile
-instead of adding a second ad-hoc selector. Before a profile runs,
-`scripts/check_test_wiring.py` compares its exact membership to the committed
-digest so a silent test drop fails closed.
-
-## Scheduled and manual validation
-
-Scheduled lanes are independent fault domains. They do not block a pull
-request, but their workflow-local gate can fail the run and scheduled failures
-are routed to the shared failure-issue action. The scheduled-validation
-watchdog and freshness workflow separately detect incomplete runs and missing
-schedules.
-
-| Cadence (UTC unless noted) | Workflow / validation | Budget | Verdict and artifacts | Reproduction |
-|---|---|---:|---|---|
-| Daily 02:17 | Fuzz: five nightly corpus targets | 60 min build; 60 min per target | Gate; corpus/crash artifacts, scheduled failure alert | `MAX_TOTAL_TIME=<seconds> ./scripts/fuzz/run.sh` |
-| Daily 03:17 | MinIO interop (EC + SSE read parity) | 40 min | Gate; scheduled failure alert | Dispatch `minio-interop.yml` or follow its pinned Docker fixture steps |
-| Daily 04:29 | Replication / cluster-fault / protocol e2e | 45 / 90 / 90 min | Three independent gates; JUnit, membership, and server logs | `cargo nextest run --profile e2e-repl-nightly -p e2e_test`; `--profile e2e-nightly`; `-j 1 --profile e2e-protocols` |
-| Daily 06:31 | Warp performance A/B | 180 min | Regression budget gate; A/B summaries and server logs | `bash scripts/run_hotpath_warp_abba.sh --help` |
-| Daily 00:07 Asia/Shanghai (16:07 UTC previous day) | Nightly GNU build and Vault lanes | 150 / 90 / 60 min | Build, live Vault, and HA failover gates | Use the commands and pinned Vault images in `nightly-gnu.yml` |
-| Daily 03:23 | Security Audit | 20 / 5 min, plus 30 min on PR dependency review | Cargo Deny and workflow-pin gates; scheduled failure alert | `cargo deny check`; `scripts/security/check_workflow_pins.sh` |
-| Daily 23:47 | Scheduled Validation Freshness | 10 min | Fails when a critical schedule was never created or is stale | Dispatch `scheduled-validation-freshness.yml` |
-| Sunday 00:11 | Full `Continuous Integration` matrix | Per-job budgets above | Weekly variant coverage, including dormant rio-v2 binary/e2e lanes | Dispatch `ci.yml` |
-| Sunday 01:13 | Seven-platform build matrix | 150 min per platform | Build/package integrity; scheduled failure alert | Dispatch `build.yml` with an exact platform set |
-| Sunday 02:19 | Ceph s3-tests full sweep: single and real four-node, four shards each | 180 min per shard | Compatibility gate; report, JUnit, exact node IDs, and server logs | `scripts/s3-tests/run.sh` against an existing single or distributed target |
-| Sunday 06:41 | Mint | 120 min | **Report-only by design**; per-suite PASS/FAIL/NA and raw `log.json` | Reproduce the pinned Docker sequence in `mint.yml` or dispatch it |
-| Sunday 07:43 | Workspace line coverage | 120 min | Report-only trend; lcov and JSON retained 90 days | `make coverage` |
-| Monthly, day 1 06:37 | Runner Hygiene | 15 min | Validates runner ephemerality; scheduled failure alert | Dispatch `runner-hygiene.yml` |
-
-Manual `workflow_dispatch` exists for the scheduled workflows above. Manual
-runs are debugging evidence and intentionally do not open scheduled-failure
-issues. A manual performance run may explicitly allow a known regression; that
-override must not be treated as an ordinary passing baseline.
-
-## Release validation
-
-Release validation is post-merge and tag-driven; it does not substitute for a
-pull-request gate.
-
-| Event | Validation | Budget | Result |
-|---|---|---:|---|
-| Push to `main` or weekly schedule | `Build and Release` platform matrix | 150 min per platform | Build artifacts for all selected targets; no release publication on a main push |
-| Valid release or preview tag | `Build and Release` plus asset checks | 150 min per platform | Draft release, checksummed assets, and publish step |
-| Successful non-preview release-tag build | Docker image build and image scan | 60 min build; 30 min scan | Multi-architecture images plus vulnerability report |
-| Successful release-tag build | DEB/RPM packaging | 30 min per architecture | Packages and checksum files uploaded to the release |
-| Successful non-preview release-tag build | Helm template test and package | 30 min build; 30 min publish | Versioned chart and repository index |
-
-Use an exact preview tag for end-to-end release rehearsal. Manual dispatches
-are backfill/debug paths and do not prove the automatic `workflow_run` chain.
-
-## Evidence requirements
-
-A green check is useful only when it proves the intended behavior ran:
-
- Record the exact commit SHA and run URL.
- Separate product failure from runner prerequisites, service readiness, and
-  cancellation. Repair the precondition, then rerun the exact workload.
- Preserve membership manifests, JUnit, raw compatibility logs, seeds, and
-  server logs where the workflow provides them.
- For a bug fix or a new fault checker, provide sensitivity evidence: the old
-  behavior or an intentional mutation must fail the new oracle, and the fixed
-  behavior must pass it.
- Never promote a report-only lane to required from one green run. Require at
-  least 14 days and 30 representative pull requests with at least 99% complete
-  execution, then update the ruleset and this table together.
-
-## Change checklist
-
-Update this file in the same pull request when any of these change:
-
- workflow triggers, job names, timeouts, or nextest profile ownership;
- required status contexts or strict/merge-queue policy;
- scheduled cadence, alert routing, artifact contract, or local reproduction;
- report-only versus gating semantics.
-
-Do not copy per-module test counts here. Update
-[e2e-suite-inventory.md](e2e-suite-inventory.md) and its enforced membership
-digest instead.
@@ -80,7 +80,6 @@
 | object_lambda_test | 16 | 🌙 |
 | object_lock | 34 |  |
 | overwrite_cleanup_regression_test | 1 |  |
-| policy | 6 |  |
 | presigned_negative_test | 7 | ✅ |
 | protocols | 16 | 🌙 |
 | quota_test | 14 |  |
@@ -100,4 +99,4 @@
 | tls_hot_reload_test | 1 | ✅ |
 | version_id_regression_test | 10 | ✅ |

