Merge branch 'main' into cxymds/fix-1852-remote-recovery

.config/nextest.toml

@@ -252,16 +252,10 @@ test-group = 'ecstore-serial-flaky'
 # cluster, so it keeps the lane's parallel-safe / no-external-dependency
 # properties. The RustFS warm backend has no loopback guard (that guard is
 # replication-only), so it needs no opt-in env for its 127.0.0.1 tier target.
-#
-# Disk compression (backlog#1848): the `compression` module joins the smoke
-# lane so the multipart disk-compression roundtrips (restored after
-# rustfs/rustfs#5169 disabled them) have PR-lane signal, not just merge-gate.
-# Single-node servers on random ports with isolated temp dirs — meets the
-# admission criteria unchanged.
 [profile.e2e-smoke]
 default-filter = """
     package(e2e_test) & (
-        test(/^(delete_marker_migration_semantics|version_id_regression|list_objects_v2_pagination|list_object_versions_regression|list_objects_duplicates|list_buckets_double_slash|list_buckets_auth|list_buckets_iam_filter|leading_slash_key|special_chars|create_bucket_region|delete_objects_versioning|head_object_consistency|head_object_range|copy_object_metadata|copy_object_tagging|copy_source_invalid_date|content_encoding|compression|multipart_storage_class|storage_class_capability|ssec_copy|anonymous_access|bucket_policy_check|presigned_negative|negative_sigv4|admin_auth|notification_webhook|tls_hot_reload|console_smoke|admin_iam_crud|admin_pools|sts_query_compat)_test::|^fake_s3_target::/)
+        test(/^(delete_marker_migration_semantics|version_id_regression|list_objects_v2_pagination|list_object_versions_regression|list_objects_duplicates|list_buckets_double_slash|list_buckets_auth|list_buckets_iam_filter|leading_slash_key|special_chars|create_bucket_region|delete_objects_versioning|head_object_consistency|head_object_range|copy_object_metadata|copy_object_tagging|copy_source_invalid_date|content_encoding|multipart_storage_class|storage_class_capability|ssec_copy|anonymous_access|bucket_policy_check|presigned_negative|negative_sigv4|admin_auth|notification_webhook|tls_hot_reload|console_smoke|admin_iam_crud|admin_pools|sts_query_compat)_test::|^fake_s3_target::/)
         | test(/^replication_extension_test::(test_replication_check_succeeds_with_remote_target|test_replication_check_rejects_target_without_object_lock|test_set_remote_target_rejects_unversioned_source_bucket|test_replication_check_rejects_unversioned_source_bucket|test_replication_check_rejects_missing_replication_config|test_replication_check_rejects_invalid_bucket|test_set_remote_target_rejects_same_bucket_on_same_deployment|test_set_remote_target_rejects_unversioned_target_bucket|test_set_remote_target_update_requires_arn|test_set_remote_target_update_rejects_missing_target|test_set_remote_target_rejects_invalid_target_url|test_set_remote_target_rejects_self_signed_https_target_without_skip_tls_verify|test_set_remote_target_rejects_private_ca_https_target_without_ca_cert_pem|test_list_remote_targets_rejects_empty_bucket|test_list_remote_targets_rejects_invalid_bucket|test_remove_remote_target_rejects_missing_target|test_remove_remote_target_rejects_missing_arn|test_remove_remote_target_rejects_invalid_bucket|test_remove_remote_target_rejects_target_used_by_replication|test_delete_bucket_replication_removes_remote_target)$/)
         | test(/^reliant::lifecycle::/)
         | test(/^reliant::tiering::/)

Cargo.lock

@@ -9459,6 +9459,7 @@ dependencies = [
  "tokio-stream",
  "tokio-util",
  "tonic",
+ "tonic-prost",
  "tower",
  "tracing",
  "tracing-core",

crates/e2e_test/src/common.rs

@@ -67,10 +67,7 @@ fn configured_capture_log_path(temp_dir: &str) -> Option<String> {
     capture_log_path(Path::new(&log_dir), temp_dir).map(|path| path.to_string_lossy().into_owned())
 }
 
-pub(crate) fn capture_command_logs(
-    command: &mut Command,
-    log_path: Option<&str>,
-) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
+fn capture_command_logs(command: &mut Command, log_path: Option<&str>) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
     let Some(log_path) = log_path else {
         return Ok(());
     };

crates/e2e_test/src/compression_test.rs

@@ -2,7 +2,6 @@
 
 use crate::common::{RustFSTestEnvironment, init_logging, rustfs_binary_path};
 use aws_sdk_s3::primitives::ByteStream;
-use aws_sdk_s3::types::{CompletedMultipartUpload, CompletedPart};
 use serial_test::serial;
 use std::fs;
 use std::path::PathBuf;
@@ -26,15 +25,6 @@ fn generate_compressible_data(size: usize) -> Vec<u8> {
     data
 }
 
-/// Deterministic 2048-byte-period binary pattern that compresses extremely well: every part
-/// yields many compressed blocks, which is exactly the shape that reproduced the mid-payload
-/// Pending truncation (rustfs/rustfs#5957).
-fn generate_high_ratio_binary_data(size: usize, seed: u8) -> Vec<u8> {
-    (0..size)
-        .map(|i| ((i as u64).wrapping_mul(2_654_435_761).wrapping_add(seed as u64) >> 3) as u8)
-        .collect()
-}
-
 fn find_part_files(temp_dir: &str, bucket: &str, object_key: &str) -> Vec<PathBuf> {
     let bucket_path = PathBuf::from(temp_dir).join(bucket);
     let mut part_files = Vec::new();
@@ -65,14 +55,9 @@ async fn start_rustfs_with_compression(env: &mut RustFSTestEnvironment) -> Resul
     env.cleanup_existing_processes().await?;
 
     let binary_path = rustfs_binary_path();
-    // Route the child's stdout/stderr through the shared RUSTFS_E2E_LOG_DIR
-    // capture (survives the temp-dir cleanup on Drop and is uploaded as a CI
-    // artifact); without the env var the child inherits stdio as before.
-    let mut command = Command::new(&binary_path);
-    command
+    let process = Command::new(&binary_path)
         .env("RUSTFS_CONSOLE_ENABLE", "false")
         .env("RUSTFS_COMPRESSION_ENABLED", "true")
-        .env("RUSTFS_COMPRESSION_MULTIPART_ENABLED", "true")
         .args([
             "--address",
             &env.address,
@@ -81,9 +66,8 @@ async fn start_rustfs_with_compression(env: &mut RustFSTestEnvironment) -> Resul
             "--secret-key",
             &env.secret_key,
             &env.temp_dir,
-        ]);
-    crate::common::capture_command_logs(&mut command, env.capture_log_path.as_deref())?;
-    let process = command.spawn()?;
+        ])
+        .spawn()?;
 
     env.process = Some(process);
 
@@ -170,647 +154,3 @@ async fn test_compression_roundtrip() -> Result<(), Box<dyn std::error::Error +
     env.stop_server();
     Ok(())
 }
-
-const MULTIPART_COMPRESSION_BUCKET: &str = "compression-multipart-bucket";
-const MPU_PART1_SIZE: usize = 5 * 1024 * 1024;
-const MPU_PART2_SIZE: usize = 1024 * 1024;
-
-async fn multipart_upload(
-    client: &aws_sdk_s3::Client,
-    bucket: &str,
-    key: &str,
-    parts: &[&[u8]],
-) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
-    let create = client.create_multipart_upload().bucket(bucket).key(key).send().await?;
-    let upload_id = create.upload_id().ok_or("missing upload id")?.to_string();
-
-    let mut completed_parts = Vec::with_capacity(parts.len());
-    for (i, part) in parts.iter().enumerate() {
-        let part_number = (i + 1) as i32;
-        let upload = client
-            .upload_part()
-            .bucket(bucket)
-            .key(key)
-            .upload_id(&upload_id)
-            .part_number(part_number)
-            .body(ByteStream::from(part.to_vec()))
-            .send()
-            .await?;
-        completed_parts.push(
-            CompletedPart::builder()
-                .part_number(part_number)
-                .e_tag(upload.e_tag().unwrap_or_default())
-                .build(),
-        );
-    }
-
-    client
-        .complete_multipart_upload()
-        .bucket(bucket)
-        .key(key)
-        .upload_id(&upload_id)
-        .multipart_upload(CompletedMultipartUpload::builder().set_parts(Some(completed_parts)).build())
-        .send()
-        .await?;
-    Ok(())
-}
-
-async fn fetch_range(
-    client: &aws_sdk_s3::Client,
-    bucket: &str,
-    key: &str,
-    range: &str,
-) -> Result<Vec<u8>, Box<dyn std::error::Error + Send + Sync>> {
-    let response = client.get_object().bucket(bucket).key(key).range(range).send().await?;
-    Ok(response.body.collect().await?.into_bytes().to_vec())
-}
-
-/// Multipart disk compression roundtrip: parts are written as independent
-/// compressed streams and every GET shape must reassemble the original bytes
-/// (rustfs/rustfs#5957: multipart uploads previously bypassed disk compression
-/// entirely).
-#[tokio::test]
-#[serial]
-async fn test_compression_multipart_roundtrip() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
-    init_logging();
-    info!("Starting multipart compression roundtrip test");
-
-    let mut env = RustFSTestEnvironment::new().await?;
-    start_rustfs_with_compression(&mut env).await?;
-
-    let client = env.create_s3_client();
-    env.create_test_bucket(MULTIPART_COMPRESSION_BUCKET).await?;
-
-    let object_key = "multipart-compressible.txt";
-    let part1 = generate_compressible_data(MPU_PART1_SIZE);
-    let part2 = generate_compressible_data(MPU_PART2_SIZE);
-    let mut original_data = part1.clone();
-    original_data.extend_from_slice(&part2);
-    let total_size = original_data.len();
-
-    multipart_upload(&client, MULTIPART_COMPRESSION_BUCKET, object_key, &[&part1, &part2]).await?;
-
-    let head_response = client
-        .head_object()
-        .bucket(MULTIPART_COMPRESSION_BUCKET)
-        .key(object_key)
-        .send()
-        .await?;
-    assert_eq!(
-        head_response.content_length().unwrap_or(0) as usize,
-        total_size,
-        "Content-Length should be the logical object size"
-    );
-
-    let part_files = find_part_files(&env.temp_dir, MULTIPART_COMPRESSION_BUCKET, object_key);
-    assert!(!part_files.is_empty(), "expected on-disk part files for the multipart object");
-    let total_physical_size: u64 = part_files.iter().filter_map(|p| fs::metadata(p).ok()).map(|m| m.len()).sum();
-    assert!(
-        total_physical_size < (total_size / 2) as u64,
-        "Physical size {total_physical_size} should be well below original size {total_size} (multipart compression applied)"
-    );
-    info!("Multipart physical storage size: {total_physical_size} bytes (compressed from {total_size} bytes)");
-
-    // Full GET must reassemble both independently compressed parts.
-    let get_response = client
-        .get_object()
-        .bucket(MULTIPART_COMPRESSION_BUCKET)
-        .key(object_key)
-        .send()
-        .await?;
-    let downloaded = get_response.body.collect().await?.into_bytes();
-    assert_eq!(downloaded.len(), total_size);
-    assert_eq!(&downloaded[..], &original_data[..], "full GET data mismatch");
-
-    // Range fully inside part 1.
-    let range_inside_part1 = fetch_range(&client, MULTIPART_COMPRESSION_BUCKET, object_key, "bytes=1024-999423").await?;
-    assert_eq!(&range_inside_part1[..], &original_data[1024..999424], "part-1 range mismatch");
-
-    // Range crossing the part boundary.
-    let boundary_start = MPU_PART1_SIZE - 128 * 1024;
-    let boundary_end = MPU_PART1_SIZE + 128 * 1024 - 1;
-    let range_crossing = fetch_range(
-        &client,
-        MULTIPART_COMPRESSION_BUCKET,
-        object_key,
-        &format!("bytes={boundary_start}-{boundary_end}"),
-    )
-    .await?;
-    assert_eq!(
-        &range_crossing[..],
-        &original_data[boundary_start..boundary_end + 1],
-        "boundary-crossing range mismatch"
-    );
-
-    // Range fully inside part 2.
-    let part2_start = MPU_PART1_SIZE + 4096;
-    let part2_end = MPU_PART1_SIZE + 256 * 1024 - 1;
-    let range_inside_part2 = fetch_range(
-        &client,
-        MULTIPART_COMPRESSION_BUCKET,
-        object_key,
-        &format!("bytes={part2_start}-{part2_end}"),
-    )
-    .await?;
-    assert_eq!(
-        &range_inside_part2[..],
-        &original_data[part2_start..part2_end + 1],
-        "part-2 range mismatch"
-    );
-
-    // Suffix range (last 128 KiB, entirely in part 2).
-    let suffix_len = 128 * 1024;
-    let suffix = fetch_range(&client, MULTIPART_COMPRESSION_BUCKET, object_key, &format!("bytes=-{suffix_len}")).await?;
-    assert_eq!(&suffix[..], &original_data[total_size - suffix_len..], "suffix range mismatch");
-
-    // partNumber GETs must return each original part.
-    for (part_number, expected) in [(1, &part1), (2, &part2)] {
-        let response = client
-            .get_object()
-            .bucket(MULTIPART_COMPRESSION_BUCKET)
-            .key(object_key)
-            .part_number(part_number)
-            .send()
-            .await?;
-        let body = response.body.collect().await?.into_bytes();
-        assert_eq!(&body[..], &expected[..], "partNumber={part_number} GET mismatch");
-    }
-
-    info!("Multipart compression roundtrip test passed");
-    env.delete_test_bucket(MULTIPART_COMPRESSION_BUCKET).await?;
-    env.stop_server();
-    Ok(())
-}
-
-const MPU_HIGH_RATIO_BUCKET: &str = "compression-mpu-high-ratio-bucket";
-
-/// High-ratio binary multipart payload: the object key is on the compression allow-list, so the
-/// disk-compression path runs and each part is stored as many compressed blocks — the shape that
-/// reproduced the mid-payload Pending truncation (rustfs/rustfs#5957). Every GET shape must return
-/// the exact original bytes, and the stored size must show the data really was compressed.
-#[tokio::test]
-#[serial]
-async fn test_compression_multipart_high_ratio_binary_roundtrip() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
-    init_logging();
-    info!("Starting multipart high-ratio binary compression roundtrip test");
-
-    let mut env = RustFSTestEnvironment::new().await?;
-    start_rustfs_with_compression(&mut env).await?;
-
-    let client = env.create_s3_client();
-    env.create_test_bucket(MPU_HIGH_RATIO_BUCKET).await?;
-
-    let object_key = "multipart-high-ratio.txt";
-    let part1 = generate_high_ratio_binary_data(MPU_PART1_SIZE, 7);
-    let part2 = generate_high_ratio_binary_data(MPU_PART2_SIZE, 61);
-    let mut original_data = part1.clone();
-    original_data.extend_from_slice(&part2);
-    let total_size = original_data.len();
-
-    multipart_upload(&client, MPU_HIGH_RATIO_BUCKET, object_key, &[&part1, &part2]).await?;
-
-    let head_response = client
-        .head_object()
-        .bucket(MPU_HIGH_RATIO_BUCKET)
-        .key(object_key)
-        .send()
-        .await?;
-    assert_eq!(
-        head_response.content_length().unwrap_or(0) as usize,
-        total_size,
-        "Content-Length should be the logical object size"
-    );
-
-    // This pattern compresses to roughly 1/50 of its logical size, so a comfortably loose 2x
-    // margin still proves the parts were stored compressed rather than raw or double-encoded.
-    let part_files = find_part_files(&env.temp_dir, MPU_HIGH_RATIO_BUCKET, object_key);
-    assert!(!part_files.is_empty(), "expected on-disk part files for the multipart object");
-    let total_physical_size: u64 = part_files.iter().filter_map(|p| fs::metadata(p).ok()).map(|m| m.len()).sum();
-    assert!(
-        total_physical_size < (total_size as u64) / 2,
-        "Physical size {total_physical_size} should be far below the logical size {total_size} for high-ratio data"
-    );
-    info!("High-ratio multipart physical storage size: {total_physical_size} bytes (logical {total_size} bytes)");
-
-    info!("step: full GET");
-    let get_response = client
-        .get_object()
-        .bucket(MPU_HIGH_RATIO_BUCKET)
-        .key(object_key)
-        .send()
-        .await?;
-    let downloaded = get_response.body.collect().await?.into_bytes();
-    assert_eq!(downloaded.len(), total_size);
-    assert_eq!(&downloaded[..], &original_data[..], "full GET data mismatch");
-
-    // Range crossing the part boundary.
-    info!("step: boundary range GET");
-    let boundary_start = MPU_PART1_SIZE - 128 * 1024;
-    let boundary_end = MPU_PART1_SIZE + 128 * 1024 - 1;
-    let range_crossing = fetch_range(
-        &client,
-        MPU_HIGH_RATIO_BUCKET,
-        object_key,
-        &format!("bytes={boundary_start}-{boundary_end}"),
-    )
-    .await?;
-    assert_eq!(
-        &range_crossing[..],
-        &original_data[boundary_start..boundary_end + 1],
-        "boundary-crossing range mismatch"
-    );
-
-    // partNumber GET for the trailing part.
-    info!("step: partNumber GET");
-    let part2_response = client
-        .get_object()
-        .bucket(MPU_HIGH_RATIO_BUCKET)
-        .key(object_key)
-        .part_number(2)
-        .send()
-        .await?;
-    let part2_body = part2_response.body.collect().await?.into_bytes();
-    assert_eq!(&part2_body[..], &part2[..], "partNumber=2 GET mismatch");
-
-    info!("Multipart high-ratio binary compression roundtrip test passed");
-    env.delete_test_bucket(MPU_HIGH_RATIO_BUCKET).await?;
-    env.stop_server();
-    Ok(())
-}
-
-const MPU_COPY_COMPRESSION_BUCKET: &str = "compression-mpu-copy-bucket";
-const MPU_COPY_SOURCE_SIZE: usize = 6 * 1024 * 1024;
-const MPU_COPY_RANGE_LEN: usize = 5 * 1024 * 1024;
-
-/// UploadPartCopy feeds a part from an already stored (and already compressed) object. The copied
-/// range must be decompressed on read and re-compressed into the destination part, so the final
-/// object has to match "source prefix + uploaded tail" byte for byte.
-#[tokio::test]
-#[serial]
-async fn test_compression_multipart_upload_part_copy_roundtrip() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
-    init_logging();
-    info!("Starting multipart upload-part-copy compression roundtrip test");
-
-    let mut env = RustFSTestEnvironment::new().await?;
-    start_rustfs_with_compression(&mut env).await?;
-
-    let client = env.create_s3_client();
-    env.create_test_bucket(MPU_COPY_COMPRESSION_BUCKET).await?;
-
-    // Source object: a plain PUT that goes through the single-stream compression path.
-    let source_key = "copy-source.txt";
-    let source_data = generate_compressible_data(MPU_COPY_SOURCE_SIZE);
-    client
-        .put_object()
-        .bucket(MPU_COPY_COMPRESSION_BUCKET)
-        .key(source_key)
-        .body(ByteStream::from(source_data.clone()))
-        .send()
-        .await?;
-
-    // Destination object: part 1 copied from the source, part 2 uploaded directly.
-    let target_key = "copy-target.txt";
-    let part2 = generate_compressible_data(MPU_PART2_SIZE);
-    let mut expected_data = source_data[..MPU_COPY_RANGE_LEN].to_vec();
-    expected_data.extend_from_slice(&part2);
-    let total_size = expected_data.len();
-
-    let create = client
-        .create_multipart_upload()
-        .bucket(MPU_COPY_COMPRESSION_BUCKET)
-        .key(target_key)
-        .send()
-        .await?;
-    let upload_id = create.upload_id().ok_or("missing upload id")?.to_string();
-
-    let copy_part = client
-        .upload_part_copy()
-        .bucket(MPU_COPY_COMPRESSION_BUCKET)
-        .key(target_key)
-        .upload_id(&upload_id)
-        .part_number(1)
-        .copy_source(format!("{MPU_COPY_COMPRESSION_BUCKET}/{source_key}"))
-        .copy_source_range(format!("bytes=0-{}", MPU_COPY_RANGE_LEN - 1))
-        .send()
-        .await?;
-    let copy_etag = copy_part
-        .copy_part_result()
-        .and_then(|r| r.e_tag())
-        .ok_or("missing copy part etag")?
-        .to_string();
-
-    let uploaded_part = client
-        .upload_part()
-        .bucket(MPU_COPY_COMPRESSION_BUCKET)
-        .key(target_key)
-        .upload_id(&upload_id)
-        .part_number(2)
-        .body(ByteStream::from(part2.clone()))
-        .send()
-        .await?;
-
-    client
-        .complete_multipart_upload()
-        .bucket(MPU_COPY_COMPRESSION_BUCKET)
-        .key(target_key)
-        .upload_id(&upload_id)
-        .multipart_upload(
-            CompletedMultipartUpload::builder()
-                .parts(CompletedPart::builder().part_number(1).e_tag(copy_etag).build())
-                .parts(
-                    CompletedPart::builder()
-                        .part_number(2)
-                        .e_tag(uploaded_part.e_tag().unwrap_or_default())
-                        .build(),
-                )
-                .build(),
-        )
-        .send()
-        .await?;
-
-    let head_response = client
-        .head_object()
-        .bucket(MPU_COPY_COMPRESSION_BUCKET)
-        .key(target_key)
-        .send()
-        .await?;
-    assert_eq!(
-        head_response.content_length().unwrap_or(0) as usize,
-        total_size,
-        "Content-Length should be the logical object size"
-    );
-
-    let part_files = find_part_files(&env.temp_dir, MPU_COPY_COMPRESSION_BUCKET, target_key);
-    assert!(!part_files.is_empty(), "expected on-disk part files for the copied object");
-    let total_physical_size: u64 = part_files.iter().filter_map(|p| fs::metadata(p).ok()).map(|m| m.len()).sum();
-    assert!(
-        total_physical_size < (total_size / 2) as u64,
-        "Physical size {total_physical_size} should be well below original size {total_size} (copied part compression applied)"
-    );
-
-    let get_response = client
-        .get_object()
-        .bucket(MPU_COPY_COMPRESSION_BUCKET)
-        .key(target_key)
-        .send()
-        .await?;
-    let downloaded = get_response.body.collect().await?.into_bytes();
-    assert_eq!(downloaded.len(), total_size);
-    assert_eq!(&downloaded[..], &expected_data[..], "copied multipart GET data mismatch");
-
-    info!("Multipart upload-part-copy compression roundtrip test passed");
-    env.delete_test_bucket(MPU_COPY_COMPRESSION_BUCKET).await?;
-    env.stop_server();
-    Ok(())
-}
-
-const MPU_THREE_PARTS_BUCKET: &str = "compression-mpu-three-parts-bucket";
-const MPU_THREE_PARTS_TAIL_SIZE: usize = 512 * 1024;
-
-/// Three-part upload with uneven part sizes: each partNumber GET must map back to exactly one
-/// compressed part stream, and a suffix range must resolve inside the trailing part.
-#[tokio::test]
-#[serial]
-async fn test_compression_multipart_three_parts_part_number_gets() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
-    init_logging();
-    info!("Starting three-part multipart compression partNumber test");
-
-    let mut env = RustFSTestEnvironment::new().await?;
-    start_rustfs_with_compression(&mut env).await?;
-
-    let client = env.create_s3_client();
-    env.create_test_bucket(MPU_THREE_PARTS_BUCKET).await?;
-
-    let object_key = "multipart-three-parts.txt";
-    let part1 = generate_compressible_data(MPU_PART1_SIZE);
-    let part2 = generate_compressible_data(MPU_PART1_SIZE);
-    let part3 = generate_compressible_data(MPU_THREE_PARTS_TAIL_SIZE);
-    let mut original_data = part1.clone();
-    original_data.extend_from_slice(&part2);
-    original_data.extend_from_slice(&part3);
-    let total_size = original_data.len();
-
-    multipart_upload(&client, MPU_THREE_PARTS_BUCKET, object_key, &[&part1, &part2, &part3]).await?;
-
-    let head_response = client
-        .head_object()
-        .bucket(MPU_THREE_PARTS_BUCKET)
-        .key(object_key)
-        .send()
-        .await?;
-    assert_eq!(
-        head_response.content_length().unwrap_or(0) as usize,
-        total_size,
-        "Content-Length should be the logical object size"
-    );
-
-    let part_files = find_part_files(&env.temp_dir, MPU_THREE_PARTS_BUCKET, object_key);
-    assert!(!part_files.is_empty(), "expected on-disk part files for the multipart object");
-    let total_physical_size: u64 = part_files.iter().filter_map(|p| fs::metadata(p).ok()).map(|m| m.len()).sum();
-    assert!(
-        total_physical_size < (total_size / 2) as u64,
-        "Physical size {total_physical_size} should be well below original size {total_size} (multipart compression applied)"
-    );
-
-    // Every partNumber GET must return exactly the bytes of the corresponding uploaded part.
-    for (part_number, expected) in [(1, &part1), (2, &part2), (3, &part3)] {
-        let response = client
-            .get_object()
-            .bucket(MPU_THREE_PARTS_BUCKET)
-            .key(object_key)
-            .part_number(part_number)
-            .send()
-            .await?;
-        let body = response.body.collect().await?.into_bytes();
-        assert_eq!(&body[..], &expected[..], "partNumber={part_number} GET mismatch");
-    }
-
-    // Suffix range (last 64 KiB) resolves inside the trailing part.
-    let suffix_len = 64 * 1024;
-    let suffix = fetch_range(&client, MPU_THREE_PARTS_BUCKET, object_key, &format!("bytes=-{suffix_len}")).await?;
-    assert_eq!(&suffix[..], &original_data[total_size - suffix_len..], "suffix range mismatch");
-
-    info!("Three-part multipart compression partNumber test passed");
-    env.delete_test_bucket(MPU_THREE_PARTS_BUCKET).await?;
-    env.stop_server();
-    Ok(())
-}
-
-const MPU_SSE_COMPRESSION_BUCKET: &str = "compression-mpu-sse-bucket";
-
-async fn start_rustfs_with_compression_and_sse(
-    env: &mut RustFSTestEnvironment,
-) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
-    use base64::Engine;
-    env.cleanup_existing_processes().await?;
-
-    let binary_path = rustfs_binary_path();
-    let master_key = base64::engine::general_purpose::STANDARD.encode([0x42u8; 32]);
-    // Server output goes to a file inside the per-test temp dir so a failing
-    // run can be diagnosed from the child's logs.
-    let server_log = std::fs::File::create(format!("{}/server.log", env.temp_dir))?;
-    let server_log_err = server_log.try_clone()?;
-    let process = Command::new(&binary_path)
-        .env("RUSTFS_CONSOLE_ENABLE", "false")
-        .env("RUSTFS_COMPRESSION_ENABLED", "true")
-        .env("RUSTFS_COMPRESSION_MULTIPART_ENABLED", "true")
-        .env("RUSTFS_SSE_S3_MASTER_KEY", master_key)
-        .env("RUST_LOG", "rustfs=info,rustfs_ecstore=info")
-        .stdout(std::process::Stdio::from(server_log))
-        .stderr(std::process::Stdio::from(server_log_err))
-        .args([
-            "--address",
-            &env.address,
-            "--access-key",
-            &env.access_key,
-            "--secret-key",
-            &env.secret_key,
-            &env.temp_dir,
-        ])
-        .spawn()?;
-
-    env.process = Some(process);
-
-    info!("Waiting for RustFS server with compression + SSE-S3 enabled on {}", env.address);
-    for i in 0..30 {
-        if TcpStream::connect(&env.address).await.is_ok() {
-            info!("RustFS server is ready after {} attempts", i + 1);
-            return Ok(());
-        }
-        if i == 29 {
-            return Err("RustFS server failed to become ready".into());
-        }
-        sleep(Duration::from_secs(1)).await;
-    }
-    Ok(())
-}
-
-/// SSE-S3 + disk compression multipart: each part is compressed and then encrypted, and every GET
-/// shape must still return the original plaintext bytes. Physical size must shrink because the
-/// compression runs before encryption.
-#[tokio::test]
-#[serial]
-async fn test_compression_multipart_sse_s3_roundtrip() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
-    use aws_sdk_s3::types::ServerSideEncryption;
-
-    init_logging();
-    info!("Starting SSE-S3 multipart compression roundtrip test");
-
-    let mut env = RustFSTestEnvironment::new().await?;
-    start_rustfs_with_compression_and_sse(&mut env).await?;
-
-    let client = env.create_s3_client();
-    env.create_test_bucket(MPU_SSE_COMPRESSION_BUCKET).await?;
-
-    let object_key = "multipart-sse-compressible.txt";
-    let part1 = generate_compressible_data(MPU_PART1_SIZE);
-    let part2 = generate_compressible_data(MPU_PART2_SIZE);
-    let mut original_data = part1.clone();
-    original_data.extend_from_slice(&part2);
-    let total_size = original_data.len();
-
-    let create = client
-        .create_multipart_upload()
-        .bucket(MPU_SSE_COMPRESSION_BUCKET)
-        .key(object_key)
-        .server_side_encryption(ServerSideEncryption::Aes256)
-        .send()
-        .await?;
-    let upload_id = create.upload_id().ok_or("missing upload id")?.to_string();
-
-    let mut completed_parts = Vec::new();
-    for (i, part) in [&part1, &part2].into_iter().enumerate() {
-        let part_number = (i + 1) as i32;
-        let upload = client
-            .upload_part()
-            .bucket(MPU_SSE_COMPRESSION_BUCKET)
-            .key(object_key)
-            .upload_id(&upload_id)
-            .part_number(part_number)
-            .body(ByteStream::from(part.clone()))
-            .send()
-            .await?;
-        completed_parts.push(
-            CompletedPart::builder()
-                .part_number(part_number)
-                .e_tag(upload.e_tag().unwrap_or_default())
-                .build(),
-        );
-    }
-
-    client
-        .complete_multipart_upload()
-        .bucket(MPU_SSE_COMPRESSION_BUCKET)
-        .key(object_key)
-        .upload_id(&upload_id)
-        .multipart_upload(CompletedMultipartUpload::builder().set_parts(Some(completed_parts)).build())
-        .send()
-        .await?;
-
-    let head_response = client
-        .head_object()
-        .bucket(MPU_SSE_COMPRESSION_BUCKET)
-        .key(object_key)
-        .send()
-        .await?;
-    assert_eq!(
-        head_response.content_length().unwrap_or(0) as usize,
-        total_size,
-        "Content-Length should be the logical object size"
-    );
-    assert_eq!(
-        head_response.server_side_encryption(),
-        Some(&ServerSideEncryption::Aes256),
-        "HEAD must report SSE-S3"
-    );
-
-    let part_files = find_part_files(&env.temp_dir, MPU_SSE_COMPRESSION_BUCKET, object_key);
-    assert!(!part_files.is_empty(), "expected on-disk part files for the multipart object");
-    let total_physical_size: u64 = part_files.iter().filter_map(|p| fs::metadata(p).ok()).map(|m| m.len()).sum();
-    assert!(
-        total_physical_size < (total_size / 2) as u64,
-        "Physical size {total_physical_size} should be well below original size {total_size} (compress-then-encrypt applied)"
-    );
-
-    let get_response = client
-        .get_object()
-        .bucket(MPU_SSE_COMPRESSION_BUCKET)
-        .key(object_key)
-        .send()
-        .await?;
-    let downloaded = get_response.body.collect().await?.into_bytes();
-    assert_eq!(downloaded.len(), total_size);
-    assert_eq!(&downloaded[..], &original_data[..], "SSE-S3 multipart full GET data mismatch");
-
-    // Range crossing the part boundary must decrypt and decompress across parts.
-    let boundary_start = MPU_PART1_SIZE - 64 * 1024;
-    let boundary_end = MPU_PART1_SIZE + 64 * 1024 - 1;
-    let range_crossing = fetch_range(
-        &client,
-        MPU_SSE_COMPRESSION_BUCKET,
-        object_key,
-        &format!("bytes={boundary_start}-{boundary_end}"),
-    )
-    .await?;
-    assert_eq!(
-        &range_crossing[..],
-        &original_data[boundary_start..boundary_end + 1],
-        "SSE-S3 boundary-crossing range mismatch"
-    );
-
-    // partNumber GET for the trailing part.
-    let part2_response = client
-        .get_object()
-        .bucket(MPU_SSE_COMPRESSION_BUCKET)
-        .key(object_key)
-        .part_number(2)
-        .send()
-        .await?;
-    let part2_body = part2_response.body.collect().await?.into_bytes();
-    assert_eq!(&part2_body[..], &part2[..], "SSE-S3 partNumber=2 GET mismatch");
-
-    info!("SSE-S3 multipart compression roundtrip test passed");
-    env.delete_test_bucket(MPU_SSE_COMPRESSION_BUCKET).await?;
-    env.stop_server();
-    Ok(())
-}