-**Total listed: 581 tests across 83 modules · PR smoke: 163 tests / 36 modules · merge/main full: 459 tests / 74 modules · nightly replication: 55 tests · nightly cluster faults: 28 tests / 7 modules · nightly protocols: 16 tests** · updated 2026-08-23.
+**Total listed: 575 tests across 82 modules · PR smoke: 163 tests / 36 modules · merge/main full: 453 tests / 73 modules · nightly replication: 55 tests · nightly cluster faults: 28 tests / 7 modules · nightly protocols: 16 tests** · updated 2026-08-23.
@@ -322,7 +322,7 @@ thiserror = { workspace = true }
 tracing.workspace = true
 url = { workspace = true }
 urlencoding = { workspace = true }
-uuid = { workspace = true, features = ["v4", "v5", "fast-rng", "macro-diagnostics"] }
+uuid = { workspace = true, features = ["v4", "fast-rng", "macro-diagnostics"] }
 zip = { workspace = true }
 libc = { workspace = true }
 rand = { workspace = true, features = ["serde"] }
@@ -336,16 +336,16 @@ opentelemetry = { workspace = true }
 tracing-opentelemetry = { workspace = true }
 # Data structures
 hashbrown = { workspace = true, features = ["serde", "rayon"] }
-rustfs-mimalloc = { workspace = true }
+mimalloc = { workspace = true }

 [target.'cfg(target_os = "linux")'.dependencies]
 libsystemd.workspace = true

 [target.'cfg(not(target_os = "windows"))'.dependencies]
-rustfs-mimalloc-sys.workspace = true
+libmimalloc-sys.workspace = true