crates/e2e_test/src/inline_fast_path_cluster_test.rs

@@ -1828,36 +1828,33 @@ async fn four_node_compressed_inline_fallback() -> TestResult {
     Ok(())
 }
 
-/// Multipart disk compression is live again, so a compression-enabled cluster classifies multipart objects as compressed and the roundtrip (full GET plus partNumber GET) must still return the original bytes.
-/// Reverting the multipart compression fix must fail this test.
 #[tokio::test]
 #[serial]
-async fn four_node_multipart_disk_compression_roundtrip() -> TestResult {
+async fn four_node_multipart_ignores_disk_compression_fallback() -> TestResult {
     init_logging();
 
     let collector = OtlpMetricCollector::start().await?;
     let mut cluster = RustFSTestClusterEnvironment::new(4).await?;
     configure_reader_metric_cluster(&mut cluster, &collector);
     cluster.set_env("RUSTFS_COMPRESSION_ENABLED", "true");
-    cluster.set_env("RUSTFS_COMPRESSION_MULTIPART_ENABLED", "true");
     cluster.start().await?;
 
-    let bucket = "inline-multipart-compression-roundtrip";
+    let bucket = "inline-multipart-compression-fallback";
     cluster.create_test_bucket(bucket).await?;
     let client = cluster.create_s3_client(0)?;
-    let key = "multipart/compressed.txt";
+    let key = "multipart/compression-disabled.txt";
     let (body, second_part, etag) = put_two_part_multipart(&client, bucket, key).await?;
 
     assert_reader_path(
         &collector,
         &client,
-        ReaderPathExpectation::for_class(ReaderObject::new(bucket, key, &body, etag.as_deref(), None), LEGACY_DUPLEX, COMPRESSED),
+        ReaderPathExpectation::for_class(ReaderObject::new(bucket, key, &body, etag.as_deref(), None), LEGACY_DUPLEX, MULTIPART),
     )
     .await?;
     assert_part_number_reader_path(
         &collector,
         &client,
-        PartNumberReaderPathExpectation::new(bucket, key, &second_part, body.len(), COMPRESSED, LEGACY_DUPLEX),
+        PartNumberReaderPathExpectation::new(bucket, key, &second_part, body.len(), MULTIPART, LEGACY_DUPLEX),
     )
     .await?;
 
@@ -1874,7 +1871,6 @@ async fn four_node_mixed_msgpack_compat_mode_preserves_fallback_controls() -> Te
     let sse_master_key = base64::engine::general_purpose::STANDARD.encode([0x42u8; 32]);
     cluster.set_env("RUSTFS_SSE_S3_MASTER_KEY", sse_master_key);
     cluster.set_env("RUSTFS_COMPRESSION_ENABLED", "true");
-    cluster.set_env("RUSTFS_COMPRESSION_MULTIPART_ENABLED", "true");
     configure_mixed_msgpack_cluster(&mut cluster, &collector)?;
     cluster.start().await?;
 
@@ -1894,21 +1890,14 @@ async fn four_node_mixed_msgpack_compat_mode_preserves_fallback_controls() -> Te
         ReaderPathExpectation::for_class(
             ReaderObject::new(bucket, multipart_key, &multipart_body, multipart_etag.as_deref(), None),
             LEGACY_DUPLEX,
-            COMPRESSED,
+            MULTIPART,
         ),
     )
     .await?;
     assert_part_number_reader_path(
         &collector,
         &client,
-        PartNumberReaderPathExpectation::new(
-            bucket,
-            multipart_key,
-            &second_part,
-            multipart_body.len(),
-            COMPRESSED,
-            LEGACY_DUPLEX,
-        ),
+        PartNumberReaderPathExpectation::new(bucket, multipart_key, &second_part, multipart_body.len(), MULTIPART, LEGACY_DUPLEX),
     )
     .await?;
     assert_msgpack_decode_observed(&collector, &decode_before).await?;
@@ -2364,11 +2353,7 @@ async fn four_node_mixed_msgpack_compat_mode_preserves_fallback_controls_during_
     hot_client.create_bucket().bucket(bucket).send().await?;
     put_lifecycle_with_transition_retry(&hot_client, bucket, &tier_name).await?;
 
-    // `.zip` sits on the disk-compression exclusion list: this test pins
-    // msgpack compat controls across ILM transition, and a compressed object
-    // would classify as `compressed` instead of `remote` (and the warm-tier
-    // read path does not decode compression — tracked separately).
-    let key = "transition/mixed-multipart.zip";
+    let key = "transition/mixed-multipart.bin";
     let (body, second_part, etag) = put_two_part_multipart(&hot_client, bucket, key).await?;
     wait_for_transition(&hot_client, bucket, key, &tier_name).await?;
     assert!(

crates/e2e_test/src/kms/kms_ilm_sse_kms_test.rs

@@ -97,7 +97,7 @@ async fn start_enforcing_ilm_server(env: &mut LocalKMSTestEnvironment) -> TestRe
 
     let envs = [
         ("RUSTFS_KMS_ALLOW_INSECURE_DEV_DEFAULTS", "true"),
-        ("RUSTFS_KMS_ENFORCE_SSE_KEY_POLICY", "true"),
+        ("RUSTFS_KMS_ENFORCE_SSE_KEY_POLICY", "false"),
         ("RUSTFS_SCANNER_CYCLE", "1"),
         ("RUSTFS_ILM_PROCESS_TIME", "1"),
         ("RUSTFS_ILM_DEBUG_DAY_SECS", "2"),
@@ -486,6 +486,7 @@ async fn ilm_expiration_on_sse_kms_bucket_under_enforcement() -> TestResult {
 /// depend on scanner scheduling; the 1s scanner cycle stays on as a backstop.
 #[tokio::test]
 #[serial]
+#[ignore = "pins rustfs/rustfs#6025: GET on a transitioned managed-SSE object silently returns corrupt bytes (fails with enforcement on AND off, so it is not an authorization regression); un-ignore with the fix"]
 async fn ilm_transition_on_sse_kms_bucket_under_enforcement_reads_back() -> TestResult {
     init_logging();
 

crates/ecstore/Cargo.toml

@@ -273,6 +273,7 @@ proptest = "1"
 rcgen.workspace = true
 insta = { workspace = true, features = ["yaml", "json"] }
 rustfs-crypto = { workspace = true }
+tonic-prost = { workspace = true }
 
 [build-dependencies]
 shadow-rs = { workspace = true, default-features = false, features = ["build", "metadata"] }

crates/ecstore/src/api/mod.rs

@@ -278,9 +278,7 @@ pub mod cluster {
 }
 
 pub mod compression {
-    pub use crate::io_support::compress::{
-        MIN_DISK_COMPRESSIBLE_SIZE, is_disk_compressible, is_disk_compression_enabled, is_multipart_disk_compression_enabled,
-    };
+    pub use crate::io_support::compress::{MIN_DISK_COMPRESSIBLE_SIZE, is_disk_compressible, is_disk_compression_enabled};
 }
 
 pub mod config {

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

@@ -46,13 +46,15 @@ use crate::bucket::lifecycle::transition_transaction::run_transition_transaction
 use crate::bucket::object_lock::ObjectLockApi;
 use crate::bucket::versioning::VersioningApi as _;
 use crate::bucket::versioning_sys::BucketVersioningSys;
+use crate::client::object_api_utils::new_getobjectreader;
 use crate::disk::error::DiskError;
 use crate::disk::{DeleteOptions, Disk, DiskAPI, RUSTFS_META_BUCKET, RUSTFS_META_MULTIPART_BUCKET, STORAGE_FORMAT_FILE};
 use crate::error::Error;
 use crate::error::StorageError;
-use crate::error::{is_err_object_not_found, is_err_read_quorum, is_err_version_not_found, is_network_or_host_down};
+use crate::error::{
+    error_resp_to_object_err, is_err_object_not_found, is_err_read_quorum, is_err_version_not_found, is_network_or_host_down,
+};
 use crate::object_api::{GetObjectReader, ObjectInfo, ObjectOptions};
-use crate::object_api::{ObjectEncryptionResolver, ReadPlan};
 use crate::services::tier::{
     tier::{TierConfigMgr, TierOperationLease, tier_destination_id_from_metadata},
     warm_backend::WarmBackendGetOpts,
@@ -4398,10 +4400,9 @@ pub async fn get_transitioned_object_reader(
     h: &HeaderMap,
     oi: &ObjectInfo,
     opts: &ObjectOptions,
-    resolver: Option<&dyn ObjectEncryptionResolver>,
 ) -> Result<GetObjectReader, std::io::Error> {
     let tier_config_mgr = runtime_sources::tier_config_mgr_handle();
-    get_transitioned_object_reader_with_tier_manager(bucket, object, rs, h, oi, opts, &tier_config_mgr, resolver).await
+    get_transitioned_object_reader_with_tier_manager(bucket, object, rs, h, oi, opts, &tier_config_mgr).await
 }
 
 fn validate_transition_remote_version(oi: &ObjectInfo) -> Result<bool, std::io::Error> {
@@ -4421,10 +4422,6 @@ fn validate_transition_remote_version(oi: &ObjectInfo) -> Result<bool, std::io::
     }
 }
 
-// The resolver joins the tier manager as the second injected port this read
-// needs; grouping the request half into a struct would churn every call site of
-// a bug fix.
-#[allow(clippy::too_many_arguments)]
 pub(crate) async fn get_transitioned_object_reader_with_tier_manager(
     bucket: &str,
     object: &str,
@@ -4433,7 +4430,6 @@ pub(crate) async fn get_transitioned_object_reader_with_tier_manager(
     oi: &ObjectInfo,
     opts: &ObjectOptions,
     tier_config_mgr: &Arc<RwLock<TierConfigMgr>>,
-    resolver: Option<&dyn ObjectEncryptionResolver>,
 ) -> Result<GetObjectReader, std::io::Error> {
     validate_transition_remote_version(oi)?;
     let expected_identity = tier_destination_id_from_metadata(&oi.user_defined)?;
@@ -4451,16 +4447,11 @@ pub(crate) async fn get_transitioned_object_reader_with_tier_manager(
 
     tgt_client.validate_remote_version_id(&oi.transitioned_object.version_id)?;
 
-    // The same read plan the local path uses, so the tier fetch is positioned in
-    // the object's *stored* coordinate system and the stream is handed the same
-    // decrypt/decompress transforms. Reading an encrypted object's ciphertext
-    // through a plaintext-coordinate range and skipping the transform is how a
-    // transitioned SSE object used to come back as silently corrupt bytes of the
-    // right length (rustfs/rustfs#6025).
-    let plan = ReadPlan::build_for_request(rs.clone(), oi, opts, h, resolver)
-        .await
-        .map_err(|err| std::io::Error::other(format!("building the read plan for {bucket}/{object} failed: {err}")))?;
-    let (off, length) = (plan.storage_offset() as i64, plan.storage_length());
+    let ret = new_getobjectreader(rs, oi, opts, h);
+    if let Err(err) = ret {
+        return Err(error_resp_to_object_err(err, vec![bucket, object]));
+    }
+    let (get_fn, off, length) = ret.expect("get_transitioned_object_reader should succeed after error check");
     let mut gopts = WarmBackendGetOpts::default();
 
     if off >= 0 && length >= 0 {
@@ -4497,10 +4488,7 @@ pub(crate) async fn get_transitioned_object_reader_with_tier_manager(
             );
             e
         })?;
-    let object_reader = plan
-        .into_object_reader(Box::new(reader), oi)
-        .map_err(|err| std::io::Error::other(format!("wrapping the tier stream for {bucket}/{object} failed: {err}")))?;
-    Ok(attach_tier_operation_lease(object_reader, tgt_client))
+    Ok(attach_tier_operation_lease(get_fn(reader, h.clone()), tgt_client))
 }
 
 struct TierOperationLeaseReader {
@@ -5788,7 +5776,6 @@ mod tests {
             &object_info,
             &ObjectOptions::default(),
             &manager,
-            None,
         )
         .await
         .expect("transitioned reader should open");
@@ -5853,7 +5840,6 @@ mod tests {
             &object_info,
             &ObjectOptions::default(),
             &manager,
-            None,
         )
         .await
         {
@@ -5894,7 +5880,6 @@ mod tests {
             &object_info,
             &ObjectOptions::default(),
             &manager,
-            None,
         )
         .await
         {
@@ -6132,7 +6117,6 @@ mod tests {
             &oi,
             &ObjectOptions::default(),
             &manager,
-            None,
         )
         .await
         {
@@ -6156,7 +6140,6 @@ mod tests {
             &oi,
             &ObjectOptions::default(),
             &manager,
-            None,
         )
         .await
         {

crates/ecstore/src/cluster/rpc/remote_disk.rs

@@ -59,7 +59,7 @@ use std::{
     path::PathBuf,
     sync::{
         Arc,
-        atomic::{AtomicU32, Ordering},
+        atomic::{AtomicBool, AtomicU32, Ordering},
     },
     time::Duration,
 };
@@ -216,6 +216,8 @@ where
 
 #[derive(Debug)]
 pub struct RemoteDisk {
+    /// Stable identity for this handle instance; replacement handles receive a new identity.
+    handle_id: Uuid,
     pub id: Mutex<Option<Uuid>>,
     pub addr: String,
     endpoint: Endpoint,
@@ -226,9 +228,22 @@ pub struct RemoteDisk {
     health: Arc<DiskHealthTracker>,
     /// Cancellation token for monitoring tasks
     cancel_token: CancellationToken,
+    recovery_monitor_active: Arc<AtomicBool>,
+    #[cfg(test)]
+    recovery_monitor_start_count: Arc<AtomicU32>,
     data_transport: Arc<dyn InternodeDataTransport>,
 }
 
+struct RecoveryMonitorLease {
+    active: Arc<AtomicBool>,
+}
+
+impl Drop for RecoveryMonitorLease {
+    fn drop(&mut self) {
+        self.active.store(false, Ordering::Release);
+    }
+}
+
 // ── Connection lifecycle (grpc-optimization P3) ──
 
 /// Whether to prewarm the internode control channel in the background at construction (default off).
@@ -368,14 +383,15 @@ impl RemoteDisk {
             .await
     }
 
-    fn recovery_monitor_span(addr: &str, endpoint: &Endpoint) -> tracing::Span {
+    fn recovery_monitor_span(addr: &str, endpoint: &Endpoint, handle_id: Uuid) -> tracing::Span {
         tracing::info_span!(
             "recovery-monitor",
             component = LOG_COMPONENT_ECSTORE,
             subsystem = LOG_SUBSYSTEM_REMOTE_DISK,
             kind = "remote_disk",
             endpoint = %endpoint,
-            addr = %addr
+            addr = %addr,
+            handle_id = %handle_id
         )
     }
 
@@ -411,6 +427,7 @@ impl RemoteDisk {
             rustfs_utils::get_env_bool(ENV_RUSTFS_DRIVE_ACTIVE_MONITORING, DEFAULT_RUSTFS_DRIVE_ACTIVE_MONITORING);
 
         let disk = Self {
+            handle_id: Uuid::new_v4(),
             id: Mutex::new(None),
             addr,
             endpoint: ep.clone(),
@@ -418,6 +435,9 @@ impl RemoteDisk {
             health_check: opt.health_check && env_health_check,
             health: Arc::new(DiskHealthTracker::new()),
             cancel_token: CancellationToken::new(),
+            recovery_monitor_active: Arc::new(AtomicBool::new(false)),
+            #[cfg(test)]
+            recovery_monitor_start_count: Arc::new(AtomicU32::new(0)),
             data_transport,
         };
         record_drive_runtime_state(ep, RuntimeDriveHealthState::Online);
@@ -435,6 +455,16 @@ impl RemoteDisk {
         self.health.runtime_state()
     }
 