 [dev-dependencies]
-uuid = { workspace = true, features = ["v4", "v5", "fast-rng", "macro-diagnostics"] }
+uuid = { workspace = true, features = ["v4", "fast-rng", "macro-diagnostics"] }
 serial_test = { workspace = true }
 tempfile = { workspace = true }
 aws-config = { workspace = true }
@@ -41,7 +41,7 @@ use crate::admin::storage_api::config::save_admin_config;
 use crate::admin::storage_api::contract::bucket::{
    BucketOperations, BucketOptions, DeleteBucketOptions, MakeBucketOptions, SRBucketDeleteOp,
 };
-use crate::admin::storage_api::error::{Error as StorageError, is_err_bucket_not_found};
+use crate::admin::storage_api::error::Error as StorageError;
 use crate::admin::storage_api::runtime::ECStore;
 use crate::admin::utils::{encode_compatible_admin_payload, read_compatible_admin_body};
 use crate::auth::constant_time_eq;
@@ -55,7 +55,6 @@ use crate::storage::storage_api::{
 use base64::Engine;
 use base64::engine::general_purpose::STANDARD as BASE64_STANDARD;
 use base64::engine::general_purpose::URL_SAFE_NO_PAD;
-use futures::StreamExt;
 use hmac::{Hmac, Mac};
 use http::header::{CONTENT_TYPE, HOST};
 use http::{HeaderMap, HeaderValue, Uri};
@@ -2097,18 +2096,6 @@ async fn remote_add_preflight_info(site: &PeerSite) -> S3Result<SiteReplicationA
            format!("invalid site replication metainfo from `{}`: {e}", site.endpoint),
        )
    })?;
-    if info.deployment_id.is_empty() {
-        // The peer will be tracked under a locally derived fallback ID
-        // (deployment_id_for_endpoint) instead of its real deployment ID.
-        warn!(
-            event = EVENT_ADMIN_SITE_REPLICATION_STATE,
-            component = LOG_COMPONENT_ADMIN,
-            subsystem = LOG_SUBSYSTEM_SITE_REPLICATION,
-            result = "peer_deployment_id_missing",
-            peer_endpoint = %site.endpoint,
-            "admin site replication state"
-        );
-    }

    let idp_body = send_peer_admin_get_request_with_client(
        &client,
@@ -2219,30 +2206,20 @@ fn site_replication_bootstrap_token(uri: &Uri) -> Option<String> {
    query_pairs(uri).get("bootstrapToken").cloned()
 }

-/// Query for a peer `make-with-versioning` bucket op. `versioningEnabled`
-/// always travels so the outbound query matches MinIO's site-replication
-/// make-bucket wire contract: MinIO's own create-bucket hook sends
-/// `versioningEnabled=true` on this op. RustFS's inbound handler
-/// force-enables versioning either way.
-fn make_with_versioning_bucket_op_path(bucket: &str, created_at: Option<&str>, lock_enabled: bool) -> String {
+fn bootstrap_bucket_make_op_path(bucket: &SRBucketInfo) -> String {
    let mut query = form_urlencoded::Serializer::new(String::new());
-    query.append_pair("bucket", bucket);
-    query.append_pair("operation", SITE_REPLICATION_BUCKET_OP_MAKE_WITH_VERSIONING);
-    query.append_pair("versioningEnabled", "true");
-    if let Some(created_at) = created_at {
-        query.append_pair("createdAt", created_at);
+    query.append_pair("bucket", &bucket.bucket);
+    query.append_pair("operation", "make-with-versioning");
+    if let Some(created_at) = bucket
+        .created_at
+        .and_then(|value| value.format(&time::format_description::well_known::Rfc3339).ok())
+    {
+        query.append_pair("createdAt", &created_at);
    }
-    if lock_enabled {
+    if bucket.object_lock_config.is_some() {
        query.append_pair("lockEnabled", "true");
    }
-    format!("{SITE_REPLICATION_PEER_BUCKET_OPS_PATH}?{}", query.finish())
-}
-
-fn bootstrap_bucket_make_op_path(bucket: &SRBucketInfo) -> String {
-    let created_at = bucket
-        .created_at
-        .and_then(|value| value.format(&time::format_description::well_known::Rfc3339).ok());
-    make_with_versioning_bucket_op_path(&bucket.bucket, created_at.as_deref(), bucket.object_lock_config.is_some())
+    format!("/rustfs/admin/v3/site-replication/peer/bucket-ops?{}", query.finish())
 }

 fn bootstrap_bucket_meta_item(bucket: &SRBucketInfo, item_type: &str, updated_at: Option<OffsetDateTime>) -> SRBucketMeta {
@@ -4269,7 +4246,16 @@ async fn broadcast_site_replication_make_bucket(
        .format(&time::format_description::well_known::Rfc3339)
        .unwrap_or_default();