+    #[cfg(test)]
+    fn recovery_monitor_is_active(&self) -> bool {
+        self.recovery_monitor_active.load(Ordering::Acquire)
+    }
+
+    #[cfg(test)]
+    fn recovery_monitor_start_count(&self) -> u32 {
+        self.recovery_monitor_start_count.load(Ordering::Acquire)
+    }
+
     pub fn offline_duration_secs(&self) -> Option<u64> {
         self.health.offline_duration().map(|duration| duration.as_secs())
     }
@@ -573,13 +603,54 @@ impl RemoteDisk {
             return;
         }
 
-        let addr = self.addr.clone();
-        let endpoint = self.endpoint.clone();
-        let health = Arc::clone(&self.health);
-        let cancel_token = self.cancel_token.clone();
-        let span = Self::recovery_monitor_span(&addr, &endpoint);
+        Self::schedule_recovery_monitor(
+            self.addr.clone(),
+            self.endpoint.clone(),
+            self.handle_id,
+            Arc::clone(&self.health),
+            self.cancel_token.clone(),
+            Arc::clone(&self.recovery_monitor_active),
+            #[cfg(test)]
+            Arc::clone(&self.recovery_monitor_start_count),
+        );
+    }
+
+    fn schedule_recovery_monitor(
+        addr: String,
+        endpoint: Endpoint,
+        handle_id: Uuid,
+        health: Arc<DiskHealthTracker>,
+        cancel_token: CancellationToken,
+        active: Arc<AtomicBool>,
+        #[cfg(test)] start_count: Arc<AtomicU32>,
+    ) {
+        if active
+            .compare_exchange(false, true, Ordering::AcqRel, Ordering::Acquire)
+            .is_err()
+        {
+            return;
+        }
+        let span = Self::recovery_monitor_span(&addr, &endpoint, handle_id);
         super::spawn_background_monitor(span, async move {
-            Self::monitor_remote_disk_recovery(addr, endpoint, health, cancel_token).await;
+            #[cfg(test)]
+            start_count.fetch_add(1, Ordering::AcqRel);
+            let lease = RecoveryMonitorLease {
+                active: Arc::clone(&active),
+            };
+            Self::monitor_remote_disk_recovery(addr.clone(), endpoint.clone(), Arc::clone(&health), cancel_token.clone()).await;
+            drop(lease);
+            if !cancel_token.is_cancelled() && health.runtime_state() != RuntimeDriveHealthState::Online {
+                Self::schedule_recovery_monitor(
+                    addr,
+                    endpoint,
+                    handle_id,
+                    health,
+                    cancel_token,
+                    active,
+                    #[cfg(test)]
+                    start_count,
+                );
+            }
         });
     }
 
@@ -588,7 +659,7 @@ impl RemoteDisk {
         let (tx, rx) = tokio::sync::oneshot::channel();
         let endpoint = self.endpoint.clone();
         let addr = self.addr.clone();
-        let span = Self::recovery_monitor_span(&addr, &endpoint);
+        let span = Self::recovery_monitor_span(&addr, &endpoint, self.handle_id);
         super::spawn_background_monitor(span, async move {
             warn!(
                 event = EVENT_REMOTE_DISK_HEALTH,
@@ -619,9 +690,11 @@ impl RemoteDisk {
         let cancel_token = self.cancel_token.clone();
         let addr = self.addr.clone();
         let endpoint = self.endpoint.clone();
+        let handle_id = self.handle_id;
+        let recovery_monitor_active = Arc::clone(&self.recovery_monitor_active);
 
         tokio::spawn(async move {
-            Self::monitor_remote_disk_health(addr, endpoint, health, cancel_token).await;
+            Self::monitor_remote_disk_health(addr, endpoint, handle_id, health, cancel_token, recovery_monitor_active).await;
         });
     }
 
@@ -629,8 +702,10 @@ impl RemoteDisk {
     async fn monitor_remote_disk_health(
         addr: String,
         endpoint: Endpoint,
+        handle_id: Uuid,
         health: Arc<DiskHealthTracker>,
         cancel_token: CancellationToken,
+        recovery_monitor_active: Arc<AtomicBool>,
     ) {
         let mut interval = time::interval(get_drive_active_check_interval());
 
@@ -655,11 +730,16 @@ impl RemoteDisk {
             let addr_clone = addr.clone();
             let endpoint_clone = endpoint.clone();
             let cancel_clone = cancel_token.clone();
-            let span = Self::recovery_monitor_span(&addr_clone, &endpoint_clone);
-
-            super::spawn_background_monitor(span, async move {
-                Self::monitor_remote_disk_recovery(addr_clone, endpoint_clone, health_clone, cancel_clone).await;
-            });
+            Self::schedule_recovery_monitor(
+                addr_clone,
+                endpoint_clone,
+                handle_id,
+                health_clone,
+                cancel_clone,
+                Arc::clone(&recovery_monitor_active),
+                #[cfg(test)]
+                Arc::new(AtomicU32::new(0)),
+            );
         }
 
         loop {
@@ -718,11 +798,16 @@ impl RemoteDisk {
                         let addr_clone = addr.clone();
                         let endpoint_clone = endpoint.clone();
                         let cancel_clone = cancel_token.clone();
-                        let span = Self::recovery_monitor_span(&addr_clone, &endpoint_clone);
-
-                        super::spawn_background_monitor(span, async move {
-                            Self::monitor_remote_disk_recovery(addr_clone, endpoint_clone, health_clone, cancel_clone).await;
-                        });
+                        Self::schedule_recovery_monitor(
+                            addr_clone,
+                            endpoint_clone,
+                            handle_id,
+                            health_clone,
+                            cancel_clone,
+                            Arc::clone(&recovery_monitor_active),
+                            #[cfg(test)]
+                            Arc::new(AtomicU32::new(0)),
+                        );
                     }
                 }
             }
@@ -973,6 +1058,7 @@ impl RemoteDisk {
                 subsystem = LOG_SUBSYSTEM_REMOTE_DISK,
                 endpoint = %self.endpoint,
                 addr = %self.addr,
+                handle_id = %self.handle_id,
                 op,
                 state = "faulty_short_circuit",
                 "Remote disk operation short-circuited by faulty state"
@@ -3116,15 +3202,23 @@ mod tests {
     use super::*;
     use crate::cluster::rpc::internode_data_transport::{InternodeDataTransportCapabilities, TcpHttpInternodeDataTransport};
     use crate::runtime::sources as runtime_sources;
+    use rustfs_protos::proto_gen::node_service::{DiskInfoResponse, ReadAllResponse};
     use serde_json::Value;
     use serial_test::serial;
+    use std::convert::Infallible;
+    use std::future::Future;
     use std::io::{self as std_io, Write};
     use std::pin::Pin;
     use std::sync::{Arc, Mutex, Mutex as StdMutex, Once};
     use std::task::{Context, Poll};
     use tokio::io::{ReadBuf, duplex};
     use tokio::net::TcpListener;
-    use tonic::transport::Endpoint as TonicEndpoint;
+    use tonic::transport::{Endpoint as TonicEndpoint, Server};
+    use tonic::{Response, Status};
+    use tonic::{
+        codegen::{Body as HttpBody, BoxFuture, StdError, http},
+        server::NamedService,
+    };
     use tracing::Level;
     use tracing_subscriber::{Registry, fmt::MakeWriter, layer::SubscriberExt};
     use uuid::Uuid;
@@ -3284,6 +3378,205 @@ mod tests {
         ns_scanner_probe_status: Arc<StdMutex<Option<u16>>>,
     }
 
+    #[derive(Clone, Debug)]
+    struct AuthenticatedReadPeer {
+        audience: String,
+        disk_info_calls: Arc<AtomicU32>,
+        read_all_calls: Arc<AtomicU32>,
+        read_all_disks: Arc<StdMutex<Vec<String>>>,
+        read_all_data: Bytes,
+    }
+
+    impl AuthenticatedReadPeer {
+        fn new(audience: String, read_all_data: Bytes) -> Self {
+            Self {
+                audience,
+                disk_info_calls: Arc::new(AtomicU32::new(0)),
+                read_all_calls: Arc::new(AtomicU32::new(0)),
+                read_all_disks: Arc::default(),
+                read_all_data,
+            }
+        }
+
+        fn disk_info_calls(&self) -> u32 {
+            self.disk_info_calls.load(Ordering::Acquire)
+        }
+
+        fn read_all_calls(&self) -> u32 {
+            self.read_all_calls.load(Ordering::Acquire)
+        }
+
+        fn read_all_disks(&self) -> Vec<String> {
+            self.read_all_disks.lock().expect("read_all disk list lock poisoned").clone()
+        }
+
+        fn verify_auth<T>(&self, request: &Request<T>, path: &str) -> std::result::Result<(), Status> {
+            let headers = request.metadata().clone().into_headers();
+            crate::cluster::rpc::verify_tonic_rpc_signature(&self.audience, path, &headers)
+                .map_err(|err| Status::unauthenticated(err.to_string()))
+        }
+    }
+
+    #[derive(Clone, Debug)]
+    struct AuthenticatedReadPeerService {
+        peer: AuthenticatedReadPeer,
+    }
+
+    impl NamedService for AuthenticatedReadPeerService {
+        const NAME: &'static str = "node_service.NodeService";
+    }
+
+    impl<B> tower::Service<http::Request<B>> for AuthenticatedReadPeerService
+    where
+        B: HttpBody + Send + 'static,
+        B::Error: Into<StdError> + Send + 'static,
+    {
+        type Response = http::Response<tonic::body::Body>;
+        type Error = Infallible;
+        type Future = BoxFuture<Self::Response, Self::Error>;
+
+        fn poll_ready(&mut self, _cx: &mut Context<'_>) -> Poll<std::result::Result<(), Self::Error>> {
+            Poll::Ready(Ok(()))
+        }
+
+        fn call(&mut self, request: http::Request<B>) -> Self::Future {
+            match request.uri().path() {
+                "/node_service.NodeService/DiskInfo" => {
+                    #[derive(Clone)]
+                    struct DiskInfoSvc(AuthenticatedReadPeer);
+
+                    impl tonic::server::UnaryService<DiskInfoRequest> for DiskInfoSvc {
+                        type Response = DiskInfoResponse;
+                        type Future = Pin<Box<dyn Future<Output = std::result::Result<Response<Self::Response>, Status>> + Send>>;
+
+                        fn call(&mut self, request: Request<DiskInfoRequest>) -> Self::Future {
+                            let peer = self.0.clone();
+                            Box::pin(async move {
+                                peer.verify_auth(&request, "/node_service.NodeService/DiskInfo")?;
+                                let request = request.into_inner();
+                                let opts = serde_json::from_str::<DiskInfoOptions>(&request.opts)
+                                    .map_err(|err| Status::invalid_argument(err.to_string()))?;
+                                if !opts.noop {
+                                    return Err(Status::invalid_argument("recovery probe must use noop disk_info"));
+                                }
+                                peer.disk_info_calls.fetch_add(1, Ordering::AcqRel);
+                                let disk_info = serde_json::to_string(&DiskInfo {
+                                    total: 1,
+                                    free: 1,
+                                    endpoint: request.disk,
+                                    ..Default::default()
+                                })
+                                .map_err(|err| Status::internal(err.to_string()))?;
+                                Ok(Response::new(DiskInfoResponse {
+                                    success: true,
+                                    disk_info,
+                                    error: None,
+                                }))
+                            })
+                        }
+                    }
+
+                    let peer = self.peer.clone();
+                    Box::pin(async move {
+                        let method = DiskInfoSvc(peer);
+                        let codec = tonic_prost::ProstCodec::default();
+                        let mut grpc = tonic::server::Grpc::new(codec);
+                        Ok(grpc.unary(method, request).await)
+                    })
+                }
+                "/node_service.NodeService/ReadAll" => {
+                    #[derive(Clone)]
+                    struct ReadAllSvc(AuthenticatedReadPeer);
+
+                    impl tonic::server::UnaryService<ReadAllRequest> for ReadAllSvc {
+                        type Response = ReadAllResponse;
+                        type Future = Pin<Box<dyn Future<Output = std::result::Result<Response<Self::Response>, Status>> + Send>>;
+
+                        fn call(&mut self, request: Request<ReadAllRequest>) -> Self::Future {
+                            let peer = self.0.clone();
+                            Box::pin(async move {
+                                peer.verify_auth(&request, "/node_service.NodeService/ReadAll")?;
+                                let request = request.into_inner();
+                                peer.read_all_calls.fetch_add(1, Ordering::AcqRel);
+                                peer.read_all_disks
+                                    .lock()
+                                    .expect("read_all disk list lock poisoned")
+                                    .push(request.disk);
+                                Ok(Response::new(ReadAllResponse {
+                                    success: true,
+                                    data: peer.read_all_data.clone(),
+                                    error: None,
+                                }))
+                            })
+                        }
+                    }
+
+                    let peer = self.peer.clone();
+                    Box::pin(async move {
+                        let method = ReadAllSvc(peer);
+                        let codec = tonic_prost::ProstCodec::default();
+                        let mut grpc = tonic::server::Grpc::new(codec);
+                        Ok(grpc.unary(method, request).await)
+                    })
+                }
+                _ => Box::pin(async move {
+                    let mut response = http::Response::new(tonic::body::Body::default());
+                    let headers = response.headers_mut();
+                    headers.insert(tonic::Status::GRPC_STATUS, (tonic::Code::Unimplemented as i32).into());
+                    headers.insert(http::header::CONTENT_TYPE, tonic::metadata::GRPC_CONTENT_TYPE);
+                    Ok(response)
+                }),
+            }
+        }
+    }
+
+    struct TestGrpcPeer {
+        addr: String,
+        peer: AuthenticatedReadPeer,
+        shutdown: CancellationToken,
+        task: tokio::task::JoinHandle<()>,
+    }
+
+    impl TestGrpcPeer {
+        async fn spawn(read_all_data: Bytes) -> Option<Self> {
+            let listener = match TcpListener::bind("127.0.0.1:0").await {
+                Ok(listener) => listener,
+                Err(err) if err.kind() == std::io::ErrorKind::PermissionDenied => return None,
+                Err(err) => panic!("test gRPC listener should bind: {err}"),
+            };
+            let socket_addr = listener.local_addr().expect("listener local address should be available");
+            let addr = format!("http://{socket_addr}");
+            let audience = crate::cluster::rpc::normalize_tonic_rpc_audience(&socket_addr.to_string())
+                .expect("test audience should normalize");
+            let peer = AuthenticatedReadPeer::new(audience, read_all_data);
+            let service = AuthenticatedReadPeerService { peer: peer.clone() };
+            let shutdown = CancellationToken::new();
+            let shutdown_for_task = shutdown.clone();
+            let incoming = futures_util::stream::unfold(listener, |listener| async {
+                Some((listener.accept().await.map(|(stream, _)| stream), listener))
+            });
+            let task = tokio::spawn(async move {
+                Server::builder()
+                    .add_service(service)
+                    .serve_with_incoming_shutdown(incoming, shutdown_for_task.cancelled_owned())
+                    .await
+                    .expect("test gRPC peer should serve");
+            });
+
+            Some(Self {
+                addr,
+                peer,
+                shutdown,
+                task,
+            })
+        }
+
+        async fn stop(self) {
+            self.shutdown.cancel();
+            let _ = self.task.await;
+        }
+    }
+
     impl RecordingInternodeDataTransport {
         fn with_ns_scanner_probe_status(status: u16) -> Self {
             Self {
@@ -4397,6 +4690,152 @@ mod tests {
         accept_task.abort();
     }
 
+    #[tokio::test]
+    async fn faulty_handle_runs_only_one_recovery_monitor() {
+        let endpoint = Endpoint {
+            url: url::Url::parse("http://remote-node:9000/data/rustfs0").expect("endpoint should parse"),
+            is_local: false,
+            pool_idx: 0,
+            set_idx: 0,
+            disk_idx: 0,
+        };
+        let disk = RemoteDisk::new(
+            &endpoint,
+            &DiskOption {
+                cleanup: false,
+                health_check: true,
+            },
+            Arc::new(TcpHttpInternodeDataTransport),
+        )
+        .await
+        .expect("remote disk should construct");
+        if !disk.health_check {
+            return;
+        }
+
+        disk.force_runtime_state_for_test(RuntimeDriveHealthState::Offline);
+        disk.spawn_recovery_monitor_if_needed();
+        disk.spawn_recovery_monitor_if_needed();
+        tokio::time::timeout(Duration::from_secs(1), async {
+            while disk.recovery_monitor_start_count() == 0 {
+                tokio::task::yield_now().await;
+            }
+        })
+        .await
+        .expect("recovery monitor should start");
+
+        assert!(disk.recovery_monitor_is_active(), "only one recovery monitor should own the handle");
+        assert_eq!(
+            disk.recovery_monitor_start_count(),
+            1,
+            "the failed compare-exchange path must not start a second monitor"
+        );
+
+        disk.cancel_token.cancel();
+        tokio::time::timeout(Duration::from_secs(1), async {
+            while disk.recovery_monitor_is_active() {
+                tokio::task::yield_now().await;
+            }
+        })
+        .await
+        .expect("cancelled recovery monitor should release its single-flight state");
+        assert!(!disk.recovery_monitor_is_active());
+    }
+
+    #[tokio::test]
+    #[serial(remote_disk_recovery_probe)]
+    async fn recovery_monitor_restores_online_then_real_reads_use_replacement_handle() {
+        runtime_sources::ensure_test_rpc_secret();
+        let Some(peer) = TestGrpcPeer::spawn(Bytes::from_static(b"replacement-data")).await else {
+            return;
+        };
+        let url = url::Url::parse(&format!("{}/data/rustfs0", peer.addr)).expect("endpoint should parse");
+        let endpoint = Endpoint {
+            url,
+            is_local: false,
+            pool_idx: 0,
+            set_idx: 0,
+            disk_idx: 0,
+        };
+        let disk = RemoteDisk::new(
+            &endpoint,
+            &DiskOption {
+                cleanup: false,
+                health_check: true,
+            },
+            Arc::new(TcpHttpInternodeDataTransport),
+        )
+        .await
+        .expect("remote disk should construct");
+        disk.force_runtime_state_for_test(RuntimeDriveHealthState::Offline);
+
+        temp_env::async_with_vars(
+            [
+                (rustfs_config::ENV_DRIVE_RETURNING_PROBE_INTERVAL_SECS, Some("1")),
+                (rustfs_config::ENV_DRIVE_RETURNING_SUCCESS_THRESHOLD, Some("3")),
+                (rustfs_config::ENV_DRIVE_ACTIVE_CHECK_TIMEOUT_SECS, Some("1")),
+            ],
+            async {
+                let monitor = tokio::spawn(RemoteDisk::monitor_remote_disk_recovery(
+                    disk.addr.clone(),
+                    endpoint.clone(),
+                    Arc::clone(&disk.health),
+                    disk.cancel_token.clone(),
+                ));
+
+                tokio::time::timeout(Duration::from_secs(5), async {
+                    while disk.runtime_state() != RuntimeDriveHealthState::Online {
+                        tokio::time::sleep(Duration::from_millis(20)).await;
+                    }
+                })
+                .await
+                .expect("three authenticated recovery probes should restore the disk online");
+                monitor.await.expect("recovery monitor should exit after restoring Online");
+
+                assert_eq!(
+                    peer.peer.disk_info_calls(),
+                    3,
+                    "RemoteDisk recovery requires the configured three successful disk_info probes"
+                );
+                let recovered_read = disk.read_all("bucket", "object").await.expect("recovered handle should read");
+                assert_eq!(recovered_read, Bytes::from_static(b"replacement-data"));
+
+                let replacement = RemoteDisk::new(
+                    &endpoint,
+                    &DiskOption {
+                        cleanup: false,
+                        health_check: false,
+                    },
+                    Arc::new(TcpHttpInternodeDataTransport),
+                )
+                .await
+                .expect("replacement remote disk should construct");
+                let replacement_id = Uuid::new_v4();
+                replacement
+                    .set_disk_id(Some(replacement_id))
+                    .await
+                    .expect("replacement disk id should set");
+                let replacement_read = replacement
+                    .read_all("bucket", "object")
+                    .await
+                    .expect("replacement handle should route real reads");
+
+                assert_eq!(replacement_read, Bytes::from_static(b"replacement-data"));
+                assert_eq!(peer.peer.read_all_calls(), 2);
+                assert_eq!(
+                    peer.peer.read_all_disks(),
+                    vec![endpoint.to_string(), replacement_id.to_string()],
+                    "real reads must use the current handle's disk reference"
+                );
+                disk.cancel_token.cancel();
+                replacement.cancel_token.cancel();
+            },
+        )
+        .await;
+
+        peer.stop().await;
+    }
+
     #[tokio::test]
     async fn test_copy_stream_with_buffer_copies_full_payload() {
         let payload = b"walk-dir-stream".repeat(1024);

crates/ecstore/src/data_usage/local_snapshot.rs

@@ -12,16 +12,6 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-//! Per-disk usage snapshots persisted under the metadata bucket.
-//!
-//! **Nothing calls into this module.** It landed complete with tests in #5307
-//! (2026-07-27) and its aggregation entry point,
-//! [`crate::data_usage::aggregate_local_snapshots`], has never had a caller in
-//! the tree's history. The live data-usage path is
-//! `load_data_usage_from_backend` / `store_data_usage_in_backend`. The items
-//! below therefore carry individual `dead_code` allows rather than a module
-//! blanket, so the gap stays greppable until it is either wired up or removed.
-
 use crate::data_usage::BucketUsageInfo;
 use crate::disk::RUSTFS_META_BUCKET;
 use crate::error::{Error, Result};
@@ -36,12 +26,10 @@ pub const DATA_USAGE_DIR: &str = "datausage";
 /// Directory used to store incremental scan state files under the metadata bucket.
 pub const DATA_USAGE_STATE_DIR: &str = "datausage/state";
 /// Snapshot file format version, allows forward compatibility if the structure evolves.
-#[allow(dead_code, reason = "unwired local usage-snapshot feature; see module docs (backlog#1823)")]
 pub const LOCAL_USAGE_SNAPSHOT_VERSION: u32 = 1;
 
 /// Additional metadata describing which disk produced the snapshot.
 #[derive(Debug, Clone, Serialize, Deserialize, Default)]
-#[allow(dead_code, reason = "unwired local usage-snapshot feature; see module docs (backlog#1823)")]
 pub struct LocalUsageSnapshotMeta {
     /// Disk UUID stored as a string for simpler serialization.
     pub disk_id: String,
@@ -55,7 +43,6 @@ pub struct LocalUsageSnapshotMeta {
 
 /// Usage snapshot produced by a single disk.
 #[derive(Debug, Clone, Serialize, Deserialize, Default)]
-#[allow(dead_code, reason = "unwired local usage-snapshot feature; see module docs (backlog#1823)")]
 pub struct LocalUsageSnapshot {
     /// Format version recorded in the snapshot.
     pub format_version: u32,
@@ -77,7 +64,6 @@ pub struct LocalUsageSnapshot {
     pub objects_total_size: u64,
 }
 
-#[allow(dead_code, reason = "unwired local usage-snapshot feature; see module docs (backlog#1823)")]
 impl LocalUsageSnapshot {
     /// Create an empty snapshot with the default format version filled in.
     pub fn new(meta: LocalUsageSnapshotMeta) -> Self {
@@ -113,13 +99,11 @@ impl LocalUsageSnapshot {
 }
 
 /// Build the snapshot file name `<disk-id>.json`.
-#[allow(dead_code, reason = "unwired local usage-snapshot feature; see module docs (backlog#1823)")]
 pub fn snapshot_file_name(disk_id: &str) -> String {
     format!("{disk_id}.json")
 }
 
 /// Build the object path relative to `RUSTFS_META_BUCKET`, e.g. `datausage/<disk-id>.json`.
-#[allow(dead_code, reason = "unwired local usage-snapshot feature; see module docs (backlog#1823)")]
 pub fn snapshot_object_path(disk_id: &str) -> String {
     format!("{}/{}", DATA_USAGE_DIR, snapshot_file_name(disk_id))
 }
@@ -135,13 +119,11 @@ pub fn data_usage_state_dir(root: &Path) -> PathBuf {
 }
 
 /// Build the absolute path to the snapshot file for the provided disk ID.
-#[allow(dead_code, reason = "unwired local usage-snapshot feature; see module docs (backlog#1823)")]
 pub fn snapshot_path(root: &Path, disk_id: &str) -> PathBuf {
     data_usage_dir(root).join(snapshot_file_name(disk_id))
 }
 
 /// Read a snapshot from disk if it exists.
-#[allow(dead_code, reason = "unwired local usage-snapshot feature; see module docs (backlog#1823)")]
 pub async fn read_snapshot(root: &Path, disk_id: &str) -> Result<Option<LocalUsageSnapshot>> {
     let path = snapshot_path(root, disk_id);
     match fs::read(&path).await {
@@ -156,7 +138,6 @@ pub async fn read_snapshot(root: &Path, disk_id: &str) -> Result<Option<LocalUsa
 }
 
 /// Persist a snapshot to disk, creating directories as needed and overwriting any existing file.
-#[allow(dead_code, reason = "unwired local usage-snapshot feature; see module docs (backlog#1823)")]
 pub async fn write_snapshot(root: &Path, disk_id: &str, snapshot: &LocalUsageSnapshot) -> Result<()> {
     let dir = data_usage_dir(root);
     fs::create_dir_all(&dir).await.map_err(Error::other)?;

crates/ecstore/src/data_usage/mod.rs

@@ -13,6 +13,7 @@
 // limitations under the License.
 
 // #730: scanner/data-usage state is partially migrated and still owns staged cache helpers.
+#![allow(dead_code)]
 
 pub mod local_snapshot;
 
@@ -33,8 +34,8 @@ use crate::{
 pub use local_snapshot::{LocalUsageSnapshot, read_snapshot as read_local_snapshot, snapshot_path};
 use rustfs_data_usage::{
     BucketTargetUsageInfo, BucketUsageInfo, CompressionTotalInfo, DATA_USAGE_OBJECT_NAME, DATA_USAGE_OBSERVED_OBJECT_NAME,
-    DataUsageCache, DataUsageInfo, DiskUsageStatus, LEGACY_DATA_USAGE_OBJECT_NAME, SizeHistogram, VersionsHistogram,
-    observed_data_usage_is_newer,
+    DataUsageCache, DataUsageEntry, DataUsageInfo, DiskUsageStatus, LEGACY_DATA_USAGE_OBJECT_NAME, SizeHistogram, SizeSummary,
+    VersionsHistogram, observed_data_usage_is_newer,
 };
 use rustfs_io_metrics::record_system_path_failure;
 use rustfs_utils::path::SLASH_SEPARATOR;
@@ -54,6 +55,7 @@ use tracing::{debug, error, info, instrument};
 // Data usage storage constants
 pub const DATA_USAGE_ROOT: &str = SLASH_SEPARATOR;
 const DATA_COMPRESSION_TOTAL_NAME: &str = ".compression.json";
+const DATA_USAGE_BLOOM_NAME: &str = ".bloomcycle.bin";
 pub const DATA_USAGE_CACHE_NAME: &str = ".usage-cache.bin";
 const DATA_USAGE_CACHE_TTL_SECS: u64 = 30;
 const LIVE_BUCKET_USAGE_MAX_ENTRIES: u64 = 1024;
@@ -311,6 +313,11 @@ lazy_static::lazy_static! {
         LEGACY_DATA_USAGE_OBJECT_NAME
     );
     static ref LEGACY_DATA_USAGE_OBJ_BACKUP_PATH: String = format!("{}.bkp", LEGACY_DATA_USAGE_OBJ_NAME_PATH.as_str());
+    pub static ref DATA_USAGE_BLOOM_NAME_PATH: String = format!("{}{}{}",
+        crate::disk::BUCKET_META_PREFIX,
+        SLASH_SEPARATOR,
+        DATA_USAGE_BLOOM_NAME
+    );
     pub static ref DATA_COMPRESSION_TOTAL_NAME_PATH: String = format!("{}{}{}",
         crate::disk::BUCKET_META_PREFIX,
         SLASH_SEPARATOR,
@@ -851,10 +858,6 @@ async fn resolve_loaded_snapshot_pair_with_source(
     }
 }
 
-#[allow(
-    dead_code,
-    reason = "primary/backup snapshot fallback asserted by this file's tests (backlog#1823)"
-)]
 async fn resolve_loaded_snapshot(
     primary: Result<Vec<u8>, Error>,
     backup: impl Future<Output = Result<Vec<u8>, Error>>,
@@ -1184,10 +1187,6 @@ pub async fn invalidate_admin_data_usage_snapshot_cache() {
 }
 
 /// Aggregate usage information from local disk snapshots.
-#[allow(
-    dead_code,
-    reason = "reached only through aggregate_local_snapshots, which has no caller (backlog#1823)"
-)]
 fn merge_snapshot(aggregated: &mut DataUsageInfo, mut snapshot: LocalUsageSnapshot, latest_update: &mut Option<SystemTime>) {
     if let Some(update) = snapshot.last_update
         && latest_update.is_none_or(|current| update > current)
@@ -1221,10 +1220,6 @@ fn merge_snapshot(aggregated: &mut DataUsageInfo, mut snapshot: LocalUsageSnapsh
     }
 }
 
-#[allow(
-    dead_code,
-    reason = "entry point of the local usage-snapshot feature, which has had no caller since it landed in #5307 (backlog#1823)"
-)]
 pub async fn aggregate_local_snapshots(store: Arc<ECStore>) -> Result<(Vec<DiskUsageStatus>, DataUsageInfo), Error> {
     let mut aggregated = DataUsageInfo::default();
     let mut latest_update: Option<SystemTime> = None;
@@ -1772,6 +1767,11 @@ pub async fn record_bucket_object_write_unknown_previous_memory(bucket: &str, ne
     entry.pending_scanner_position = None;
 }
 
+/// Fast in-memory increment for immediate quota consistency.
+pub async fn increment_bucket_usage_memory(bucket: &str, size_increment: u64) {
+    record_bucket_object_write_memory(bucket, None, size_increment).await;
+}
+
 /// Fast in-memory update for successful object deletes.
 pub async fn record_bucket_object_delete_memory(bucket: &str, deleted_size: u64, removed_current_object: bool) {
     ensure_bucket_usage_cached(bucket).await;
@@ -1814,6 +1814,11 @@ pub async fn record_bucket_delete_marker_memory(bucket: &str) {
     entry.pending_scanner_position = None;
 }
 
+/// Fast in-memory decrement for immediate quota consistency
+pub async fn decrement_bucket_usage_memory(bucket: &str, size_decrement: u64) {
+    record_bucket_object_delete_memory(bucket, size_decrement, size_decrement > 0).await;
+}
+
 /// Get bucket usage from the authoritative cache for this topology.
 async fn get_persisted_bucket_usage(bucket: &str) -> Option<u64> {
     let store = runtime_sources::object_store_handle()?;
@@ -2008,6 +2013,91 @@ pub async fn apply_bucket_usage_memory_overlay(data_usage_info: &mut DataUsageIn
     apply_bucket_usage_memory_overlay_if_authoritative(data_usage_info, authoritative).await;
 }
 
+/// Sync memory cache with backend data (called by scanner)
+pub async fn sync_memory_cache_with_backend() -> Result<(), Error> {
+    if let Some(store) = runtime_sources::object_store_handle() {
+        match load_data_usage_from_backend(store.clone()).await {
+            Ok(data_usage_info) => {
+                replace_bucket_usage_memory_from_info(&data_usage_info).await;
+            }
+            Err(e) => {
+                debug!("Failed to sync memory cache with backend: {}", e);
+            }
+        }
+    }
+    Ok(())
+}
+
+/// Create a data usage cache entry from size summary
+pub fn create_cache_entry_from_summary(summary: &SizeSummary) -> DataUsageEntry {
+    let mut entry = DataUsageEntry::default();
+    entry.add_sizes(summary);
+    entry
+}
+
+/// Convert data usage cache to DataUsageInfo
+pub fn cache_to_data_usage_info(
+    cache: &DataUsageCache,
+    path: &str,
+    buckets: &[crate::storage_api_contracts::bucket::BucketInfo],
+) -> DataUsageInfo {
+    let e = match cache.find(path) {
+        Some(e) => e,
+        None => return DataUsageInfo::default(),
+    };
+    let flat = cache.flatten(&e);
+
+    let mut buckets_usage = HashMap::new();
+    for bucket in buckets.iter() {
+        let e = match cache.find(&bucket.name) {
+            Some(e) => e,
+            None => continue,
+        };
+        let flat = cache.flatten(&e);
+        let mut bui = BucketUsageInfo {
+            size: flat.size as u64,
+            versions_count: flat.versions as u64,
+            objects_count: flat.objects as u64,
+            delete_markers_count: flat.delete_markers as u64,
+            object_size_histogram: flat.obj_sizes.to_map(),
+            object_versions_histogram: flat.obj_versions.to_map(),
+            ..Default::default()
+        };
+
+        if let Some(rs) = &flat.replication_stats {
+            bui.replica_size = rs.replica_size;
+            bui.replica_count = rs.replica_count;
+
+            for (arn, stat) in rs.targets.iter() {
+                bui.replication_info.insert(
+                    arn.clone(),
+                    BucketTargetUsageInfo {
+                        replication_pending_size: stat.pending_size,
+                        replicated_size: stat.replicated_size,
+                        replication_failed_size: stat.failed_size,
+                        replication_pending_count: stat.pending_count,
+                        replication_failed_count: stat.failed_count,
+                        replicated_count: stat.replicated_count,
+                        ..Default::default()
+                    },
+                );
+            }
+        }
+        buckets_usage.insert(bucket.name.clone(), bui);
+    }
+
+    DataUsageInfo {
+        last_update: cache.info.last_update,
+        objects_total_count: flat.objects as u64,
+        versions_total_count: flat.versions as u64,
+        delete_markers_total_count: flat.delete_markers as u64,
+        objects_total_size: flat.size as u64,
+        buckets_count: e.children.len() as u64,
+        buckets_usage,
+        ..Default::default()
+    }
+}
+
 // Helper functions for DataUsageCache operations
 pub async fn load_data_usage_cache(store: &crate::set_disk::SetDisks, name: &str) -> crate::error::Result<DataUsageCache> {
     use crate::disk::{BUCKET_META_PREFIX, RUSTFS_META_BUCKET};

crates/ecstore/src/disk/disk_store.rs

@@ -418,6 +418,17 @@ pub struct DiskHealthTracker {
     pub last_capacity_free: AtomicU64,
     /// Last successful capacity probe timestamp
     pub last_capacity_probe_unix_secs: AtomicI64,
+    /// Authoritative atomically published runtime/status pair.
+    state_snapshot: AtomicU64,
+    transition_lock: std::sync::Mutex<()>,
+}
+
+fn pack_health_state(runtime_state: RuntimeDriveHealthState, status: u32) -> u64 {
+    (u64::from(runtime_state as u32) << 32) | u64::from(status)
+}
+
+fn unpack_health_state(snapshot: u64) -> (RuntimeDriveHealthState, u32) {
+    (RuntimeDriveHealthState::from_u32((snapshot >> 32) as u32), snapshot as u32)
 }
 
 #[derive(Debug)]
@@ -730,6 +741,8 @@ impl DiskHealthTracker {
             last_capacity_used: AtomicU64::new(0),
             last_capacity_free: AtomicU64::new(0),
             last_capacity_probe_unix_secs: AtomicI64::new(0),
+            state_snapshot: AtomicU64::new(pack_health_state(RuntimeDriveHealthState::Online, DISK_HEALTH_OK)),
+            transition_lock: std::sync::Mutex::new(()),
         }
     }
 
@@ -766,38 +779,56 @@ impl DiskHealthTracker {
 
     /// Check if disk is faulty
     pub fn is_faulty(&self) -> bool {
-        self.status.load(Ordering::Acquire) == DISK_HEALTH_FAULTY
+        unpack_health_state(self.state_snapshot.load(Ordering::Acquire)).1 == DISK_HEALTH_FAULTY
+    }
+
+    pub fn health_state_snapshot(&self) -> (RuntimeDriveHealthState, bool) {
+        let (runtime_state, status) = unpack_health_state(self.state_snapshot.load(Ordering::Acquire));
+        (runtime_state, status == DISK_HEALTH_FAULTY)
+    }
+
+    fn publish_state(&self, runtime_state: RuntimeDriveHealthState, status: u32) {
+        self.state_snapshot
+            .store(pack_health_state(runtime_state, status), Ordering::Release);
+        self.runtime_state.store(runtime_state as u32, Ordering::Release);
+        self.status.store(status, Ordering::Release);
     }
 
     /// Set disk as faulty
     pub fn set_faulty(&self) {
-        self.status.store(DISK_HEALTH_FAULTY, Ordering::Release);
+        let _guard = self.transition_lock.lock().unwrap_or_else(|poisoned| poisoned.into_inner());
+        self.publish_state(RuntimeDriveHealthState::Offline, DISK_HEALTH_FAULTY);
     }
 
     /// Set disk as OK
     pub fn set_ok(&self) {
-        self.status.store(DISK_HEALTH_OK, Ordering::Release);
+        let _guard = self.transition_lock.lock().unwrap_or_else(|poisoned| poisoned.into_inner());
+        self.publish_state(RuntimeDriveHealthState::Online, DISK_HEALTH_OK);
     }
 
     #[cfg(test)]
     pub fn force_runtime_state_for_test(&self, state: RuntimeDriveHealthState) {
-        self.runtime_state.store(state as u32, Ordering::Release);
-        match state {
-            RuntimeDriveHealthState::Offline => self.set_faulty(),
-            RuntimeDriveHealthState::Online | RuntimeDriveHealthState::Suspect | RuntimeDriveHealthState::Returning => {
-                self.set_ok();
-            }
-        }
+        let _guard = self.transition_lock.lock().unwrap_or_else(|poisoned| poisoned.into_inner());
+        let status = if state == RuntimeDriveHealthState::Offline {
+            DISK_HEALTH_FAULTY
+        } else {
+            DISK_HEALTH_OK
+        };
+        self.publish_state(state, status);
     }
 
     pub fn swap_ok_to_faulty(&self) -> bool {
-        self.status
-            .compare_exchange(DISK_HEALTH_OK, DISK_HEALTH_FAULTY, Ordering::AcqRel, Ordering::Relaxed)
-            .is_ok()
+        let _guard = self.transition_lock.lock().unwrap_or_else(|poisoned| poisoned.into_inner());
+        let (_, status) = unpack_health_state(self.state_snapshot.load(Ordering::Acquire));
+        if status != DISK_HEALTH_OK {
+            return false;
+        }
+        self.publish_state(RuntimeDriveHealthState::Offline, DISK_HEALTH_FAULTY);
+        true
     }
 
     pub fn runtime_state(&self) -> RuntimeDriveHealthState {
-        RuntimeDriveHealthState::from_u32(self.runtime_state.load(Ordering::Acquire))
+        unpack_health_state(self.state_snapshot.load(Ordering::Acquire)).0
     }
 
     pub fn offline_duration(&self) -> Option<Duration> {
@@ -813,6 +844,7 @@ impl DiskHealthTracker {
     }
 
     pub fn mark_failure(&self, endpoint: &Endpoint, reason: &'static str) -> bool {
+        let _guard = self.transition_lock.lock().unwrap_or_else(|poisoned| poisoned.into_inner());
         let current = self.runtime_state();
         let now = current_unix_secs();
         let next = match current {
@@ -841,23 +873,18 @@ impl DiskHealthTracker {
         };
 
         let became_offline = next == RuntimeDriveHealthState::Offline && current != RuntimeDriveHealthState::Offline;
-        if next == RuntimeDriveHealthState::Offline {
-            self.status.store(DISK_HEALTH_FAULTY, Ordering::Release);
-        } else {
-            self.status.store(DISK_HEALTH_OK, Ordering::Release);
-        }
         self.transition_state(endpoint, current, next, reason);
         became_offline
     }
 
     pub fn mark_offline(&self, endpoint: &Endpoint, reason: &'static str) -> bool {
+        let _guard = self.transition_lock.lock().unwrap_or_else(|poisoned| poisoned.into_inner());
         let current = self.runtime_state();
         if current == RuntimeDriveHealthState::Offline {
             return false;
         }
 
         self.consecutive_successes.store(0, Ordering::Release);
-        self.status.store(DISK_HEALTH_FAULTY, Ordering::Release);
         self.transition_state(endpoint, current, RuntimeDriveHealthState::Offline, reason);
         true
     }
@@ -871,11 +898,10 @@ impl DiskHealthTracker {
     }
 
     fn reset_for_store_init_retry_at(&self, endpoint: &Endpoint, now: Duration) {
+        let _guard = self.transition_lock.lock().unwrap_or_else(|poisoned| poisoned.into_inner());
         let now_nanos = unix_nanos(now);
         let now_secs = unix_secs_i64(now);
-        self.status.store(DISK_HEALTH_OK, Ordering::Release);
-        self.runtime_state
-            .store(RuntimeDriveHealthState::Online as u32, Ordering::Release);
+        self.publish_state(RuntimeDriveHealthState::Online, DISK_HEALTH_OK);
         self.consecutive_failures.store(0, Ordering::Release);
         self.consecutive_successes.store(0, Ordering::Release);
         self.offline_since_unix_secs.store(0, Ordering::Release);
@@ -887,6 +913,7 @@ impl DiskHealthTracker {
     }
 
     pub fn mark_recovery_success(&self, endpoint: &Endpoint, reason: &'static str) -> bool {
+        let _guard = self.transition_lock.lock().unwrap_or_else(|poisoned| poisoned.into_inner());
         let current = self.runtime_state();
         let next = match current {
             RuntimeDriveHealthState::Online => RuntimeDriveHealthState::Online,
@@ -907,7 +934,6 @@ impl DiskHealthTracker {
 
         let became_online = next == RuntimeDriveHealthState::Online;
         if became_online {
-            self.status.store(DISK_HEALTH_OK, Ordering::Release);
             self.consecutive_failures.store(0, Ordering::Release);
             self.consecutive_successes.store(0, Ordering::Release);
         }
@@ -937,7 +963,13 @@ impl DiskHealthTracker {
             return;
         }
 
-        self.runtime_state.store(next as u32, Ordering::Release);
+        let current_status = unpack_health_state(self.state_snapshot.load(Ordering::Acquire)).1;
+        let status = match next {
+            RuntimeDriveHealthState::Offline => DISK_HEALTH_FAULTY,
+            RuntimeDriveHealthState::Returning => current_status,
+            RuntimeDriveHealthState::Online | RuntimeDriveHealthState::Suspect => DISK_HEALTH_OK,
+        };
+        self.publish_state(next, status);
         self.last_transition_unix_secs
             .store(current_unix_secs() as i64, Ordering::Release);
 
@@ -1223,7 +1255,7 @@ impl LocalDiskWrapper {
                         return;
                     }
 
-                    if health.status.load(Ordering::Relaxed) != DISK_HEALTH_OK {
+                    if health.is_faulty() {
                         continue;
                     }
 
@@ -2929,6 +2961,35 @@ mod tests {
         });
     }
 
+    #[test]
+    fn concurrent_failure_and_recovery_publish_one_health_snapshot() {
+        let endpoint = Endpoint::try_from("/tmp/concurrent-health-snapshot").expect("endpoint should parse");
+        let health = Arc::new(DiskHealthTracker::new());
+        let workers = (0..8)
+            .map(|_| {
+                let health = Arc::clone(&health);
+                let endpoint = endpoint.clone();
+                std::thread::spawn(move || {
+                    for _ in 0..32 {
+                        health.mark_failure(&endpoint, "concurrent_test");
+                        health.mark_recovery_success(&endpoint, "concurrent_test");
+                        let (runtime, faulty) = health.health_state_snapshot();
+                        assert!(matches!(
+                            (runtime, faulty),
+                            (RuntimeDriveHealthState::Online, false)
+                                | (RuntimeDriveHealthState::Suspect, false)
+                                | (RuntimeDriveHealthState::Offline, true)
+                                | (RuntimeDriveHealthState::Returning, true)
+                        ));
+                    }
+                })
+            })
+            .collect::<Vec<_>>();
+        for worker in workers {
+            worker.join().expect("health transition worker should not panic");
+        }
+    }
+
     #[test]
     fn operation_success_recovers_suspect_drive_without_faulting() {
         let endpoint = Endpoint::try_from("/tmp/runtime-state-suspect-success").expect("endpoint should parse");

crates/ecstore/src/error/mod.rs

@@ -1075,6 +1075,9 @@ pub struct GenericError {
 
 #[derive(Debug, thiserror::Error, PartialEq, Eq)]
 pub enum ObjectApiError {
+    #[error("BackendDown")]
+    BackendDown(String),
+
     #[error("The operation is not valid for the current state of the object {}/{}({})", .0.bucket, .0.object, .0.version_id)]
     InvalidObjectState(GenericError),
 }
@@ -1091,6 +1094,72 @@ pub struct ErrorResponse {
     pub host_id: String,
 }
 
+pub fn error_resp_to_object_err(err: ErrorResponse, params: Vec<&str>) -> std::io::Error {
+    let mut bucket = "";
+    let mut object = "";
+    let mut version_id = "";
+    if !params.is_empty() {
+        bucket = params[0];
+    }
+    if params.len() >= 2 {
+        object = params[1];
+    }
+    if params.len() >= 3 {
+        version_id = params[2];
+    }
+
+    if is_network_or_host_down(&err.to_string(), false) {
+        return std::io::Error::other(ObjectApiError::BackendDown(format!("{err}")));
+    }
+
+    let err_ = std::io::Error::other(err.to_string());
+    let r_err = err;
+    let err;
+    let bucket = bucket.to_string();
+    let object = object.to_string();
+    let version_id = version_id.to_string();
+
+    match r_err.code {
+        S3ErrorCode::BucketNotEmpty => {
+            err = std::io::Error::other(StorageError::BucketNotEmpty("".to_string()).to_string());
+        }
+        S3ErrorCode::InvalidBucketName => {
+            err = std::io::Error::other(StorageError::BucketNameInvalid(bucket));
+        }
+        S3ErrorCode::InvalidPart => {
+            err = std::io::Error::other(StorageError::InvalidPart(0, bucket, object /* , version_id */));
+        }
+        S3ErrorCode::NoSuchBucket => {
+            err = std::io::Error::other(StorageError::BucketNotFound(bucket));
+        }
+        S3ErrorCode::NoSuchKey => {
+            if !object.is_empty() {
+                err = std::io::Error::other(StorageError::ObjectNotFound(bucket, object));
+            } else {
+                err = std::io::Error::other(StorageError::BucketNotFound(bucket));
+            }
+        }
+        S3ErrorCode::NoSuchVersion => {
+            if !object.is_empty() {
+                err = std::io::Error::other(StorageError::ObjectNotFound(bucket, object)); //, version_id);
+            } else {
+                err = std::io::Error::other(StorageError::BucketNotFound(bucket));
+            }
+        }
+        S3ErrorCode::AccessDenied => {
+            err = std::io::Error::other(StorageError::PrefixAccessDenied(bucket, object));
+        }
+        S3ErrorCode::NoSuchUpload => {
+            err = std::io::Error::other(StorageError::InvalidUploadID(bucket, object, version_id));
+        }
+        _ => {
+            err = err_;
+        }
+    }
+
+    err
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;

crates/ecstore/src/event/targetlist.rs

@@ -20,10 +20,6 @@ use std::sync::atomic::AtomicI64;
 /// this type never grew past its counter. `total_events` is read by the
 /// notifier's log line but nothing increments it, so that field reports zero.
 #[derive(Default)]
-#[allow(
-    dead_code,
-    reason = "held only by the dead ecstore EventNotifier; see services/event_notification.rs (backlog#1823)"
-)]
 pub struct TargetList {
     pub total_events: AtomicI64,
 }

crates/ecstore/src/io_support/compress.rs

@@ -31,13 +31,6 @@ pub const ENV_DISK_COMPRESSION_MIME_TYPES: &str = "RUSTFS_COMPRESSION_MIME_TYPES
 // Environment variable for additional extensions to exclude from compression (comma-separated, e.g. ".foo,.bar")
 pub const ENV_ADDED_EXCLUDE_COMPRESS_EXTENSIONS: &str = "RUSTFS_ADDED_EXCLUDE_COMPRESS_EXTENSIONS";
 
-// Environment variable to additionally enable disk compression for multipart uploads.
-// Default off: nodes from before the resumable decompressor fix fail transient reads of
-// compressed objects, so multipart compression stays dark until the operator confirms the
-// fleet has converged on a fixed build.
-// RUSTFS_COMPAT_TODO(multipart-compression-default-off-window): staged rollout switch for restored multipart compression, flipping the default to enabled on retirement. Remove after the minimum supported direct-upgrade release ships the resumable DecompressReader.
-pub const ENV_DISK_COMPRESSION_MULTIPART_ENABLED: &str = "RUSTFS_COMPRESSION_MULTIPART_ENABLED";
-
 pub const DEFAULT_DISK_COMPRESS_EXTENSIONS: &str = ".txt,.log,.csv,.json,.tar,.xml,.bin";
 pub const DEFAULT_DISK_COMPRESS_MIME_TYPES: &str = "text/*,application/json,application/xml,binary/octet-stream";
 