-    let path = make_with_versioning_bucket_op_path(bucket, Some(&created_at), lock_enabled);
+    let path = {
+        let mut query = form_urlencoded::Serializer::new(String::new());
+        query.append_pair("bucket", bucket);
+        query.append_pair("operation", "make-with-versioning");
+        query.append_pair("createdAt", &created_at);
+        if lock_enabled {
+            query.append_pair("lockEnabled", "true");
+        }
+        format!("/rustfs/admin/v3/site-replication/peer/bucket-ops?{}", query.finish())
+    };
    let path = if let Some(token) = bootstrap_token {
        with_site_replication_bootstrap_token(&path, token)
    } else {
@@ -10220,25 +10206,13 @@ impl Operation for SiteReplicationStatusHandler {
    }
 }

-/// `POST /v3/site-replication/devnull` — peer link-check upload drain.
-/// MinIO streams multi-megabyte probe bodies here during site netperf link
-/// checks and expects an unbounded discard (its handler copies to io.Discard);
-/// buffering through the 1MB admin body cap turned any larger probe into a
-/// 400 and a false link failure. Stream and discard instead — no size cap.
-async fn drain_site_replication_devnull(mut input: Body) -> S3Result<()> {
-    while let Some(chunk) = input.next().await {
-        chunk.map_err(|e| s3_error!(InvalidRequest, "failed to read devnull stream: {}", e))?;
-    }
-    Ok(())
-}
-
 pub struct SiteReplicationDevNullHandler {}

 #[async_trait::async_trait]
 impl Operation for SiteReplicationDevNullHandler {
    async fn call(&self, req: S3Request<Body>, _params: Params<'_, '_>) -> S3Result<S3Response<(StatusCode, Body)>> {
        validate_site_replication_admin_request(&req, AdminAction::SiteReplicationOperationAction).await?;
-        drain_site_replication_devnull(req.input).await?;
+        let _ = read_plain_admin_body(req.input).await?;
        Ok(empty_response(StatusCode::NO_CONTENT))
    }
 }
@@ -10497,19 +10471,6 @@ impl Operation for SRPeerJoinHandler {
    }
 }

-/// Outcome of a peer-driven `purge-deleted-bucket` replay. A bucket that is
-/// already gone means the purge raced an earlier replay or a local delete —
-/// that is success — but any other failure must reach the sender like the
-/// sibling delete branches do: swallowing it answered 200 while the bucket
-/// survived on this site.
-fn purge_deleted_bucket_result(result: Result<(), StorageError>) -> S3Result<()> {
-    match result {
-        Ok(()) => Ok(()),
-        Err(err) if is_err_bucket_not_found(&err) => Ok(()),
-        Err(err) => Err(ApiError::from(err).into()),
-    }
-}
-
 pub struct SRPeerBucketOpsHandler {}

 #[async_trait::async_trait]
@@ -10609,8 +10570,7 @@ impl Operation for SRPeerBucketOpsHandler {
                    .map_err(ApiError::from)?;
            }
            "purge-deleted-bucket" => {
-                purge_deleted_bucket_result(
-                    store
+                let _ = store
                    .delete_bucket(
                        &bucket,
                        &DeleteBucketOptions {
@@ -10619,8 +10579,7 @@ impl Operation for SRPeerBucketOpsHandler {
                            ..Default::default()
                        },
                    )
-                        .await,
-                )?;
+                    .await;
            }
            _ => return Err(s3_error!(InvalidRequest, "unsupported site replication bucket operation")),
        }
@@ -13966,54 +13925,6 @@ mod tests {
        assert!(!query_flag(&uri, "missing"));
    }

-    /// A5 red-light: a `purge-deleted-bucket` replay must report success when
-    /// the bucket is already gone, and must propagate every other failure —
-    /// the swallowed error answered 200 while the bucket survived.
-    #[test]
-    fn test_purge_deleted_bucket_result_tolerates_only_missing_bucket() {
-        assert!(purge_deleted_bucket_result(Ok(())).is_ok());
-        assert!(purge_deleted_bucket_result(Err(StorageError::BucketNotFound("photos".to_string()))).is_ok());
-        assert!(purge_deleted_bucket_result(Err(StorageError::VolumeNotFound)).is_ok());
-        let err = purge_deleted_bucket_result(Err(StorageError::StorageFull))
-            .expect_err("non-not-found delete failures must propagate");
-        assert_ne!(*err.code(), S3ErrorCode::NoSuchBucket);
-    }
-
-    /// C5 red-light: the site-replication devnull drain must accept bodies
-    /// beyond the 1MB admin body cap — MinIO's link check streams large
-    /// probe bodies and treats a 400 as a broken link.
-    #[tokio::test]
-    async fn test_site_replication_devnull_drains_body_beyond_admin_cap() {
-        let body = Body::from(vec![0u8; MAX_ADMIN_REQUEST_BODY_SIZE + 1]);
-        drain_site_replication_devnull(body)
-            .await
-            .expect("devnull must drain bodies larger than the admin body cap");
-    }
-
-    /// A3 red-light: `versioningEnabled` must travel on every outbound
-    /// make-with-versioning bucket op so the query matches MinIO's
-    /// site-replication make-bucket wire contract (MinIO's own hook sends
-    /// `versioningEnabled=true` on this op).
-    #[test]
-    fn test_make_with_versioning_op_paths_send_versioning_enabled() {
-        let bucket = SRBucketInfo {
-            bucket: "photos".to_string(),
-            created_at: Some(OffsetDateTime::UNIX_EPOCH),
-            object_lock_config: Some(BASE64_STANDARD.encode("<ObjectLockConfiguration/>")),
-            ..Default::default()
-        };
-        let bootstrap = bootstrap_bucket_make_op_path(&bucket);
-        assert!(bootstrap.contains("operation=make-with-versioning"), "{bootstrap}");
-        assert!(bootstrap.contains("versioningEnabled=true"), "{bootstrap}");
-        assert!(bootstrap.contains("createdAt="), "{bootstrap}");
-        assert!(bootstrap.contains("lockEnabled=true"), "{bootstrap}");
-
-        // The broadcast path (create-bucket hook) shares the same builder.
-        let broadcast = make_with_versioning_bucket_op_path("photos", Some("1970-01-01T00:00:00Z"), false);
-        assert!(broadcast.contains("versioningEnabled=true"), "{broadcast}");
-        assert!(!broadcast.contains("lockEnabled"), "{broadcast}");
-    }
-
    #[tokio::test]
    #[serial]
    async fn test_add_bootstrap_scope_only_allows_expected_bucket_setup_until_guard_drops() {
@@ -13,9 +13,9 @@
 // limitations under the License.

 use rustfs_madmin::{PeerInfo, SyncStatus};
-use std::collections::BTreeMap;
+use std::collections::{BTreeMap, hash_map::DefaultHasher};
+use std::hash::{Hash, Hasher};
 use url::Url;
-use uuid::Uuid;

 fn has_http_scheme(endpoint: &str) -> bool {
    endpoint.get(..7).is_some_and(|prefix| prefix.eq_ignore_ascii_case("http://"))
@@ -66,12 +66,10 @@ pub fn site_identity_key(endpoint: &str) -> String {
        .unwrap_or_else(|| trimmed.to_ascii_lowercase())
 }

-/// Fallback deployment ID for a peer that reported none. UUIDv5 over the
-/// canonical endpoint: the ID is persisted in site-replication state and
-/// broadcast to peers, so it must be identical across Rust toolchains
-/// (`DefaultHasher` is not) and across spellings of the same endpoint.
 pub fn deployment_id_for_endpoint(endpoint: &str) -> String {
-    Uuid::new_v5(&Uuid::NAMESPACE_URL, canonical_endpoint(endpoint).as_bytes()).to_string()
+    let mut hasher = DefaultHasher::new();
+    endpoint.hash(&mut hasher);
+    format!("{:016x}", hasher.finish())
 }

 pub fn same_identity_endpoint(left: &str, right: &str) -> bool {
@@ -176,23 +174,6 @@ mod tests {
        }
    }