@@ -178,21 +171,6 @@ pub fn is_disk_compression_enabled() -> bool {
     DISK_COMPRESSION_CONFIG.get_or_init(parse_disk_compression_config).enabled
 }
 
-// Parsed once at first use, mirroring DISK_COMPRESSION_CONFIG.
-static MULTIPART_DISK_COMPRESSION_ENABLED: OnceLock<bool> = OnceLock::new();
-
-/// Whether multipart uploads may advertise disk compression. Requires the
-/// regular disk-compression gates to pass as well; this is the staged-rollout
-/// switch that keeps multipart compression dark during rolling upgrades from
-/// builds whose decompressor was not yet resumable.
-pub fn is_multipart_disk_compression_enabled() -> bool {
-    *MULTIPART_DISK_COMPRESSION_ENABLED.get_or_init(|| {
-        env::var(ENV_DISK_COMPRESSION_MULTIPART_ENABLED)
-            .map(|s| matches!(s.to_ascii_lowercase().as_str(), "true" | "on" | "1"))
-            .unwrap_or(false)
-    })
-}
-
 fn is_disk_compressible_with_config(headers: &http::HeaderMap, object_name: &str, config: &DiskCompressionConfig) -> bool {
     // Check if disk compression is enabled (read once at first use, then fixed for process lifetime)
     if !config.enabled {

crates/ecstore/src/object_api/mod.rs

@@ -22,7 +22,7 @@ use crate::bucket::replication::{
 use crate::bucket::versioning::VersioningApi as _;
 use crate::config::storageclass;
 use crate::error::{Error, Result};
-use crate::io_support::rio::{HardLimitReader, HashReader};
+use crate::io_support::rio::{HashReader, LimitReader};
 use crate::storage_api_contracts::{
     lifecycle::{ExpirationOptions, TransitionedObject},
     range::HTTPRangeSpec,

crates/ecstore/src/object_api/readers.rs

@@ -479,15 +479,7 @@ enum ReadTransform {
     },
 }
 
-/// How an object's stored bytes must be fetched and transformed to serve a
-/// request.
-///
-/// Public so callers that fetch the stored bytes from somewhere other than the
-/// local erasure set — the remote-tier read path — can position their own fetch
-/// with [`ReadPlan::storage_offset`] / [`ReadPlan::storage_length`] and then
-/// hand the resulting stream to [`ReadPlan::into_object_reader`], instead of
-/// reimplementing the transform decisions (rustfs/rustfs#6025).
-pub struct ReadPlan {
+struct ReadPlan {
     storage_offset: usize,
     storage_length: i64,
     object_size: i64,
@@ -495,43 +487,6 @@ pub struct ReadPlan {
 }
 
 impl ReadPlan {
-    /// Byte offset into the object's **stored** bytes where the fetch must
-    /// start. Encrypted and compressed objects address their storage in a
-    /// different coordinate system than the plaintext range the caller asked
-    /// for, which is exactly the distinction this plan resolves.
-    pub fn storage_offset(&self) -> usize {
-        self.storage_offset
-    }
-
-    /// Number of **stored** bytes the fetch must deliver, in the same
-    /// coordinate system as [`Self::storage_offset`].
-    pub fn storage_length(&self) -> i64 {
-        self.storage_length
-    }
-
-    /// Build the plan for a request without consuming a stream, so a caller
-    /// that has to issue its own positioned fetch can read the offsets first.
-    pub async fn build_for_request(
-        rs: Option<HTTPRangeSpec>,
-        oi: &ObjectInfo,
-        opts: &ObjectOptions,
-        h: &HeaderMap<HeaderValue>,
-        resolver: Option<&dyn ObjectEncryptionResolver>,
-    ) -> Result<Self> {
-        Self::build_with_resolver(rs, oi, opts, h, resolver).await
-    }
-
-    /// Wrap `reader` — the stored bytes this plan asked for, already positioned
-    /// at [`Self::storage_offset`] — in the transforms that turn them into the
-    /// bytes the caller requested.
-    pub fn into_object_reader(
-        self,
-        reader: Box<dyn AsyncRead + Unpin + Send + Sync>,
-        oi: &ObjectInfo,
-    ) -> Result<GetObjectReader> {
-        self.into_reader(reader, oi).map(|(reader, _, _)| reader)
-    }
-
     #[cfg(test)]
     async fn build(rs: Option<HTTPRangeSpec>, oi: &ObjectInfo, opts: &ObjectOptions, h: &HeaderMap<HeaderValue>) -> Result<Self> {
         Self::build_with_resolver(rs, oi, opts, h, Some(&tests::TEST_RESOLVER)).await
@@ -545,17 +500,8 @@ impl ReadPlan {
         resolver: Option<&dyn ObjectEncryptionResolver>,
     ) -> Result<Self> {
         let mut rs = rs;
-        // A part number addresses the object's PLAINTEXT bytes. A restore read
-        // serves the stored representation instead (see
-        // [`restore_request_active`]), where that synthesized range would be
-        // reinterpreted as a storage range and truncate an encrypted or
-        // compressed payload by exactly its encoding overhead — the copy-back
-        // then fails its length check partway through
-        // (rustfs/rustfs#6025). An explicit caller range is already in storage
-        // coordinates on that path and is still honored.
         if let Some(part_number) = opts.part_number
             && rs.is_none()
-            && !restore_request_active(opts)
         {
             rs = http_range_spec_from_object_info(oi, part_number);
         }
@@ -808,7 +754,7 @@ impl ReadPlan {
                         }
                     }
                 } else {
-                    Box::new(HardLimitReader::new(dec_reader, decompressed_length))
+                    Box::new(LimitReader::new(dec_reader, total_plaintext_size))
                 };
 
                 let mut object_info = oi.clone();
@@ -900,7 +846,7 @@ impl ReadPlan {
                             )?;
                             Box::new(ranged_reader)
                         } else {
-                            Box::new(HardLimitReader::new(decompressed_reader, total_plaintext_size_i64))
+                            Box::new(LimitReader::new(decompressed_reader, total_plaintext_size))
                         }
                     } else if plaintext_offset > 0 || plaintext_length != total_plaintext_size_i64 {
                         Box::new(RangedDecompressReader::new(
@@ -910,7 +856,7 @@ impl ReadPlan {
                             total_plaintext_size,
                         )?)
                     } else {
-                        Box::new(HardLimitReader::new(decrypted_reader, total_plaintext_size_i64))
+                        Box::new(LimitReader::new(decrypted_reader, total_plaintext_size))
                     };
 
                 let mut object_info = oi.clone();
@@ -1781,423 +1727,6 @@ mod tests {
         assert_eq!(actual, b"fghijkl");
     }
 
-    /// Compresses one multipart part exactly like the write path does
-    /// (`WritePlan::with_compression` wraps each part in its own
-    /// `compression_reader`), returning the on-disk bytes and the storage-format
-    /// compression index.
-    async fn compressed_part_fixture(data: &[u8]) -> (Vec<u8>, Option<Bytes>) {
-        use crate::io_support::rio::TryGetIndex as _;
-        let mut compressor =
-            crate::io_support::rio::compression_reader(Cursor::new(data.to_vec()), CompressionAlgorithm::default(), false);
-        let mut compressed = Vec::new();
-        compressor.read_to_end(&mut compressed).await.expect("compress part stream");
-        let index = compressor
-            .try_get_index()
-            .map(crate::io_support::rio::compression_index_storage_bytes);
-        (compressed, index)
-    }
-
-    struct CompressedMultipartFixture {
-        object_info: ObjectInfo,
-        stored: Vec<u8>,
-        plaintext: Vec<u8>,
-    }
-
-    /// Builds the on-disk representation of a compressed multipart object: each
-    /// part is an independent compressed stream and the storage layer serves
-    /// their concatenation.
-    async fn compressed_multipart_fixture(part_sizes: &[usize]) -> CompressedMultipartFixture {
-        let pattern = b"compressed multipart read path fixture data ";
-        let mut plaintext = Vec::new();
-        let mut stored = Vec::new();
-        let mut parts = Vec::with_capacity(part_sizes.len());
-
-        for (i, part_size) in part_sizes.iter().enumerate() {
-            let mut part_plaintext = Vec::with_capacity(*part_size);
-            while part_plaintext.len() < *part_size {
-                part_plaintext.extend_from_slice(pattern);
-                part_plaintext.push(i as u8);
-            }
-            part_plaintext.truncate(*part_size);
-
-            let (compressed, index) = compressed_part_fixture(&part_plaintext).await;
-            parts.push(ObjectPartInfo {
-                number: i + 1,
-                size: compressed.len(),
-                actual_size: *part_size as i64,
-                index,
-                ..Default::default()
-            });
-            stored.extend_from_slice(&compressed);
-            plaintext.extend_from_slice(&part_plaintext);
-        }
-
-        let mut user_defined = HashMap::new();
-        rustfs_utils::http::insert_str(
-            &mut user_defined,
-            rustfs_utils::http::SUFFIX_COMPRESSION,
-            crate::io_support::rio::compression_metadata_value(CompressionAlgorithm::default()),
-        );
-        rustfs_utils::http::insert_str(&mut user_defined, rustfs_utils::http::SUFFIX_ACTUAL_SIZE, plaintext.len().to_string());
-
-        let object_info = ObjectInfo {
-            bucket: "test-bucket".to_string(),
-            name: "compressed-multipart".to_string(),
-            size: stored.len() as i64,
-            etag: Some(format!("6bcf86bed8807b8e78f0fc6e0a53079d-{}", part_sizes.len())),
-            parts: Arc::new(parts),
-            user_defined: Arc::new(user_defined),
-            ..Default::default()
-        };
-
-        CompressedMultipartFixture {
-            object_info,
-            stored,
-            plaintext,
-        }
-    }
-
-    /// Plans the read once to learn the storage window, then serves exactly that
-    /// window — mirroring how `set_disk` feeds the erasure read into the
-    /// returned reader.
-    async fn read_compressed_multipart(
-        fixture: &CompressedMultipartFixture,
-        rs: Option<HTTPRangeSpec>,
-        opts: &ObjectOptions,
-    ) -> Vec<u8> {
-        let headers = HeaderMap::new();
-        let (_, offset, length) =
-            GetObjectReader::new(Box::new(Cursor::new(Vec::new())), rs.clone(), &fixture.object_info, opts, &headers)
-                .await
-                .expect("plan compressed multipart read");
-
-        let end = offset + usize::try_from(length).expect("storage window length must be non-negative");
-        assert!(
-            end <= fixture.stored.len(),
-            "planned storage window {offset}..{end} exceeds stored stream of {} bytes",
-            fixture.stored.len()
-        );
-        let window = fixture.stored[offset..end].to_vec();
-
-        let (mut reader, replay_offset, replay_length) =
-            GetObjectReader::new(Box::new(Cursor::new(window)), rs, &fixture.object_info, opts, &headers)
-                .await
-                .expect("build compressed multipart reader");
-        assert_eq!((replay_offset, replay_length), (offset, length), "read plan must be deterministic");
-
-        reader.read_all().await.expect("read compressed multipart stream")
-    }
-
-    /// Byte pattern with a 2 KiB period: it compresses extremely well while
-    /// looking nothing like ASCII fixtures. Mirrors the e2e generator that
-    /// exposed a truncated full GET on high-ratio multipart payloads.
-    fn high_ratio_binary_payload(size: usize, seed: u8) -> Vec<u8> {
-        (0..size)
-            .map(|i| ((i as u64).wrapping_mul(2_654_435_761).wrapping_add(seed as u64) >> 3) as u8)
-            .collect()
-    }
-
-    #[tokio::test]
-    async fn compressed_multipart_full_get_handles_high_ratio_binary_payload() {
-        let part_sizes = [5 * 1024 * 1024_usize, 1024 * 1024];
-        let mut plaintext = Vec::new();
-        let mut stored = Vec::new();
-        let mut parts = Vec::with_capacity(part_sizes.len());
-
-        for (i, part_size) in part_sizes.iter().enumerate() {
-            let part_plaintext = high_ratio_binary_payload(*part_size, if i == 0 { 7 } else { 61 });
-            let (compressed, index) = compressed_part_fixture(&part_plaintext).await;
-            parts.push(ObjectPartInfo {
-                number: i + 1,
-                size: compressed.len(),
-                actual_size: *part_size as i64,
-                index,
-                ..Default::default()
-            });
-            stored.extend_from_slice(&compressed);
-            plaintext.extend_from_slice(&part_plaintext);
-        }
-
-        let mut user_defined = HashMap::new();
-        rustfs_utils::http::insert_str(
-            &mut user_defined,
-            rustfs_utils::http::SUFFIX_COMPRESSION,
-            crate::io_support::rio::compression_metadata_value(CompressionAlgorithm::default()),
-        );
-        rustfs_utils::http::insert_str(&mut user_defined, rustfs_utils::http::SUFFIX_ACTUAL_SIZE, plaintext.len().to_string());
-        let fixture = CompressedMultipartFixture {
-            object_info: ObjectInfo {
-                bucket: "test-bucket".to_string(),
-                name: "high-ratio-multipart".to_string(),
-                size: stored.len() as i64,
-                etag: Some("6bcf86bed8807b8e78f0fc6e0a53079d-2".to_string()),
-                parts: Arc::new(parts),
-                user_defined: Arc::new(user_defined),
-                ..Default::default()
-            },
-            stored,
-            plaintext,
-        };
-
-        let read = read_compressed_multipart(&fixture, None, &ObjectOptions::default()).await;
-
-        assert_eq!(read.len(), fixture.plaintext.len(), "full GET must return the logical size");
-        assert_eq!(read, fixture.plaintext, "high-ratio multipart payload must survive the roundtrip");
-    }
-
-    /// Full GET over a compressed multipart object must decode across part
-    /// boundaries: every part is an independent compressed stream (this is also
-    /// the on-disk shape written by builds before rustfs/rustfs#5169 disabled
-    /// multipart compression, so this pins legacy-object readability).
-    #[tokio::test]
-    async fn compressed_multipart_full_get_decodes_across_part_boundaries() {
-        let fixture = compressed_multipart_fixture(&[3 * 1024 * 1024, 2 * 1024 * 1024, 512 * 1024]).await;
-
-        let read = read_compressed_multipart(&fixture, None, &ObjectOptions::default()).await;
-
-        assert_eq!(read.len(), fixture.plaintext.len(), "full GET must return the logical size");
-        assert_eq!(read, fixture.plaintext, "full GET must reassemble all parts");
-    }
-
-    #[tokio::test]
-    async fn compressed_multipart_range_get_crosses_part_boundary() {
-        let fixture = compressed_multipart_fixture(&[3 * 1024 * 1024, 2 * 1024 * 1024]).await;
-        let boundary = 3 * 1024 * 1024_i64;
-        let rs = HTTPRangeSpec {
-            is_suffix_length: false,
-            start: boundary - 100_000,
-            end: boundary + 100_000 - 1,
-        };
-
-        let read = read_compressed_multipart(&fixture, Some(rs), &ObjectOptions::default()).await;
-
-        let expected = &fixture.plaintext[(boundary - 100_000) as usize..(boundary + 100_000) as usize];
-        assert_eq!(read, expected, "boundary-crossing range must splice both parts");
-    }
-
-    #[tokio::test]
-    async fn compressed_multipart_range_get_seeks_into_later_part() {
-        let fixture = compressed_multipart_fixture(&[3 * 1024 * 1024, 4 * 1024 * 1024]).await;
-        // Deep inside part 2 so the plan skips part 1 entirely and (when the
-        // part carries an index) seeks within part 2.
-        let start = 3 * 1024 * 1024_i64 + 2 * 1024 * 1024_i64 + 137;
-        let rs = HTTPRangeSpec {
-            is_suffix_length: false,
-            start,
-            end: start + 64 * 1024 - 1,
-        };
-
-        let read = read_compressed_multipart(&fixture, Some(rs), &ObjectOptions::default()).await;
-
-        let expected = &fixture.plaintext[start as usize..(start + 64 * 1024) as usize];
-        assert_eq!(read, expected, "range inside a later part must decode from that part");
-    }
-
-    /// Parts written without a compression index (small parts skip the index in
-    /// the rio-v2 backend) must still be rangeable: the plan starts at the part
-    /// boundary and skips decompressed bytes.
-    #[tokio::test]
-    async fn compressed_multipart_range_get_works_without_part_indexes() {
-        let mut fixture = compressed_multipart_fixture(&[1024 * 1024, 1024 * 1024]).await;
-        let parts = fixture
-            .object_info
-            .parts
-            .iter()
-            .map(|part| ObjectPartInfo {
-                index: None,
-                ..part.clone()
-            })
-            .collect::<Vec<_>>();
-        fixture.object_info.parts = Arc::new(parts);
-
-        let start = 1024 * 1024_i64 + 4096;
-        let rs = HTTPRangeSpec {
-            is_suffix_length: false,
-            start,
-            end: start + 32 * 1024 - 1,
-        };
-
-        let read = read_compressed_multipart(&fixture, Some(rs), &ObjectOptions::default()).await;
-
-        let expected = &fixture.plaintext[start as usize..(start + 32 * 1024) as usize];
-        assert_eq!(read, expected, "index-less parts must fall back to part-boundary skip");
-    }
-
-    #[tokio::test]
-    async fn compressed_multipart_part_number_get_returns_single_part() {
-        let part_sizes = [3 * 1024 * 1024, 2 * 1024 * 1024, 512 * 1024];
-        let fixture = compressed_multipart_fixture(&part_sizes).await;
-
-        let mut logical_offset = 0_usize;
-        for (i, part_size) in part_sizes.iter().enumerate() {
-            let opts = ObjectOptions {
-                part_number: Some(i + 1),
-                ..Default::default()
-            };
-
-            let read = read_compressed_multipart(&fixture, None, &opts).await;
-
-            let expected = &fixture.plaintext[logical_offset..logical_offset + part_size];
-            assert_eq!(read.len(), *part_size, "partNumber={} GET must return the part's logical size", i + 1);
-            assert_eq!(read, expected, "partNumber={} GET must return the original part bytes", i + 1);
-            logical_offset += part_size;
-        }
-    }
-
-    #[tokio::test]
-    async fn compressed_multipart_suffix_range_reads_tail() {
-        let fixture = compressed_multipart_fixture(&[3 * 1024 * 1024, 1024 * 1024]).await;
-        let suffix_len = 128 * 1024_i64;
-        let rs = HTTPRangeSpec {
-            is_suffix_length: true,
-            start: suffix_len,
-            end: -1,
-        };
-
-        let read = read_compressed_multipart(&fixture, Some(rs), &ObjectOptions::default()).await;
-
-        let expected = &fixture.plaintext[fixture.plaintext.len() - suffix_len as usize..];
-        assert_eq!(read, expected, "suffix range must return the tail of the last part");
-    }
-
-    /// Builds an SSE-C + disk-compression multipart object exactly like the
-    /// write path: each part is compressed into its own stream and then
-    /// encrypted with the per-part key schedule. The fixture is
-    /// legacy-encryption-specific (`rustfs_rio::EncryptReader`), matching the
-    /// pre-existing `build_legacy_ssec_multipart_fixture` shape, while the
-    /// compression layer follows the active backend feature.
-    async fn compressed_encrypted_multipart_fixture(key_bytes: [u8; 32], part_sizes: &[usize]) -> CompressedMultipartFixture {
-        let pattern = b"compressed encrypted multipart fixture data ";
-        let mut plaintext = Vec::new();
-        let mut stored = Vec::new();
-        let mut parts = Vec::with_capacity(part_sizes.len());
-
-        for (i, part_size) in part_sizes.iter().enumerate() {
-            let part_number = i + 1;
-            let mut part_plaintext = Vec::with_capacity(*part_size);
-            while part_plaintext.len() < *part_size {
-                part_plaintext.extend_from_slice(pattern);
-                part_plaintext.push(part_number as u8);
-            }
-            part_plaintext.truncate(*part_size);
-
-            let (compressed, index) = compressed_part_fixture(&part_plaintext).await;
-            let mut part_cipher = Vec::new();
-            rustfs_rio::EncryptReader::new_multipart(Cursor::new(compressed), key_bytes, LEGACY_FIXTURE_BASE_NONCE, part_number)
-                .read_to_end(&mut part_cipher)
-                .await
-                .expect("encrypt compressed fixture part");
-
-            parts.push(ObjectPartInfo {
-                number: part_number,
-                size: part_cipher.len(),
-                actual_size: *part_size as i64,
-                index,
-                ..Default::default()
-            });
-            stored.extend_from_slice(&part_cipher);
-            plaintext.extend_from_slice(&part_plaintext);
-        }
-
-        let mut user_defined = legacy_ssec_multipart_metadata(key_bytes, plaintext.len());
-        rustfs_utils::http::insert_str(
-            &mut user_defined,
-            rustfs_utils::http::SUFFIX_COMPRESSION,
-            crate::io_support::rio::compression_metadata_value(CompressionAlgorithm::default()),
-        );
-        rustfs_utils::http::insert_str(&mut user_defined, rustfs_utils::http::SUFFIX_ACTUAL_SIZE, plaintext.len().to_string());
-
-        let object_info = ObjectInfo {
-            bucket: "test-bucket".to_string(),
-            name: "compressed-encrypted-multipart".to_string(),
-            size: stored.len() as i64,
-            etag: Some(format!("6bcf86bed8807b8e78f0fc6e0a53079d-{}", part_sizes.len())),
-            parts: Arc::new(parts),
-            user_defined: Arc::new(user_defined),
-            ..Default::default()
-        };
-
-        CompressedMultipartFixture {
-            object_info,
-            stored,
-            plaintext,
-        }
-    }
-
-    async fn read_compressed_encrypted_multipart(
-        fixture: &CompressedMultipartFixture,
-        key_bytes: [u8; 32],
-        rs: Option<HTTPRangeSpec>,
-        opts: &ObjectOptions,
-    ) -> Vec<u8> {
-        let headers = ssec_headers_from_key(key_bytes);
-        let (_, offset, length) =
-            GetObjectReader::new(Box::new(Cursor::new(Vec::new())), rs.clone(), &fixture.object_info, opts, &headers)
-                .await
-                .expect("plan compressed encrypted multipart read");
-
-        let end = offset + usize::try_from(length).expect("storage window length must be non-negative");
-        assert!(
-            end <= fixture.stored.len(),
-            "planned storage window {offset}..{end} exceeds stored stream of {} bytes",
-            fixture.stored.len()
-        );
-        let window = fixture.stored[offset..end].to_vec();
-
-        let (mut reader, replay_offset, replay_length) =
-            GetObjectReader::new(Box::new(Cursor::new(window)), rs, &fixture.object_info, opts, &headers)
-                .await
-                .expect("build compressed encrypted multipart reader");
-        assert_eq!((replay_offset, replay_length), (offset, length), "read plan must be deterministic");
-
-        reader.read_all().await.expect("read compressed encrypted multipart stream")
-    }
-
-    #[tokio::test]
-    async fn compressed_encrypted_multipart_full_get_roundtrip() {
-        let key_bytes = [0x6Eu8; 32];
-        let fixture = compressed_encrypted_multipart_fixture(key_bytes, &[3 * 1024 * 1024, 1024 * 1024]).await;
-
-        let read = read_compressed_encrypted_multipart(&fixture, key_bytes, None, &ObjectOptions::default()).await;
-
-        assert_eq!(read.len(), fixture.plaintext.len(), "full GET must return the logical size");
-        assert_eq!(read, fixture.plaintext, "SSE-C + compression full GET must reassemble all parts");
-    }
-
-    #[tokio::test]
-    async fn compressed_encrypted_multipart_range_crosses_part_boundary() {
-        let key_bytes = [0x6Eu8; 32];
-        let fixture = compressed_encrypted_multipart_fixture(key_bytes, &[3 * 1024 * 1024, 1024 * 1024]).await;
-        let boundary = 3 * 1024 * 1024_i64;
-        let rs = HTTPRangeSpec {
-            is_suffix_length: false,
-            start: boundary - 65_536,
-            end: boundary + 65_536 - 1,
-        };
-
-        let read = read_compressed_encrypted_multipart(&fixture, key_bytes, Some(rs), &ObjectOptions::default()).await;
-
-        let expected = &fixture.plaintext[(boundary - 65_536) as usize..(boundary + 65_536) as usize];
-        assert_eq!(read, expected, "SSE-C + compression boundary-crossing range must splice both parts");
-    }
-
-    #[tokio::test]
-    async fn compressed_encrypted_multipart_part_number_get_returns_single_part() {
-        let key_bytes = [0x6Eu8; 32];
-        let part_sizes = [3 * 1024 * 1024, 1024 * 1024];
-        let fixture = compressed_encrypted_multipart_fixture(key_bytes, &part_sizes).await;
-
-        let opts = ObjectOptions {
-            part_number: Some(2),
-            ..Default::default()
-        };
-        let read = read_compressed_encrypted_multipart(&fixture, key_bytes, None, &opts).await;
-
-        let expected = &fixture.plaintext[part_sizes[0]..];
-        assert_eq!(read.len(), part_sizes[1], "partNumber=2 GET must return the part's logical size");
-        assert_eq!(read, expected, "partNumber=2 GET must return the original part bytes");
-    }
-
     #[tokio::test]
     async fn test_get_object_reader_rejects_ssec_read_without_headers() {
         let object_info = ObjectInfo {

crates/ecstore/src/services/batch_processor.rs

@@ -23,10 +23,6 @@ use std::sync::{Arc, Mutex};
 use std::time::{Duration, Instant};
 use tokio::task::JoinSet;
 
-#[allow(
-    dead_code,
-    reason = "default operation label for the test-only AsyncBatchProcessor::new (backlog#1823)"
-)]
 const BATCH_PROCESSOR_OPERATION_CUSTOM: &str = "custom";
 const BATCH_PROCESSOR_OPERATION_READ: &str = "read";
 const BATCH_PROCESSOR_OPERATION_WRITE: &str = "write";
@@ -215,7 +211,6 @@ pub struct AsyncBatchProcessor {
 }
 
 impl AsyncBatchProcessor {
-    #[allow(dead_code, reason = "constructor used only by this file's tests (backlog#1823)")]
     pub fn new(max_concurrent: usize) -> Self {
         Self::new_with_operation(max_concurrent, BATCH_PROCESSOR_OPERATION_CUSTOM)
     }

crates/ecstore/src/services/event_notification.rs

@@ -26,26 +26,11 @@ use std::sync::atomic::Ordering;
 use tokio::sync::RwLock;
 use tracing::warn;
 
-/// Dead ecstore-side notification skeleton.
-///
-/// The working notification stack is `rustfs-notify`, whose own `EventNotifier`
-/// is the one bucket configuration actually drives. Nothing calls the methods
-/// below; `init_bucket_targets` even logs that it is a no-op in this build.
-/// Removing it means also retiring the `InstanceContext` slot that holds it
-/// (backlog#939 Phase 5), so it is left explicit here rather than half-removed.
-#[allow(
-    dead_code,
-    reason = "ecstore-side notification skeleton superseded by rustfs-notify; see module note (backlog#1823)"
-)]
 pub struct EventNotifier {
     target_list: TargetList,
     //bucket_rules_map: HashMap<String , HashMap<EventName, Rules>>,
 }
 
-#[allow(
-    dead_code,
-    reason = "ecstore-side notification skeleton superseded by rustfs-notify; see module note (backlog#1823)"
-)]
 impl EventNotifier {
     pub fn new() -> Arc<RwLock<Self>> {
         Arc::new(RwLock::new(Self {

crates/ecstore/src/services/mod.rs

@@ -13,6 +13,7 @@
 // limitations under the License.
 