-    /// B8 red-light: the fallback deployment ID must be a toolchain-stable
-    /// UUIDv5 over the canonical endpoint — `DefaultHasher` output is not
-    /// guaranteed stable across Rust releases, yet the ID is persisted in
-    /// site-replication state and broadcast to peers.
-    #[test]
-    fn deployment_id_for_endpoint_is_stable_uuid_v5_over_canonical_endpoint() {
-        let endpoint = "https://node-a.example.com:9000";
-        let id = deployment_id_for_endpoint(endpoint);
-        let parsed = uuid::Uuid::parse_str(&id).expect("fallback deployment ID must be a UUID");
-        assert_eq!(parsed.get_version_num(), 5, "fallback deployment ID must be UUIDv5");
-        // Deterministic for the same endpoint and for spelling variants that
-        // share a canonical form; distinct endpoints stay distinct.
-        assert_eq!(id, deployment_id_for_endpoint(endpoint));
-        assert_eq!(id, deployment_id_for_endpoint(" HTTPS://Node-A.Example.Com:9000/ "));
-        assert_ne!(id, deployment_id_for_endpoint("https://node-b.example.com:9000"));
-    }
-
    #[test]
    fn canonical_endpoint_accepts_case_insensitive_scheme() {
        assert_eq!(
@@ -51,7 +51,7 @@ mod ecstore_disk {
 }

 mod ecstore_error {
-    pub(crate) use crate::storage::storage_api::ecstore_error::{StorageError, is_err_bucket_not_found};
+    pub(crate) use crate::storage::storage_api::ecstore_error::StorageError;
 }

 #[allow(unused_imports)]
@@ -919,7 +919,6 @@ pub(crate) mod contract {
 }

 pub(crate) mod error {
-    pub(crate) use super::ecstore_error::is_err_bucket_not_found;
    pub(crate) use super::{Error, StorageError};
 }

@@ -369,8 +369,14 @@ pub fn allocator_reclaim_controller_snapshot(ctx: &CancellationToken) -> Allocat
 }

 #[cfg(not(target_os = "windows"))]
+#[allow(unsafe_code)]
 fn collect_allocator_memory(force: bool) -> Result<(), String> {
-    rustfs_mimalloc::MiMalloc::collect(force);
+    // SAFETY: `mi_collect` is provided by the active global allocator backend
+    // on this target family. It is explicitly intended to reclaim retained
+    // pages/segments and does not require additional invariants from the caller.
+    unsafe {
+        libmimalloc_sys::mi_collect(force);
+    }
    Ok(())
 }

@@ -26,22 +26,22 @@ struct MiMallocAllocator;
 unsafe impl GlobalAlloc for MiMallocAllocator {
    unsafe fn alloc(&self, layout: Layout) -> *mut u8 {
        // SAFETY: the caller upholds GlobalAlloc's contract for layout.
-        unsafe { rustfs_mimalloc::MiMalloc.alloc(layout) }
+        unsafe { mimalloc::MiMalloc.alloc(layout) }
    }

    unsafe fn alloc_zeroed(&self, layout: Layout) -> *mut u8 {
        // SAFETY: the caller upholds GlobalAlloc's contract for layout.
-        unsafe { rustfs_mimalloc::MiMalloc.alloc_zeroed(layout) }
+        unsafe { mimalloc::MiMalloc.alloc_zeroed(layout) }
    }

    unsafe fn dealloc(&self, ptr: *mut u8, layout: Layout) {
        // SAFETY: ptr and layout came from this allocator and are forwarded unchanged.
-        unsafe { rustfs_mimalloc::MiMalloc.dealloc(ptr, layout) }
+        unsafe { mimalloc::MiMalloc.dealloc(ptr, layout) }
    }

    unsafe fn realloc(&self, ptr: *mut u8, layout: Layout, new_size: usize) -> *mut u8 {
        // SAFETY: ptr and layout came from this allocator and are forwarded unchanged.
-        unsafe { rustfs_mimalloc::MiMalloc.realloc(ptr, layout, new_size) }
+        unsafe { mimalloc::MiMalloc.realloc(ptr, layout, new_size) }
    }
 }

@@ -51,7 +51,7 @@ static GLOBAL: hotpath::CountingAllocator<MiMallocAllocator> = hotpath::Counting

 #[cfg(not(all(feature = "hotpath", feature = "hotpath-alloc")))]
 #[global_allocator]
-static GLOBAL: rustfs_mimalloc::MiMalloc = rustfs_mimalloc::MiMalloc;
+static GLOBAL: mimalloc::MiMalloc = mimalloc::MiMalloc;

 fn main() {
    let _hotpath_guard = hotpath::HotpathGuardBuilder::new("main").build();
@@ -71,9 +71,8 @@ mod tests {
        allocation.extend_from_slice(&[7_u8; 64]);

        assert_eq!(allocation.len(), 64);
-        let heap = rustfs_mimalloc::heap::Heap::main();
        // SAFETY: the live Vec pointer is valid to inspect for heap ownership.
-        assert!(unsafe { heap.contains(allocation.as_ptr()) });
+        assert!(unsafe { libmimalloc_sys::mi_is_in_heap_region(allocation.as_ptr().cast()) });
    }

    #[test]
@@ -86,13 +85,12 @@ mod tests {
        let layout = Layout::from_size_align(32, 8).expect("valid test allocation layout");
        let grown_layout = Layout::from_size_align(64, 8).expect("valid grown test allocation layout");
        let allocator = super::MiMallocAllocator;
-        let heap = rustfs_mimalloc::heap::Heap::main();

        // SAFETY: The pointer is checked for null before use and later released
        // through the same allocator with the corresponding layout.
        let ptr = unsafe { allocator.alloc_zeroed(layout) };
        assert!(!ptr.is_null());
-        assert!(unsafe { heap.contains(ptr) });
+        assert!(unsafe { libmimalloc_sys::mi_is_in_heap_region(ptr.cast()) });
        assert!(unsafe { std::slice::from_raw_parts(ptr, 32).iter().all(|byte| *byte == 0) });

        // SAFETY: `ptr` was allocated by `allocator` with `layout`; on failure
@@ -104,7 +102,7 @@ mod tests {
            panic!("mimalloc realloc failed in allocator smoke test");
        }

-        assert!(unsafe { heap.contains(grown_ptr) });
+        assert!(unsafe { libmimalloc_sys::mi_is_in_heap_region(grown_ptr.cast()) });
        // SAFETY: `grown_ptr` was reallocated by `allocator` and is released
        // with the matching grown layout.
        unsafe { allocator.dealloc(grown_ptr, grown_layout) };
@@ -17,7 +17,10 @@ use rustfs_io_metrics::{
    record_cpu_usage, record_memory_usage, record_process_memory_split,
 };
 use serde::Serialize;
+#[cfg(any(test, not(target_os = "windows")))]
 use serde_json::Value;
+#[cfg(not(target_os = "windows"))]
+use std::ffi::CStr;
 use std::path::Path;
 use std::sync::{Arc, Mutex, OnceLock};
 use std::time::Duration;
@@ -228,18 +231,7 @@ fn read_cgroup_memory_snapshot() -> Option<CgroupMemorySnapshot> {
    read_cgroup_v2().or_else(read_cgroup_v1)
 }

-fn read_allocator_memory_snapshot() -> Option<AllocatorMemorySnapshot> {
-    let json = rustfs_mimalloc::MiMalloc::stats_json();
-    if json.is_empty() {
-        return None;
-    }
-    let observation = parse_mimalloc_stats_json(&json)?;
-    Some(AllocatorMemorySnapshot {
-        backend: crate::allocator_reclaim::allocator_backend(),
-        observation,
-    })
-}
-
+#[cfg(any(test, not(target_os = "windows")))]
 fn numeric_json_value(value: &Value) -> Option<u64> {
    match value {
        Value::Number(number) => number
@@ -250,6 +242,7 @@ fn numeric_json_value(value: &Value) -> Option<u64> {
    }
 }