 // #730: background service owners still contain staged notification/rebalance/tier paths.
+#![allow(dead_code)]
 
 pub(crate) mod batch_processor;
 pub(crate) mod event_notification;

crates/ecstore/src/services/notification_sys.rs

@@ -1623,7 +1623,6 @@ impl NotificationSys {
         workers.peers.remove(host);
     }
 
-    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
     fn tier_config_reload_worker_active(&self, host: &str) -> bool {
         self.tier_config_reload_workers
             .lock()
@@ -1797,7 +1796,6 @@ where
         .map_err(|_| Error::other(format!("scanner activity peer {host} timed out after {timeout_duration:?}")))?
 }
 
-#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 async fn call_peer_with_timeout<F, Fut>(
     timeout_dur: Duration,
     host_label: &str,

crates/ecstore/src/services/rebalance/meta.rs

@@ -864,10 +864,6 @@ pub(super) fn merge_rebalance_meta(remote: &mut RebalanceMeta, local: &Rebalance
     RebalanceMetaMergeOutcome::Merged
 }
 
-#[allow(
-    dead_code,
-    reason = "stop-transition helper retained beside stop_rebalance_meta_snapshot; no caller yet (backlog#1823)"
-)]
 pub(super) fn mark_started_rebalance_pools_stopped(meta: &mut RebalanceMeta, stop_time: OffsetDateTime) {
     for pool_stat in meta.pool_stats.iter_mut() {
         if pool_stat.info.status == RebalStatus::Started {
@@ -968,7 +964,6 @@ pub(super) fn rollback_rebalance_start_meta_snapshot_for_id(
     })
 }
 
-#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 pub(super) fn stop_rebalance_meta_snapshot(meta: Option<&mut RebalanceMeta>, now: OffsetDateTime) -> Option<RebalanceMeta> {
     let meta = meta?;
     stop_rebalance_state(meta, now);

crates/ecstore/src/services/rebalance/migration.rs

@@ -171,7 +171,6 @@ where
 }
 
 #[allow(clippy::too_many_arguments)]
-#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 pub(super) async fn migrate_entry_version_with_retry_wait<Backend, F, Fut, D, DFut, W, WFut>(
     set: &Backend,
     bucket: String,

crates/ecstore/src/services/rebalance/types.rs

@@ -1,4 +1,5 @@
 use serde::{Deserialize, Serialize};
+use std::sync::Arc;
 use time::OffsetDateTime;
 use tokio_util::sync::CancellationToken;
 
@@ -31,6 +32,8 @@ pub struct RebalanceStats {
     pub cleanup_warnings: RebalanceCleanupWarnings,
 }
 
+pub type RStats = Vec<Arc<RebalanceStats>>;
+
 #[derive(Debug, Default)]
 pub(super) struct RebalanceBucketConfigs {
     pub(super) bucket_incarnation_id: Option<uuid::Uuid>,

crates/ecstore/src/services/tier/mod.rs

@@ -30,5 +30,6 @@ pub mod warm_backend_minio;
 pub mod warm_backend_r2;
 pub mod warm_backend_rustfs;
 pub mod warm_backend_s3;
+pub mod warm_backend_s3sdk;
 pub mod warm_backend_tencent;
 pub mod warm_backend_wasabi;

crates/ecstore/src/services/tier/tier.rs

@@ -488,7 +488,6 @@ impl TierCandidateMutation {
         targets
     }
 
-    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
     fn affected_targets(
         &self,
         manager: &TierConfigMgr,
@@ -803,7 +802,6 @@ fn tier_persisted_reference_blocks_any_target(
         .any(|target| tier_persisted_reference_blocks_target(tier_name, backend_identity, target))
 }
 
-#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 fn tier_object_blocks_target_rebind(object: &ObjectInfo, target: &TierMutationIntentTarget) -> io::Result<bool> {
     tier_object_blocks_any_target_rebind(object, std::slice::from_ref(target))
 }
@@ -2728,6 +2726,14 @@ impl TierConfigMgr {
         Self::publish_candidate_owned(handle, candidate, driver_tier.map(str::to_string), update).await
     }
 
+    fn begin_publish_transition(
+        handle: &Arc<RwLock<Self>>,
+        manager: &mut Self,
+        candidate: &Self,
+    ) -> std::result::Result<TierPublishTransition, AdminError> {
+        Self::begin_publish_transition_with_allowed_mutation_blocks(handle, manager, candidate, None)
+    }
+
     fn begin_publish_transition_with_allowed_mutation_blocks(
         handle: &Arc<RwLock<Self>>,
         manager: &mut Self,
@@ -2813,6 +2819,14 @@ impl TierConfigMgr {
         })
     }
 
+    async fn publish_candidate_inner(
+        handle: &Arc<RwLock<Self>>,
+        candidate: Self,
+        driver_tier: Option<&str>,
+    ) -> std::result::Result<(), AdminError> {
+        Self::publish_candidate_inner_with_allowed_mutation_blocks(handle, candidate, driver_tier, None).await
+    }
+
     async fn publish_candidate_inner_with_allowed_mutation_blocks(
         handle: &Arc<RwLock<Self>>,
         candidate: Self,
@@ -2925,7 +2939,6 @@ impl TierConfigMgr {
         admin_err
     }
 
-    #[allow(dead_code, reason = "reached only through #[cfg(test)] helpers in this file (backlog#1823)")]
     async fn publish_candidate_owned(
         handle: &Arc<RwLock<Self>>,
         candidate: Self,
@@ -3528,7 +3541,6 @@ impl TierConfigMgr {
         Self::update_candidate_with_config_lock(handle, api, TierCandidateMutation::Remove(tier_name.to_string(), force)).await
     }
 
-    #[allow(dead_code, reason = "reached only through #[cfg(test)] helpers in this file (backlog#1823)")]
     async fn remove_and_save_with<S>(
         handle: &Arc<RwLock<Self>>,
         api: Arc<S>,
@@ -3562,7 +3574,6 @@ impl TierConfigMgr {
         Self::update_candidate_with_config_lock(handle, api, TierCandidateMutation::Clear(force)).await
     }
 
-    #[allow(dead_code, reason = "reached only through #[cfg(test)] helpers in this file (backlog#1823)")]
     async fn clear_and_save_with<S>(
         handle: &Arc<RwLock<Self>>,
         api: Arc<S>,
@@ -3601,10 +3612,6 @@ impl TierConfigMgr {
     }
 
     #[cfg(test)]
-    #[allow(
-        dead_code,
-        reason = "lease accounting asserted by a bucket_lifecycle_ops test behind `--features test-util` (backlog#1823)"
-    )]
     pub(crate) async fn active_operation_lease_count(handle: &Arc<RwLock<Self>>, tier_name: &str) -> usize {
         let manager = handle.read().await;
         let Some(runtime) = registered_tier_driver_runtime(&manager) else {
@@ -3710,6 +3717,10 @@ impl TierConfigMgr {
         Ok(())
     }
 
+    fn retire_driver(&mut self, tier_name: &str) {
+        self.revoke_driver(tier_name);
+    }
+
     fn revoke_all_drivers(&mut self) {
         if let Some(runtime) = registered_tier_driver_runtime(self) {
             let mut runtime = lock_unpoisoned(&runtime);
@@ -3873,7 +3884,6 @@ impl TierConfigMgr {
         self.save_config(api, &config_file, data).await
     }
 
-    #[allow(dead_code, reason = "reached only through #[cfg(test)] helpers in this file (backlog#1823)")]
     async fn save_tiering_config_if_current<S>(
         &self,
         api: Arc<S>,

crates/ecstore/src/services/tier/tier_mutation_intent.rs

@@ -305,10 +305,6 @@ impl TierMutationIntent {
     }
 }
 
-#[allow(
-    dead_code,
-    reason = "intent-record persistence asserted by store::init tests (backlog#1823)"
-)]
 pub(crate) fn tier_mutation_intent_record_object_name(mutation_id: Uuid) -> Result<String> {
     tier_mutation_intent_record_object_name_with_prefix(TIER_MUTATION_INTENT_RECORD_PREFIX, mutation_id)
 }
@@ -321,10 +317,6 @@ fn tier_mutation_intent_record_object_name_with_prefix(prefix: &str, mutation_id
     Ok(format!("{}/{}/{}/{}.json", prefix, &mutation_key[..2], &mutation_key[2..4], mutation_key))
 }
 
-#[allow(
-    dead_code,
-    reason = "intent-record persistence asserted by store::init tests (backlog#1823)"
-)]
 pub(crate) fn tier_mutation_intent_id_from_record_object_name(object: &str) -> Result<Uuid> {
     tier_mutation_intent_id_from_record_object_name_with_prefix(TIER_MUTATION_INTENT_RECORD_PREFIX, object)
 }
@@ -363,10 +355,6 @@ fn tier_mutation_intent_id_from_record_object_name_with_prefix(prefix: &str, obj
     Uuid::parse_str(mutation_key).map_err(|_| TierMutationIntentError::Corrupt("intent record path has invalid uuid"))
 }
 
-#[allow(
-    dead_code,
-    reason = "intent-record persistence asserted by store::init tests (backlog#1823)"
-)]
 pub(crate) async fn save_tier_mutation_intent_record<S>(api: Arc<S>, intent: &TierMutationIntent) -> EcstoreResult<()>
 where
     S: EcstoreObjectIO,
@@ -458,10 +446,6 @@ where
     Ok((intent, etag))
 }
 
-#[allow(
-    dead_code,
-    reason = "intent-record persistence asserted by store::init tests (backlog#1823)"
-)]
 pub(crate) async fn save_tier_mutation_intent_record_if_current<S>(
     api: Arc<S>,
     intent: &TierMutationIntent,

crates/ecstore/src/services/tier/warm_backend_gcs.rs

@@ -41,7 +41,10 @@ use crate::services::tier::{
 };
 use tracing::warn;
 
+const MAX_MULTIPART_PUT_OBJECT_SIZE: i64 = 1024 * 1024 * 1024 * 1024 * 5;
+const MAX_PARTS_COUNT: i64 = 10000;
 const _MAX_PART_SIZE: i64 = 1024 * 1024 * 1024 * 5;
+const MIN_PART_SIZE: i64 = 1024 * 1024 * 128;
 
 fn parse_generation(remote_version: &str) -> Result<Option<i64>, Error> {
     if remote_version.is_empty() {
@@ -61,6 +64,7 @@ pub struct WarmBackendGCS {
     pub control: Arc<StorageControl>,
     pub bucket: String,
     pub prefix: String,
+    pub storage_class: String,
 }
 
 impl WarmBackendGCS {
@@ -100,6 +104,7 @@ impl WarmBackendGCS {
             control,
             bucket: conf.bucket.clone(),
             prefix: conf.prefix.strip_suffix("/").unwrap_or(&conf.prefix).to_owned(),
+            storage_class: "".to_string(),
         })
     }
 

crates/ecstore/src/services/tier/warm_backend_s3.rs

@@ -33,6 +33,8 @@ use crate::client::{
     transition_api::{BucketLookupType, Options, TransitionClient, TransitionCore},
     transition_api::{ReadCloser, ReaderImpl},
 };
+use crate::error::ErrorResponse;
+use crate::error::error_resp_to_object_err;
 use crate::services::tier::{
     tier_config::TierS3,
     warm_backend::{

crates/ecstore/src/services/tier/warm_backend_s3sdk.rs

@@ -0,0 +1,200 @@
+// Copyright 2024 RustFS Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+#![allow(unused_imports)]
+#![allow(unused_variables)]
+#![allow(unused_mut)]
+#![allow(unused_assignments)]
+#![allow(unused_must_use)]
+#![allow(clippy::all)]
+
+use std::collections::HashMap;
+use std::sync::Arc;
+use url::Url;
+
+use aws_config::meta::region::RegionProviderChain;
+use aws_sdk_s3::Client;
+use aws_sdk_s3::config::{Credentials, Region};
+use aws_sdk_s3::primitives::ByteStream;
+
+use crate::client::{
+    api_get_options::GetObjectOptions,
+    api_put_object::PutObjectOptions,
+    api_remove::RemoveObjectOptions,
+    transition_api::{ReadCloser, ReaderImpl},
+};
+use crate::error::ErrorResponse;
+use crate::error::error_resp_to_object_err;
+use crate::services::tier::{
+    tier_config::TierS3,
+    warm_backend::{WarmBackend, WarmBackendGetOpts},
+};
+
+pub struct WarmBackendS3 {
+    pub client: Arc<Client>,
+    pub bucket: String,
+    pub prefix: String,
+    pub storage_class: String,
+}
+
+impl WarmBackendS3 {
+    pub async fn new(conf: &TierS3, tier: &str) -> Result<Self, std::io::Error> {
+        let u = match Url::parse(&conf.endpoint) {
+            Ok(u) => u,
+            Err(err) => {
+                return Err(std::io::Error::other(err.to_string()));
+            }
+        };
+
+        if conf.aws_role_web_identity_token_file == "" && conf.aws_role_arn != ""
+            || conf.aws_role_web_identity_token_file != "" && conf.aws_role_arn == ""
+        {
+            return Err(std::io::Error::other("both the token file and the role ARN are required"));
+        } else if conf.access_key == "" && conf.secret_key != "" || conf.access_key != "" && conf.secret_key == "" {
+            return Err(std::io::Error::other("both the access and secret keys are required"));
+        } else if conf.aws_role
+            && (conf.aws_role_web_identity_token_file != ""
+                || conf.aws_role_arn != ""
+                || conf.access_key != ""
+                || conf.secret_key != "")
+        {
+            return Err(std::io::Error::other(
+                "AWS Role cannot be activated with static credentials or the web identity token file",
+            ));
+        } else if conf.bucket == "" {
+            return Err(std::io::Error::other("no bucket name was provided"));
+        }
+
+        let creds;
+        if conf.access_key != "" && conf.secret_key != "" {
+            creds = Credentials::new(
+                conf.access_key.clone(), // access_key_id
+                conf.secret_key.clone(), // secret_access_key
+                None,                    // session_token (optional)
+                None,
+                "Static",
+            );
+        } else {
+            return Err(std::io::Error::other("insufficient parameters for S3 backend authentication"));
+        }
+        let region_provider = RegionProviderChain::default_provider().or_else(Region::new(conf.region.clone()));
+        #[allow(deprecated)]
+        let config = aws_config::from_env()
+            .endpoint_url(conf.endpoint.clone())
+            .region(region_provider)
+            .credentials_provider(creds)
+            .load()
+            .await;
+        let client = Client::new(&config);
+        let client = Arc::new(client);
+        Ok(Self {
+            client,
+            bucket: conf.bucket.clone(),
+            prefix: conf.prefix.clone().trim_matches('/').to_string(),
+            storage_class: conf.storage_class.clone(),
+        })
+    }
+
+    pub fn get_dest(&self, object: &str) -> String {
+        let mut dest_obj = object.to_string();
+        if self.prefix != "" {
+            dest_obj = format!("{}/{}", &self.prefix, object);
+        }
+        return dest_obj;
+    }
+}
+
+#[async_trait::async_trait]
+impl WarmBackend for WarmBackendS3 {
+    async fn put_with_meta(
+        &self,
+        object: &str,
+        r: ReaderImpl,
+        length: i64,
+        meta: HashMap<String, String>,
+    ) -> Result<String, std::io::Error> {
+        let client = self.client.clone();
+        let Ok(res) = client
+            .put_object()
+            .bucket(&self.bucket)
+            .key(&self.get_dest(object))
+            .body(match r {
+                ReaderImpl::Body(content_body) => ByteStream::from(content_body.to_vec()),
+                ReaderImpl::ObjectBody(mut content_body) => ByteStream::from(content_body.read_all().await?),
+            })
+            .send()
+            .await
+        else {
+            return Err(std::io::Error::other("put_object error"));
+        };
+
+        Ok(res.version_id().unwrap_or("").to_string())
+    }
+
+    async fn put(&self, object: &str, r: ReaderImpl, length: i64) -> Result<String, std::io::Error> {
+        self.put_with_meta(object, r, length, HashMap::new()).await
+    }
+
+    async fn get(&self, object: &str, rv: &str, opts: WarmBackendGetOpts) -> Result<ReadCloser, std::io::Error> {
+        let client = self.client.clone();
+        let mut req = client.get_object().bucket(&self.bucket).key(&self.get_dest(object));
+
+        if !rv.is_empty() {
+            req = req.version_id(rv);
+        }
+
+        if opts.start_offset >= 0 && opts.length > 0 {
+            let end = opts
+                .start_offset
+                .checked_add(opts.length)
+                .and_then(|v| v.checked_sub(1))
+                .ok_or_else(|| std::io::Error::other("invalid range: overflow"))?;
+            req = req.range(format!("bytes={}-{}", opts.start_offset, end));
+        }
+
+        let res = req.send().await.map_err(|e| std::io::Error::other(e.to_string()))?;
+
+        Ok(ReadCloser::new(std::io::Cursor::new(
+            res.body.collect().await.map(|data| data.into_bytes().to_vec())?,
+        )))
+    }
+
+    async fn remove(&self, object: &str, rv: &str) -> Result<(), std::io::Error> {
+        let client = self.client.clone();
+        let mut req = client.delete_object().bucket(&self.bucket).key(&self.get_dest(object));
+
+        if !rv.is_empty() {
+            req = req.version_id(rv);
+        }
+
+        req.send().await.map_err(|e| std::io::Error::other(e.to_string()))?;
+
+        Ok(())
+    }
+
+    async fn in_use(&self) -> Result<bool, std::io::Error> {
+        let client = self.client.clone();
+        let Ok(res) = client
+            .list_objects_v2()
+            .bucket(&self.bucket)
+            //.max_keys(10)
+            //.into_paginator()
+            .send()
+            .await
+        else {
+            return Err(std::io::Error::other("list_objects_v2 error"));
+        };
+
+        Ok(res.common_prefixes.unwrap_or_default().len() > 0 || res.contents.unwrap_or_default().len() > 0)
+    }
+}

crates/ecstore/src/set_disk/ops/object.rs

@@ -315,41 +315,6 @@ async fn get_object_reader_with_context(
     GetObjectReader::new_with_resolver(reader, range, object_info, opts, headers, ctx.object_encryption_resolver()).await
 }
 
-async fn get_legacy_object_reader_with_context<R>(
-    ctx: &InstanceContext,
-    reader: R,
-    terminal: tokio::sync::oneshot::Receiver<Result<()>>,
-    range: Option<HTTPRangeSpec>,
-    object_info: &ObjectInfo,
-    opts: &ObjectOptions,
-    headers: &HeaderMap<HeaderValue>,
-) -> Result<(GetObjectReader, usize, i64)>
-where
-    R: AsyncRead + Unpin + Send + Sync + 'static,
-{
-    // ReadPlan validates this size below; failure here only keeps the terminal
-    // guard inside the transform until that validation returns its typed error.
-    let full_plaintext_size = object_info.get_actual_size().ok();
-    let whole_object = opts.part_number.is_none()
-        && match (&range, full_plaintext_size) {
-            (None, _) => true,
-            (Some(range), Some(size)) => range
-                .get_offset_length(size)
-                .is_ok_and(|(offset, length)| offset == 0 && length == size),
-            (Some(_), None) => false,
-        };
-    let (source, terminal): (Box<dyn AsyncRead + Unpin + Send + Sync>, _) = if whole_object {
-        (Box::new(reader), Some(terminal))
-    } else {
-        (Box::new(LegacyDuplexProducerReader::new(reader, terminal)), None)
-    };
-    let (mut reader, offset, length) = get_object_reader_with_context(ctx, source, range, object_info, opts, headers).await?;
-    if let Some(terminal) = terminal {
-        reader.stream = Box::new(LegacyDuplexProducerReader::new(reader.stream, terminal));
-    }
-    Ok((reader, offset, length))
-}
-
 fn data_read_metadata_early_stop_request_shape_allowed(range: &Option<HTTPRangeSpec>, opts: &ObjectOptions) -> bool {
     range.is_none()
         && opts.part_number.is_none()
@@ -934,7 +899,6 @@ impl crate::storage_api_contracts::object::ObjectIO for SetDisks {
                 &object_info,
                 &opts,
                 &self.ctx.tier_config_mgr(),
-                self.ctx.object_encryption_resolver(),
             )
             .await?;
             return Ok(finish_set_disk_read_lock(gr, read_lock_guard.take(), bucket, object));
@@ -1125,9 +1089,8 @@ impl crate::storage_api_contracts::object::ObjectIO for SetDisks {
         let (rd, wd) = tokio::io::duplex(duplex_buffer_size);
         debug!(bucket, object, duplex_buffer_size, "Created duplex pipe for object data transfer");
 
-        let (producer_terminal_tx, producer_terminal_rx) = tokio::sync::oneshot::channel();
         let (mut reader, offset, length) =
-            get_legacy_object_reader_with_context(&self.ctx, rd, producer_terminal_rx, range, &object_info, opts, &h).await?;
+            get_object_reader_with_context(&self.ctx, Box::new(rd), range, &object_info, opts, &h).await?;
         // Carry the hook probe result so the app layer skips its now-redundant
         // lookup on the streaming miss path (ODC-16).
         reader.body_source = body_source;
@@ -1147,7 +1110,7 @@ impl crate::storage_api_contracts::object::ObjectIO for SetDisks {
             // `get_object_with_fileinfo` also waits on `writer`, so an outer timeout
             // would incorrectly treat downstream backpressure as disk-read latency.
             // Disk read timeouts must be enforced at the actual disk I/O operations.
-            let producer_result = Self::get_object_with_fileinfo(
+            if let Err(e) = Self::get_object_with_fileinfo(
                 &bucket,
                 &object,
                 erasure_cache,
@@ -1165,9 +1128,9 @@ impl crate::storage_api_contracts::object::ObjectIO for SetDisks {
                 object_class.as_str(),
                 size_bucket,
             )
-            .await;
-            if let Err(e) = &producer_result {
-                let reason = classify_storage_error(e);
+            .await
+            {
+                let reason = classify_storage_error(&e);
                 if reason == GetObjectFailureReason::DownstreamClosed {
                     debug!(
                         event = EVENT_SET_DISK_WRITE,
@@ -1206,7 +1169,6 @@ impl crate::storage_api_contracts::object::ObjectIO for SetDisks {
                     );
                 }
             };
-            let _ = producer_terminal_tx.send(producer_result.map(|_| ()));
         });
 
         Ok(reader)
@@ -2595,420 +2557,6 @@ impl<R: AsyncRead + Unpin> AsyncRead for TransitionUploadReader<R> {
     }
 }
 
-struct LegacyDuplexProducerReader<R> {
-    inner: Option<R>,
-    terminal: Option<tokio::sync::oneshot::Receiver<Result<()>>>,
-    inner_eof: bool,
-}
-
-impl<R> LegacyDuplexProducerReader<R> {
-    fn new(inner: R, terminal: tokio::sync::oneshot::Receiver<Result<()>>) -> Self {
-        Self {
-            inner: Some(inner),
-            terminal: Some(terminal),
-            inner_eof: false,
-        }
-    }
-}
-
-impl<R: AsyncRead + Unpin> AsyncRead for LegacyDuplexProducerReader<R> {
-    fn poll_read(mut self: Pin<&mut Self>, cx: &mut Context<'_>, buf: &mut ReadBuf<'_>) -> Poll<std::io::Result<()>> {
-        if buf.remaining() == 0 {
-            return Poll::Ready(Ok(()));
-        }
-        if !self.inner_eof {
-            let before = buf.filled().len();
-            if let Some(inner) = self.inner.as_mut() {
-                match Pin::new(inner).poll_read(cx, buf) {
-                    Poll::Pending => return Poll::Pending,
-                    Poll::Ready(Err(err)) => return Poll::Ready(Err(err)),
-                    Poll::Ready(Ok(())) if buf.filled().len() > before => return Poll::Ready(Ok(())),
-                    Poll::Ready(Ok(())) => {
-                        self.inner_eof = true;
-                        self.inner = None;
-                    }
-                }
-            } else {
-                self.inner_eof = true;
-            }
-        }
-
-        let Some(terminal) = self.terminal.as_mut() else {
-            return Poll::Ready(Ok(()));
-        };
-        match Pin::new(terminal).poll(cx) {
-            Poll::Pending => Poll::Pending,
-            Poll::Ready(Ok(Ok(()))) => {
-                self.terminal = None;
-                Poll::Ready(Ok(()))
-            }
-            Poll::Ready(Ok(Err(err))) => {
-                self.terminal = None;
-                Poll::Ready(Err(std::io::Error::other(err)))
-            }
-            Poll::Ready(Err(_)) => {
-                self.terminal = None;
-                Poll::Ready(Err(std::io::Error::other(StorageError::Unexpected)))
-            }
-        }
-    }
-}
-
-#[cfg(test)]
-mod legacy_duplex_producer_reader_tests {
-    use super::*;
-    use crate::object_api::{EncryptionResolutionError, ObjectEncryptionResolver, ReadEncryptionMaterial, ReadEncryptionMode};
-    use rustfs_utils::CompressionAlgorithm;
-    use std::collections::HashMap;
-    use std::sync::Arc;
-    use tokio::io::{AsyncReadExt, AsyncWriteExt};
-
-    const TEST_DUPLEX_CAPACITY: usize = 64 * 1024;
-
-    fn storage_error_source(error: &std::io::Error) -> &StorageError {
-        error
-            .get_ref()
-            .and_then(|source| source.downcast_ref::<StorageError>())
-            .expect("legacy duplex terminal error should retain StorageError source")
-    }
-
-    async fn compressed_fixture(plaintext: Vec<u8>, recorded_size: usize) -> (Vec<u8>, ObjectInfo) {
-        let mut compressor = rustfs_rio::CompressReader::new(std::io::Cursor::new(plaintext), CompressionAlgorithm::default());
-        let mut compressed = Vec::new();
-        compressor
-            .read_to_end(&mut compressed)
-            .await
-            .expect("compress test plaintext");
-
-        let mut metadata = HashMap::new();
-        rustfs_utils::http::insert_str(
-            &mut metadata,
-            rustfs_utils::http::SUFFIX_COMPRESSION,
-            CompressionAlgorithm::default().to_string(),
-        );
-        rustfs_utils::http::insert_str(&mut metadata, rustfs_utils::http::SUFFIX_ACTUAL_SIZE, recorded_size.to_string());
-        let object_info = ObjectInfo {
-            size: i64::try_from(compressed.len()).expect("compressed fixture length should fit in i64"),
-            user_defined: Arc::new(metadata),
-            ..Default::default()
-        };
-        (compressed, object_info)
-    }
-
-    #[tokio::test]
-    async fn legacy_duplex_reader_allows_clean_completion() {
-        let (mut writer, reader) = tokio::io::duplex(64);
-        let (terminal_tx, terminal_rx) = tokio::sync::oneshot::channel();
-        writer
-            .write_all(b"complete")
-            .await
-            .expect("duplex write should fit in buffer");
-        drop(writer);
-        terminal_tx.send(Ok(())).expect("terminal receiver should remain installed");
-
-        let mut reader = LegacyDuplexProducerReader::new(reader, terminal_rx);
-        let mut out = Vec::new();
-        reader
-            .read_to_end(&mut out)
-            .await
-            .expect("clean producer completion should surface clean EOF");
-
-        assert_eq!(out, b"complete");
-    }
-
-    #[tokio::test]
-    async fn legacy_duplex_reader_ignores_zero_capacity_read_buf() {
-        let (mut writer, reader) = tokio::io::duplex(64);
-        let (terminal_tx, terminal_rx) = tokio::sync::oneshot::channel();
-        writer.write_all(b"body").await.expect("duplex write should fit in buffer");
-        drop(writer);
-        terminal_tx
-            .send(Err(StorageError::FileCorrupt))
-            .expect("terminal receiver should remain installed");
-
-        let mut reader = LegacyDuplexProducerReader::new(reader, terminal_rx);
-        let mut empty = [];
-        std::future::poll_fn(|cx| {
-            let mut read_buf = ReadBuf::new(&mut empty);
-            Pin::new(&mut reader).poll_read(cx, &mut read_buf)
-        })
-        .await
-        .expect("zero-capacity reads should complete without observing EOF or terminal state");
-        assert!(!reader.inner_eof);
-        assert!(reader.terminal.is_some());
-
-        let mut out = Vec::new();
-        let err = reader
-            .read_to_end(&mut out)
-            .await
-            .expect_err("subsequent reads must still receive data and the terminal error");
-        assert_eq!(out, b"body");
-        assert!(matches!(storage_error_source(&err), StorageError::FileCorrupt));
-    }
-
-    #[tokio::test]
-    async fn legacy_duplex_reader_surfaces_terminal_error_after_partial_data() {
-        let (mut writer, reader) = tokio::io::duplex(64);
-        let (terminal_tx, terminal_rx) = tokio::sync::oneshot::channel();
-        writer.write_all(b"partial").await.expect("duplex write should fit in buffer");
-        drop(writer);
-        terminal_tx
-            .send(Err(StorageError::FileCorrupt))
-            .expect("terminal receiver should remain installed");
-
-        let mut reader = LegacyDuplexProducerReader::new(reader, terminal_rx);
-        let mut out = Vec::new();
-        let err = reader
-            .read_to_end(&mut out)
-            .await
-            .expect_err("terminal producer error must not become clean EOF");
-
-        assert_eq!(out, b"partial");
-        assert!(matches!(storage_error_source(&err), StorageError::FileCorrupt));
-    }
-
-    #[tokio::test]
-    async fn legacy_duplex_reader_surfaces_terminal_error_after_declared_length() {
-        let (mut writer, reader) = tokio::io::duplex(64);
-        let (terminal_tx, terminal_rx) = tokio::sync::oneshot::channel();
-        writer.write_all(b"exact").await.expect("duplex write should fit in buffer");
-        drop(writer);
-        terminal_tx
-            .send(Err(StorageError::Io(std::io::Error::new(
-                std::io::ErrorKind::ConnectionReset,
-                "remote body reset after final byte",
-            ))))
-            .expect("terminal receiver should remain installed");
-
-        let reader = LegacyDuplexProducerReader::new(reader, terminal_rx);
-        let mut reader =
-            HashReader::from_stream(reader, 5, 5, None, None, false).expect("hash reader should accept exact declared length");
-        let mut out = Vec::new();
-        let err = reader
-            .read_to_end(&mut out)
-            .await
-            .expect_err("producer terminal error after the declared length must still fail");
-
-        assert_eq!(out, b"exact");
-        assert!(
-            matches!(storage_error_source(&err), StorageError::Io(io_error) if io_error.kind() == std::io::ErrorKind::ConnectionReset)
-        );
-    }
-
-    #[tokio::test]
-    async fn legacy_compressed_reader_surfaces_terminal_error_after_complete_plaintext() {
-        let plaintext = b"compressed terminal result must survive the plaintext limit".repeat(16);
-        let (compressed, object_info) = compressed_fixture(plaintext.clone(), plaintext.len()).await;
-        let full_range = HTTPRangeSpec {
-            is_suffix_length: false,
-            start: 0,
-            end: i64::try_from(plaintext.len()).expect("plaintext fixture length should fit in i64") - 1,
-        };
-        for range in [None, Some(full_range)] {
-            let (mut writer, reader) = tokio::io::duplex(compressed.len().max(1));
-            writer
-                .write_all(&compressed)
-                .await
-                .expect("compressed body should fit in duplex buffer");
-            drop(writer);
-            let (terminal_tx, terminal_rx) = tokio::sync::oneshot::channel();
-            terminal_tx
-                .send(Err(StorageError::FileCorrupt))
-                .expect("terminal receiver should remain installed");
-
-            let (mut reader, _, _) = get_legacy_object_reader_with_context(
-                &InstanceContext::new(),
-                reader,
-                terminal_rx,
-                range,
-                &object_info,
-                &ObjectOptions::default(),
-                &HeaderMap::new(),
-            )
-            .await
-            .expect("compressed read plan should build");
-            let mut out = Vec::new();
-            let err = reader
-                .read_to_end(&mut out)
-                .await
-                .expect_err("terminal error after complete decompression must not become clean EOF");
-
-            assert_eq!(out, plaintext);
-            assert!(matches!(storage_error_source(&err), StorageError::FileCorrupt));
-        }
-    }
-
-    #[tokio::test]
-    async fn legacy_exact_reader_rejects_extra_data_without_backpressure_deadlock() {
-        let payload = vec![0x5a; TEST_DUPLEX_CAPACITY * 2];
-        let (mut writer, reader) = tokio::io::duplex(TEST_DUPLEX_CAPACITY);
-        let (terminal_tx, terminal_rx) = tokio::sync::oneshot::channel();
-        let producer = tokio::spawn(async move {
-            let result = writer.write_all(&payload).await;
-            drop(writer);
-            let terminal_result = result
-                .as_ref()
-                .map(|_| ())
-                .map_err(|err| StorageError::Io(std::io::Error::new(err.kind(), err.to_string())));
-            let _ = terminal_tx.send(terminal_result);
-            result
-        });
-        let reader = crate::io_support::rio::HardLimitReader::new(reader, 1);
-        let mut reader = LegacyDuplexProducerReader::new(reader, terminal_rx);
-
-        let mut out = Vec::new();
-        tokio::time::timeout(std::time::Duration::from_secs(1), reader.read_to_end(&mut out))
-            .await
-            .expect("extra data beyond the declared size must not deadlock")
-            .expect_err("extra data beyond the declared size must fail closed");
-        assert_eq!(out, [0x5a]);
-        drop(reader);
-        let _ = tokio::time::timeout(std::time::Duration::from_secs(1), producer)
-            .await
-            .expect("producer must unblock after the read fails")
-            .expect("producer task should not panic");
-    }
-
-    #[tokio::test]
-    async fn legacy_terminal_reader_releases_unconsumed_source_before_waiting() {
-        let payload = vec![0x5a; TEST_DUPLEX_CAPACITY * 2];
-        let (mut writer, reader) = tokio::io::duplex(TEST_DUPLEX_CAPACITY);
-        let (terminal_tx, terminal_rx) = tokio::sync::oneshot::channel();
-        let producer = tokio::spawn(async move {
-            let result = writer.write_all(&payload).await;
-            drop(writer);
-            let terminal_result = result
-                .as_ref()
-                .map(|_| ())
-                .map_err(|err| StorageError::Io(std::io::Error::new(err.kind(), err.to_string())));
-            let _ = terminal_tx.send(terminal_result);
-            result
-        });
-        let reader = rustfs_rio::LimitReader::new(reader, 1);
-        let mut reader = LegacyDuplexProducerReader::new(reader, terminal_rx);
-
-        let mut out = Vec::new();
-        let err = tokio::time::timeout(std::time::Duration::from_secs(1), reader.read_to_end(&mut out))
-            .await
-            .expect("terminal wait must not deadlock behind unconsumed source data")
-            .expect_err("unconsumed source data must fail the producer terminal result");
-        assert_eq!(out, [0x5a]);
-        assert!(
-            matches!(storage_error_source(&err), StorageError::Io(io_error) if io_error.kind() == std::io::ErrorKind::BrokenPipe)
-        );
-        producer
-            .await
-            .expect("producer task should not panic")
-            .expect_err("source should close early");
-    }
-
-    struct FixedEncryptionResolver {
-        key_bytes: [u8; 32],
-        base_nonce: [u8; 12],
-    }
-
-    #[async_trait::async_trait]
-    impl ObjectEncryptionResolver for FixedEncryptionResolver {
-        async fn resolve_read_material(
-            &self,
-            _request: crate::object_api::ReadEncryptionRequest<'_>,
-        ) -> std::result::Result<Option<ReadEncryptionMaterial>, EncryptionResolutionError> {
-            Ok(Some(ReadEncryptionMaterial {
-                key_bytes: self.key_bytes,
-                mode: ReadEncryptionMode::Direct {
-                    base_nonce: self.base_nonce,
-                },
-            }))
-        }
-    }
-
-    #[tokio::test]
-    async fn legacy_encrypted_reader_surfaces_terminal_error_after_complete_plaintext() {
-        let plaintext = b"encrypted terminal result must survive the plaintext limit".repeat(16);
-        let key_bytes = [0x31; 32];
-        let base_nonce = [0x42; 12];
-        let mut encryptor = rustfs_rio::EncryptReader::new(std::io::Cursor::new(plaintext.clone()), key_bytes, base_nonce);
-        let mut encrypted = Vec::new();
-        encryptor.read_to_end(&mut encrypted).await.expect("encrypt test plaintext");
-
-        let object_info = ObjectInfo {
-            bucket: "bucket".to_string(),
-            name: "encrypted-object".to_string(),
-            size: i64::try_from(encrypted.len()).expect("encrypted fixture length should fit in i64"),
-            user_defined: Arc::new(HashMap::from([
-                ("x-amz-server-side-encryption-customer-algorithm".to_string(), "AES256".to_string()),
-                (
-                    "x-amz-server-side-encryption-customer-original-size".to_string(),
-                    plaintext.len().to_string(),
-                ),
-            ])),
-            ..Default::default()
-        };
-        let ctx = InstanceContext::new();
-        assert!(
-            ctx.set_object_encryption_resolver(Arc::new(FixedEncryptionResolver { key_bytes, base_nonce }))
-                .is_ok(),
-            "fresh context should accept resolver"
-        );
-        let full_range = HTTPRangeSpec {
-            is_suffix_length: false,
-            start: 0,
-            end: i64::try_from(plaintext.len()).expect("plaintext fixture length should fit in i64") - 1,
-        };
-        for range in [None, Some(full_range)] {
-            let (mut writer, reader) = tokio::io::duplex(encrypted.len().max(1));
-            writer
-                .write_all(&encrypted)
-                .await
-                .expect("encrypted body should fit in duplex buffer");
-            drop(writer);
-            let (terminal_tx, terminal_rx) = tokio::sync::oneshot::channel();
-            terminal_tx
-                .send(Err(StorageError::FileCorrupt))
-                .expect("terminal receiver should remain installed");
-
-            let (mut reader, _, _) = get_legacy_object_reader_with_context(
-                &ctx,
-                reader,
-                terminal_rx,
-                range,
-                &object_info,
-                &ObjectOptions::default(),
-                &HeaderMap::new(),
-            )
-            .await
-            .expect("encrypted read plan should build");
-            let mut out = Vec::new();
-            let err = reader
-                .read_to_end(&mut out)
-                .await
-                .expect_err("terminal error after complete decryption must not become clean EOF");
-
-            assert_eq!(out, plaintext);
-            assert!(matches!(storage_error_source(&err), StorageError::FileCorrupt));
-        }
-    }
-
-    #[tokio::test]
-    async fn legacy_duplex_reader_fails_closed_when_terminal_channel_closes() {
-        let (mut writer, reader) = tokio::io::duplex(64);
-        let (terminal_tx, terminal_rx) = tokio::sync::oneshot::channel::<Result<()>>();
-        writer.write_all(b"body").await.expect("duplex write should fit in buffer");
-        drop(writer);
-        drop(terminal_tx);
-
-        let mut reader = LegacyDuplexProducerReader::new(reader, terminal_rx);
-        let mut out = Vec::new();
-        let err = reader
-            .read_to_end(&mut out)
-            .await
-            .expect_err("producer disappearance must fail closed");
-
-        assert_eq!(out, b"body");
-        assert!(matches!(storage_error_source(&err), StorageError::Unexpected));
-    }
-}
-
 struct TransitionUploadWriter<W> {
     inner: W,
     produced: u64,
@@ -6517,7 +6065,6 @@ impl crate::storage_api_contracts::object::ObjectOperations for SetDisks {
                 &oi,
                 &opts,
                 &self_.ctx.tier_config_mgr(),
-                self_.ctx.object_encryption_resolver(),
             )
             .await;
             if let Err(err) = gr {
@@ -6587,7 +6134,6 @@ impl crate::storage_api_contracts::object::ObjectOperations for SetDisks {
                     &oi,
                     &part_opts,
                     &self_.ctx.tier_config_mgr(),
-                    self_.ctx.object_encryption_resolver(),
                 )
                 .await
                 .map_err(StorageError::Io)?;
@@ -11634,7 +11180,7 @@ mod put_object_tmp_cleanup_tests {
     use tokio::io::AsyncReadExt;
 
     /// Large enough that the erasure shards are written as real tmp files
-    /// (never inlined into xl.meta), so the cleanup tests exercise actual cleanup.
+    /// (never inlined into xl.meta), so both tests exercise actual cleanup.
     const TEST_OBJECT_SIZE: usize = 1 << 20;
 
     /// Entries under `.rustfs.sys/tmp` on every disk, excluding the `.trash`
@@ -11658,18 +11204,6 @@ mod put_object_tmp_cleanup_tests {
         leftovers
     }
 
-    async fn wait_for_tmp_workspace_to_drain(temp_dirs: &[TempDir], failure_context: &str) {
-        let deadline = tokio::time::Instant::now() + Duration::from_secs(10);
-        loop {
-            let leftovers = non_trash_tmp_entries(temp_dirs).await;
-            if leftovers.is_empty() {
-                break;
-            }
-            assert!(tokio::time::Instant::now() < deadline, "{failure_context}, leftovers: {leftovers:?}");
-            tokio::time::sleep(Duration::from_millis(25)).await;
-        }
-    }
-
     #[tokio::test]
     async fn put_object_success_eventually_cleans_tmp_workspace() {
         let (temp_dirs, disk_stores, set_disks) = hermetic_set_disks(4).await;
@@ -11685,39 +11219,22 @@ mod put_object_tmp_cleanup_tests {
             .await
             .expect("put_object should succeed");
 
-        wait_for_tmp_workspace_to_drain(&temp_dirs, "tmp workspace should drain after a successful PUT").await;
-
-        drop(temp_dirs);
-    }
-
-    #[tokio::test]
-    async fn cancelled_put_before_rename_cleans_tmp_workspace() {
-        let (temp_dirs, disk_stores, set_disks) = hermetic_set_disks(4).await;
-
-        let bucket = "tmp-clean-cancelled-bucket";
-        let object = "cancelled-object";
-        for disk in &disk_stores {
-            disk.make_volume(bucket).await.expect("bucket volume should be created");
+        // The speculative cleanup runs on a spawned task off the PUT response
+        // path, so poll for the tmp workspace to drain instead of asserting
+        // immediately.
+        let deadline = tokio::time::Instant::now() + Duration::from_secs(10);
+        loop {
+            let leftovers = non_trash_tmp_entries(&temp_dirs).await;
+            if leftovers.is_empty() {
+                break;
+            }
+            assert!(
+                tokio::time::Instant::now() < deadline,
+                "tmp workspace should drain after a successful PUT, leftovers: {leftovers:?}"
+            );
+            tokio::time::sleep(Duration::from_millis(25)).await;
         }
 
-        let barrier = PutObjectCommitBarrier::install(bucket, object, PutObjectCommitPause::AfterQuotaReservation);
-        let cancelled_set = set_disks.clone();
-        let put = tokio::spawn(async move {
-            let mut reader = PutObjReader::from_vec(vec![8u8; TEST_OBJECT_SIZE]);
-            cancelled_set
-                .put_object(bucket, object, &mut reader, &ObjectOptions::default())
-                .await
-        });
-        barrier.wait_until_paused().await;
-        put.abort();
-        let join_error = put.await.expect_err("the paused PUT task must be cancelled");
-        assert!(join_error.is_cancelled(), "the paused PUT task must not panic");
-
-        // Keep the barrier armed so a detached child cannot proceed and hide
-        // missing cancellation cleanup.
-        wait_for_tmp_workspace_to_drain(&temp_dirs, "cancelling before rename should drain the tmp workspace").await;
-
-        drop(barrier);
         drop(temp_dirs);
     }
 

crates/ecstore/src/store/list.rs

@@ -0,0 +1,81 @@
+// 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 super::*;
+
+impl ECStore {
+    #[instrument(level = "trace", skip(self))]
+    #[allow(clippy::too_many_arguments)]
+    pub(super) async fn handle_list_objects_v2(
+        self: Arc<Self>,
+        bucket: &str,
+        prefix: &str,
+        continuation_token: Option<String>,
+        delimiter: Option<String>,
+        max_keys: i32,
+        fetch_owner: bool,
+        start_after: Option<String>,
+        incl_deleted: bool,
+    ) -> Result<ListObjectsV2Info> {
+        self.inner_list_objects_v2(
+            bucket,
+            prefix,
+            continuation_token,
+            delimiter,
+            max_keys,
+            fetch_owner,
+            start_after,
+            incl_deleted,
+        )
+        .await
+    }
+
+    #[instrument(skip(self))]
+    pub(super) async fn handle_list_object_versions(
+        self: Arc<Self>,
+        bucket: &str,
+        prefix: &str,
+        marker: Option<String>,
+        version_marker: Option<String>,
+        delimiter: Option<String>,
+        max_keys: i32,
+    ) -> Result<ListObjectVersionsInfo> {
+        self.inner_list_object_versions(bucket, prefix, marker, version_marker, delimiter, max_keys)
+            .await
+    }
+
+    pub(crate) async fn list_object_versions_for_lifecycle(
+        self: Arc<Self>,
+        bucket: &str,
+        prefix: &str,
+        marker: Option<String>,
+        version_marker: Option<String>,
+        delimiter: Option<String>,
+        max_keys: i32,
+    ) -> Result<ListObjectVersionsInfo> {
+        self.inner_list_object_versions_for_lifecycle(bucket, prefix, marker, version_marker, delimiter, max_keys)
+            .await
+    }
+
+    pub(super) async fn handle_walk(
+        self: Arc<Self>,
+        rx: CancellationToken,
+        bucket: &str,
+        prefix: &str,
+        result: tokio::sync::mpsc::Sender<ObjectInfoOrErr>,
+        opts: WalkOptions,
+    ) -> Result<()> {
+        self.walk_internal(rx, bucket, prefix, result, opts).await
+    }
+}

crates/ecstore/src/store/list_objects.rs

@@ -3845,7 +3845,7 @@ impl ECStore {
         .await
     }
 
-    pub(crate) async fn list_object_versions_for_lifecycle(
+    pub(crate) async fn inner_list_object_versions_for_lifecycle(
         self: Arc<Self>,
         bucket: &str,
         prefix: &str,

crates/ecstore/src/store/mod.rs

@@ -148,6 +148,7 @@ mod heal_walk;
 pub use heal_walk::HealWalkVersion;
 mod init;
 pub(crate) mod init_format;
+mod list;
 pub(crate) mod list_objects;
 mod multipart;
 mod object;
@@ -600,7 +601,7 @@ impl crate::storage_api_contracts::list::ListOperations for ECStore {
         start_after: Option<String>,
         incl_deleted: bool,
     ) -> Result<ListObjectsV2Info> {
-        self.inner_list_objects_v2(
+        self.handle_list_objects_v2(
             bucket,
             prefix,
             continuation_token,
@@ -623,7 +624,7 @@ impl crate::storage_api_contracts::list::ListOperations for ECStore {
         delimiter: Option<String>,
         max_keys: i32,
     ) -> Result<ListObjectVersionsInfo> {
-        self.inner_list_object_versions(bucket, prefix, marker, version_marker, delimiter, max_keys)
+        self.handle_list_object_versions(bucket, prefix, marker, version_marker, delimiter, max_keys)
             .await
     }
 