+#[cfg(any(test, not(target_os = "windows")))]
 fn numeric_json_field(value: &Value, field: &str) -> Option<u64> {
    match value {
        Value::Object(fields) => fields
@@ -261,6 +254,7 @@ fn numeric_json_field(value: &Value, field: &str) -> Option<u64> {
    }
 }

+#[cfg(any(test, not(target_os = "windows")))]
 fn mimalloc_stat_field(value: &Value, metric: &str, field: &str) -> Option<u64> {
    match value {
        Value::Object(fields) => {
@@ -277,10 +271,12 @@ fn mimalloc_stat_field(value: &Value, metric: &str, field: &str) -> Option<u64>
    }
 }

+#[cfg(any(test, not(target_os = "windows")))]
 fn mimalloc_stat_current(value: &Value, metric: &str) -> Option<u64> {
    mimalloc_stat_field(value, metric, "current")
 }

+#[cfg(any(test, not(target_os = "windows")))]
 fn mimalloc_stat_sum(value: &Value, metrics: &[&str], field: &str) -> Option<u64> {
    metrics
        .iter()
@@ -289,6 +285,7 @@ fn mimalloc_stat_sum(value: &Value, metrics: &[&str], field: &str) -> Option<u64
        .filter(|value| *value > 0)
 }

+#[cfg(any(test, not(target_os = "windows")))]
 fn parse_mimalloc_stats_json(stats_json: &str) -> Option<AllocatorMemoryObservation> {
    let value = serde_json::from_str::<Value>(stats_json).ok()?;
    let malloc_metrics = ["malloc_normal", "malloc_huge"];
@@ -315,6 +312,33 @@ fn parse_mimalloc_stats_json(stats_json: &str) -> Option<AllocatorMemoryObservat
    }
 }

+#[cfg(not(target_os = "windows"))]
+#[allow(unsafe_code)]
+fn read_allocator_memory_snapshot() -> Option<AllocatorMemorySnapshot> {
+    // SAFETY: `mi_stats_get_json` returns a null-terminated JSON buffer owned by
+    // mimalloc when called with a null input buffer. The mimalloc API requires
+    // freeing that buffer with `mi_free`; parsing finishes before the buffer is freed.
+    let observation = unsafe {
+        let stats_ptr = libmimalloc_sys::mi_stats_get_json(0, std::ptr::null_mut());
+        if stats_ptr.is_null() {
+            return None;
+        }
+
+        let observation = CStr::from_ptr(stats_ptr).to_str().ok().and_then(parse_mimalloc_stats_json);
+        libmimalloc_sys::mi_free(stats_ptr.cast());
+        observation?
+    };
+    Some(AllocatorMemorySnapshot {
+        backend: crate::allocator_reclaim::allocator_backend(),
+        observation,
+    })
+}
+
+#[cfg(target_os = "windows")]
+fn read_allocator_memory_snapshot() -> Option<AllocatorMemorySnapshot> {
+    None
+}
+
 fn configured_memory_observability_interval_secs() -> u64 {
    rustfs_utils::get_env_u64(ENV_MEMORY_OBSERVABILITY_INTERVAL_SECS, DEFAULT_MEMORY_OBSERVABILITY_INTERVAL_SECS).max(1)
 }
@@ -542,13 +566,6 @@ mod tests {
        assert_eq!(parse_mimalloc_stats_json(r#"{ "allocator": "unknown" }"#), None);
    }

-    #[test]
-    fn read_allocator_memory_snapshot_uses_mimalloc_stats_json() {
-        let snapshot = super::read_allocator_memory_snapshot();
-        #[cfg(not(target_os = "windows"))]
-        assert!(snapshot.is_some(), "allocator snapshot should be available on non-Windows");
-    }
-
    #[test]
    fn memory_observability_snapshot_reports_disabled_when_metrics_are_disabled() {
        let snapshot = build_memory_observability_status_snapshot(false, 15, false);