@@ -635,7 +636,7 @@ impl crate::storage_api_contracts::list::ListOperations for ECStore {
         result: tokio::sync::mpsc::Sender<ObjectInfoOrErr>,
         opts: WalkOptions,
     ) -> Result<()> {
-        self.walk_internal(rx, bucket, prefix, result, opts).await
+        self.handle_walk(rx, bucket, prefix, result, opts).await
     }
 }
 

crates/ecstore/tests/ecstore_contract_compat_test.rs

@@ -157,18 +157,6 @@ fn ecstore_implements_storage_list_operations_contract() {
     assert!(storage_list_operations_type_name::<ECStore>().ends_with("::ECStore"));
 }
 
-#[test]
-fn ecstore_pools_expose_storage_list_operations_contract() {
-    fn assert_contract(store: &ECStore) {
-        let future = store.pools[0]
-            .clone()
-            .list_objects_v2("bucket", "", None, None, 1, false, None, false);
-        drop(future);
-    }
-
-    let _ = assert_contract;
-}
-
 #[test]
 fn ecstore_implements_storage_multipart_operations_contract() {
     assert!(storage_multipart_operations_type_name::<ECStore>().ends_with("::ECStore"));

crates/log-analyzer/src/rules/seed/heal.rs

@@ -103,7 +103,11 @@ pub(super) fn rules() -> Vec<Rule> {
                 P2Degraded,
                 "heal",
                 "heal 任务调度/执行失败",
-                any([prefix("Heal task timeout"), prefix("Heal task execution failed")]),
+                any([
+                    prefix("Heal task timeout"),
+                    prefix("Heal task execution failed"),
+                    contains("Heal manager is not running"),
+                ]),
                 "heal 任务调度/执行层故障。",
                 "检查 heal 后台服务状态与资源压力。",
             )

crates/rio/src/compress_reader.rs

@@ -71,7 +71,6 @@ where
 
     /// Optional: allow users to customize block_size
     pub fn with_block_size(inner: R, block_size: usize, compression_algorithm: CompressionAlgorithm) -> Self {
-        debug_assert!(block_size > 0, "CompressReader block_size must be non-zero");
         Self {
             inner,
             buffer: Vec::new(),
@@ -184,21 +183,11 @@ pin_project! {
         buffer: Vec<u8>,
         buffer_pos: usize,
         finished: bool,
-        // A previously surfaced stream error is sticky: without this, a caller
-        // that polls again after an error would restart at the header phase and
-        // read a truncated tail as a clean EOF, converting the error into a
-        // silently short body.
-        poisoned: bool,
         // Fields for saving header read progress across polls
         header_buf: [u8; 8],
         header_read: usize,
-        // Fields for saving compressed block read progress across polls.
-        // `compressed_len > 0` means a block payload is in flight: the header has
-        // been fully parsed and `compressed_read` bytes of the payload are already
-        // consumed from the inner stream. The header phase must not run again (and
-        // must not reset `compressed_read`) until this block completes, or a
-        // `Poll::Pending` in the middle of a payload would silently drop the bytes
-        // read so far and desynchronize the block framing.
+        header_done: bool,
+        // Fields for saving compressed block read progress across polls
         compressed_buf: Vec<u8>,
         compressed_read: usize,
         compressed_len: usize,
@@ -216,9 +205,9 @@ where
             buffer: Vec::new(),
             buffer_pos: 0,
             finished: false,
-            poisoned: false,
             header_buf: [0u8; 8],
             header_read: 0,
+            header_done: false,
             compressed_buf: Vec::new(),
             compressed_read: 0,
             compressed_len: 0,
@@ -247,74 +236,54 @@ where
         if *this.finished {
             return Poll::Ready(Ok(()));
         }
-        if *this.poisoned {
-            return Poll::Ready(Err(io::Error::new(io::ErrorKind::InvalidData, "decompress reader previously failed")));
-        }
-
-        if *this.compressed_len == 0 {
-            // Read the 8-byte block header, resuming across polls via `header_read`.
-            while *this.header_read < HEADER_LEN {
-                let mut temp = [0u8; HEADER_LEN];
-                let mut temp_buf = ReadBuf::new(&mut temp[0..HEADER_LEN - *this.header_read]);
-                match this.inner.as_mut().poll_read(cx, &mut temp_buf) {
-                    Poll::Pending => return Poll::Pending,
-                    Poll::Ready(Ok(())) => {
-                        let n = temp_buf.filled().len();
-                        if n == 0 {
-                            if *this.header_read == 0 {
-                                // Clean EOF on a block boundary.
-                                *this.finished = true;
-                                return Poll::Ready(Ok(()));
-                            }
-                            *this.poisoned = true;
-                            return Poll::Ready(Err(io::Error::new(
-                                io::ErrorKind::UnexpectedEof,
-                                "unexpected EOF while reading compressed block header",
-                            )));
-                        }
-                        this.header_buf[*this.header_read..*this.header_read + n].copy_from_slice(&temp_buf.filled()[..n]);
-                        *this.header_read += n;
-                    }
-                    Poll::Ready(Err(e)) => {
-                        // error!("DecompressReader poll_read: read header error: {e}");
-                        *this.poisoned = true;
-                        return Poll::Ready(Err(e));
+        // Read header
+        while !*this.header_done && *this.header_read < HEADER_LEN {
+            let mut temp = [0u8; HEADER_LEN];
+            let mut temp_buf = ReadBuf::new(&mut temp[0..HEADER_LEN - *this.header_read]);
+            match this.inner.as_mut().poll_read(cx, &mut temp_buf) {
+                Poll::Pending => return Poll::Pending,
+                Poll::Ready(Ok(())) => {
+                    let n = temp_buf.filled().len();
+                    if n == 0 {
+                        break;
                     }
+                    this.header_buf[*this.header_read..*this.header_read + n].copy_from_slice(&temp_buf.filled()[..n]);
+                    *this.header_read += n;
+                }
+                Poll::Ready(Err(e)) => {
+                    // error!("DecompressReader poll_read: read header error: {e}");
+                    return Poll::Ready(Err(e));
                 }
             }
+            if *this.header_read < HEADER_LEN {
+                return Poll::Pending;
+            }
+        }
+        if !*this.header_done && *this.header_read == 0 {
+            return Poll::Ready(Ok(()));
+        }
+        let typ = this.header_buf[0];
+        let len = (this.header_buf[1] as usize) | ((this.header_buf[2] as usize) << 8) | ((this.header_buf[3] as usize) << 16);
+        let crc = (this.header_buf[4] as u32)
+            | ((this.header_buf[5] as u32) << 8)
+            | ((this.header_buf[6] as u32) << 16)
+            | ((this.header_buf[7] as u32) << 24);
+        *this.header_read = 0;
+        *this.header_done = true;
 
-            let typ = this.header_buf[0];
-            let len =
-                (this.header_buf[1] as usize) | ((this.header_buf[2] as usize) << 8) | ((this.header_buf[3] as usize) << 16);
-            *this.header_read = 0;
-
-            // `CompressReader` never emits an end block — a stream terminates on
-            // inner EOF, which is what lets concatenated per-part streams decode as
-            // one. This branch is kept for streams that do carry the marker.
-            if typ == COMPRESS_TYPE_END {
-                *this.compressed_read = 0;
-                *this.compressed_len = 0;
-                *this.finished = true;
-                return Poll::Ready(Ok(()));
-            }
-            if typ != COMPRESS_TYPE_COMPRESSED && typ != COMPRESS_TYPE_UNCOMPRESSED {
-                // error!("DecompressReader unknown compression type: {typ}");
-                *this.poisoned = true;
-                return Poll::Ready(Err(io::Error::new(io::ErrorKind::InvalidData, "Unknown compression type")));
-            }
-            if len == 0 {
-                *this.poisoned = true;
-                return Poll::Ready(Err(io::Error::new(io::ErrorKind::InvalidData, "Invalid compressed block length")));
-            }
-
-            if this.compressed_buf.len() < len {
-                this.compressed_buf.resize(len, 0);
-            }
-            *this.compressed_len = len;
+        if typ == COMPRESS_TYPE_END {
             *this.compressed_read = 0;
+            *this.compressed_len = 0;
+            *this.finished = true;
+            return Poll::Ready(Ok(()));
         }
 
-        // Fill the in-flight block payload, resuming across polls via `compressed_read`.
+        if this.compressed_buf.len() < len {
+            this.compressed_buf.resize(len, 0);
+        }
+        *this.compressed_len = len;
+        *this.compressed_read = 0;
+
         while *this.compressed_read < *this.compressed_len {
             let mut temp_buf = ReadBuf::new(&mut this.compressed_buf[*this.compressed_read..*this.compressed_len]);
             match this.inner.as_mut().poll_read(cx, &mut temp_buf) {
@@ -322,13 +291,7 @@ where
                 Poll::Ready(Ok(())) => {
                     let n = temp_buf.filled().len();
                     if n == 0 {
-                        *this.compressed_read = 0;
-                        *this.compressed_len = 0;
-                        *this.poisoned = true;
-                        return Poll::Ready(Err(io::Error::new(
-                            io::ErrorKind::UnexpectedEof,
-                            "unexpected EOF while reading compressed block payload",
-                        )));
+                        break;
                     }
                     *this.compressed_read += n;
                 }
@@ -336,17 +299,10 @@ where
                     // error!("DecompressReader poll_read: read compressed block error: {e}");
                     *this.compressed_read = 0;
                     *this.compressed_len = 0;
-                    *this.poisoned = true;
                     return Poll::Ready(Err(e));
                 }
             }
         }
-
-        let typ = this.header_buf[0];
-        let crc = (this.header_buf[4] as u32)
-            | ((this.header_buf[5] as u32) << 8)
-            | ((this.header_buf[6] as u32) << 16)
-            | ((this.header_buf[7] as u32) << 24);
         let compressed_buf = &this.compressed_buf[..*this.compressed_len];
         // `compressed_buf`'s length comes from the untrusted 24-bit header length field, so it
         // can be shorter than 16 bytes. `uvarint` is safe on any slice length (reads at most 10
@@ -360,7 +316,6 @@ where
         if uvarint <= 0 || uvarint as usize > compressed_buf.len() {
             *this.compressed_read = 0;
             *this.compressed_len = 0;
-            *this.poisoned = true;
             return Poll::Ready(Err(io::Error::new(io::ErrorKind::InvalidData, "Invalid compressed block length prefix")));
         }
         let compressed_data = &compressed_buf[uvarint as usize..];
@@ -371,29 +326,21 @@ where
                     // error!("DecompressReader decompress_block error: {e}");
                     *this.compressed_read = 0;
                     *this.compressed_len = 0;
-                    *this.poisoned = true;
                     return Poll::Ready(Err(e));
                 }
             }
-        } else {
-            // The header phase already rejected every type other than
-            // COMPRESS_TYPE_COMPRESSED / COMPRESS_TYPE_UNCOMPRESSED.
+        } else if typ == COMPRESS_TYPE_UNCOMPRESSED {
             compressed_data.to_vec()
-        };
-        if decompressed.is_empty() {
-            // The writer never emits zero-length plaintext blocks; an empty
-            // decode surfacing as Ready(Ok) with no bytes would read as EOF and
-            // silently truncate the stream.
-            *this.poisoned = true;
+        } else {
+            // error!("DecompressReader unknown compression type: {typ}");
             *this.compressed_read = 0;
             *this.compressed_len = 0;
-            return Poll::Ready(Err(io::Error::new(io::ErrorKind::InvalidData, "Empty compressed block")));
-        }
+            return Poll::Ready(Err(io::Error::new(io::ErrorKind::InvalidData, "Unknown compression type")));
+        };
         if decompressed.len() != uncompress_len as usize {
             // error!("DecompressReader decompressed length mismatch: {} != {}", decompressed.len(), uncompress_len);
             *this.compressed_read = 0;
             *this.compressed_len = 0;
-            *this.poisoned = true;
             return Poll::Ready(Err(io::Error::new(io::ErrorKind::InvalidData, "Decompressed length mismatch")));
         }
         let actual_crc = {
@@ -405,13 +352,13 @@ where
             // error!("DecompressReader CRC32 mismatch: actual {actual_crc} != expected {crc}");
             *this.compressed_read = 0;
             *this.compressed_len = 0;
-            *this.poisoned = true;
             return Poll::Ready(Err(io::Error::new(io::ErrorKind::InvalidData, "CRC32 mismatch")));
         }
         *this.buffer = decompressed;
         *this.buffer_pos = 0;
         *this.compressed_read = 0;
         *this.compressed_len = 0;
+        *this.header_done = false;
         let to_copy = min(buf.remaining(), this.buffer.len());
         buf.put_slice(&this.buffer[..to_copy]);
         *this.buffer_pos += to_copy;
@@ -546,184 +493,6 @@ mod tests {
         assert_eq!(&decompressed, &data);
     }
 
-    /// Wraps a reader so every other poll returns `Poll::Pending` and every
-    /// `Ready` poll serves at most `chunk` bytes. This is the shape a duplex
-    /// pipe produces when the erasure writer is slower than the decoder, which
-    /// is exactly what desynchronized the block framing before the resumable
-    /// payload state was added (rustfs/rustfs#5957 multipart GET truncation).
-    struct PendingChunkReader<R> {
-        inner: R,
-        chunk: usize,
-        pending_next: bool,
-    }
-
-    impl<R> PendingChunkReader<R> {
-        fn new(inner: R, chunk: usize) -> Self {
-            Self {
-                inner,
-                chunk,
-                pending_next: true,
-            }
-        }
-    }
-
-    impl<R: AsyncRead + Unpin> AsyncRead for PendingChunkReader<R> {
-        fn poll_read(
-            mut self: std::pin::Pin<&mut Self>,
-            cx: &mut std::task::Context<'_>,
-            buf: &mut tokio::io::ReadBuf<'_>,
-        ) -> std::task::Poll<std::io::Result<()>> {
-            if self.pending_next {
-                self.pending_next = false;
-                cx.waker().wake_by_ref();
-                return std::task::Poll::Pending;
-            }
-            self.pending_next = true;
-            let cap = self.chunk.min(buf.remaining());
-            let mut scratch = vec![0u8; cap];
-            let mut inner_buf = tokio::io::ReadBuf::new(&mut scratch);
-            match std::pin::Pin::new(&mut self.inner).poll_read(cx, &mut inner_buf) {
-                std::task::Poll::Ready(Ok(())) => {
-                    buf.put_slice(inner_buf.filled());
-                    std::task::Poll::Ready(Ok(()))
-                }
-                other => other,
-            }
-        }
-    }
-
-    fn patterned_payload(size: usize, seed: u8) -> Vec<u8> {
-        (0..size)
-            .map(|i| ((i as u64).wrapping_mul(2_654_435_761).wrapping_add(seed as u64) >> 3) as u8)
-            .collect()
-    }
-
-    /// Root-cause regression for the multipart compressed GET truncation: a
-    /// `Poll::Pending` in the middle of a block payload must not drop the bytes
-    /// already consumed. Before the resumable payload state, the decoder reset
-    /// `compressed_read` on every re-poll and surfaced
-    /// `LZ4 error: ERROR_frameType_unknown` mid-stream.
-    #[tokio::test]
-    async fn test_decompress_reader_survives_pending_mid_payload() {
-        let data = patterned_payload(100 * 1024, 7);
-        let mut compress_reader =
-            CompressReader::with_block_size(Cursor::new(data.clone()), 8192, CompressionAlgorithm::default());
-        let mut compressed = Vec::new();
-        compress_reader.read_to_end(&mut compressed).await.unwrap();
-
-        for chunk in [1usize, 3, 7, 8, 17, 1000, 8192] {
-            let inner = PendingChunkReader::new(Cursor::new(compressed.clone()), chunk);
-            let mut decompress_reader = DecompressReader::new(inner, CompressionAlgorithm::default());
-            let mut decompressed = Vec::new();
-            decompress_reader.read_to_end(&mut decompressed).await.unwrap();
-            assert_eq!(decompressed, data, "pending-chunked decode must be byte-exact for chunk={chunk}");
-        }
-    }
-
-    /// Two independently compressed streams concatenated back to back — the
-    /// on-disk shape of a compressed multipart object — must decode across the
-    /// stream boundary even when every poll can suspend mid-block.
-    #[tokio::test]
-    async fn test_decompress_reader_survives_pending_across_concatenated_streams() {
-        let part1 = patterned_payload(64 * 1024, 7);
-        let part2 = patterned_payload(24 * 1024, 61);
-
-        let mut stored = Vec::new();
-        for part in [&part1, &part2] {
-            let mut compress_reader =
-                CompressReader::with_block_size(Cursor::new(part.clone()), 8192, CompressionAlgorithm::default());
-            let mut compressed = Vec::new();
-            compress_reader.read_to_end(&mut compressed).await.unwrap();
-            stored.extend_from_slice(&compressed);
-        }
-
-        let mut expected = part1;
-        expected.extend_from_slice(&part2);
-
-        for chunk in [1usize, 5, 8, 13, 4096] {
-            let inner = PendingChunkReader::new(Cursor::new(stored.clone()), chunk);
-            let mut decompress_reader = DecompressReader::new(inner, CompressionAlgorithm::default());
-            let mut decompressed = Vec::new();
-            decompress_reader.read_to_end(&mut decompressed).await.unwrap();
-            assert_eq!(
-                decompressed, expected,
-                "concatenated part streams must decode byte-exact for chunk={chunk}"
-            );
-        }
-    }
-
-    /// After the first stream error, every further poll must keep failing.
-    /// Without the sticky poison a retrying caller would restart at the header
-    /// phase and read the truncated tail as a clean EOF — converting a hard
-    /// error into a silently short body.
-    #[tokio::test]
-    async fn test_decompress_reader_error_is_sticky() {
-        let data = patterned_payload(32 * 1024, 7);
-        let mut compress_reader = CompressReader::with_block_size(Cursor::new(data), 8192, CompressionAlgorithm::default());
-        let mut compressed = Vec::new();
-        compress_reader.read_to_end(&mut compressed).await.unwrap();
-        compressed.truncate(compressed.len() - 3);
-
-        let mut decompress_reader = DecompressReader::new(Cursor::new(compressed), CompressionAlgorithm::default());
-        let mut out = Vec::new();
-        let first = decompress_reader
-            .read_to_end(&mut out)
-            .await
-            .expect_err("truncated payload must error");
-        assert_eq!(first.kind(), std::io::ErrorKind::UnexpectedEof);
-
-        let mut retry = Vec::new();
-        let second = decompress_reader
-            .read_to_end(&mut retry)
-            .await
-            .expect_err("a poll after the first error must not turn into a clean EOF");
-        assert_eq!(second.kind(), std::io::ErrorKind::InvalidData);
-        assert!(retry.is_empty(), "no bytes may be produced after the stream failed");
-    }
-
-    /// A stream cut off in the middle of a block payload must fail with a clean
-    /// UnexpectedEof instead of decoding a short buffer.
-    #[tokio::test]
-    async fn test_decompress_reader_truncated_payload_is_unexpected_eof() {
-        let data = patterned_payload(32 * 1024, 7);
-        let mut compress_reader = CompressReader::with_block_size(Cursor::new(data), 8192, CompressionAlgorithm::default());
-        let mut compressed = Vec::new();
-        compress_reader.read_to_end(&mut compressed).await.unwrap();
-
-        compressed.truncate(compressed.len() - 3);
-        let mut decompress_reader = DecompressReader::new(Cursor::new(compressed), CompressionAlgorithm::default());
-        let mut out = Vec::new();
-        let err = decompress_reader
-            .read_to_end(&mut out)
-            .await
-            .expect_err("truncated payload must error");
-        assert_eq!(err.kind(), std::io::ErrorKind::UnexpectedEof);
-    }
-
-    /// A stream cut off in the middle of a block header must fail with a clean
-    /// UnexpectedEof instead of parsing a garbage header.
-    #[tokio::test]
-    async fn test_decompress_reader_truncated_header_is_unexpected_eof() {
-        let data = patterned_payload(12 * 1024, 7);
-        let mut compress_reader = CompressReader::with_block_size(Cursor::new(data), 8192, CompressionAlgorithm::default());
-        let mut compressed = Vec::new();
-        compress_reader.read_to_end(&mut compressed).await.unwrap();
-
-        // Keep the first full block plus 3 bytes of the next header.
-        let ln = (compressed[1] as usize) | ((compressed[2] as usize) << 8) | ((compressed[3] as usize) << 16);
-        let first_block_end = 8 + ln;
-        assert!(compressed.len() > first_block_end, "fixture must contain more than one block");
-        compressed.truncate(first_block_end + 3);
-
-        let mut decompress_reader = DecompressReader::new(Cursor::new(compressed), CompressionAlgorithm::default());
-        let mut out = Vec::new();
-        let err = decompress_reader
-            .read_to_end(&mut out)
-            .await
-            .expect_err("truncated header must error");
-        assert_eq!(err.kind(), std::io::ErrorKind::UnexpectedEof);
-    }
-
     // Regression: a corrupted block whose 24-bit length field is < 16 must not panic.
     // Header layout (HEADER_LEN = 8): [type, len_lo, len_mid, len_hi, crc0..crc3], then `len`
     // bytes of block body. Pre-fix, poll_read sliced `compressed_buf[0..16]` unconditionally,
@@ -749,85 +518,6 @@ mod tests {
         assert_eq!(res.unwrap_err().kind(), std::io::ErrorKind::InvalidData);
     }
 
-    // Header-level fail-closed matrix, built by hand so the decoder is exercised against bytes no
-    // encoder in this crate can produce. Header layout (HEADER_LEN = 8):
-    // [type, len_lo, len_mid, len_hi, crc0..crc3], then `len` body bytes = uvarint(plain_len) + data.
-    #[tokio::test]
-    async fn test_decompress_reader_header_validation_matrix() {
-        // Build a block whose body is `uvarint(plain.len()) + plain` (i.e. the
-        // COMPRESS_TYPE_UNCOMPRESSED shape), with the header CRC taken over the plaintext exactly
-        // like the production writer does.
-        fn build_raw_block(typ: u8, plain: &[u8], len_override: Option<usize>) -> Vec<u8> {
-            let crc = {
-                let mut hasher = crc_fast::Digest::new(crc_fast::CrcAlgorithm::Crc32IsoHdlc);
-                hasher.update(plain);
-                hasher.finalize() as u32
-            };
-            let mut uvarint_buf = [0u8; 10];
-            let int_len = put_uvarint(&mut uvarint_buf[..], plain.len() as u64);
-            let body_len = int_len + plain.len();
-            let len = len_override.unwrap_or(body_len);
-
-            let mut out = Vec::with_capacity(HEADER_LEN + body_len);
-            out.push(typ);
-            out.push((len & 0xFF) as u8);
-            out.push(((len >> 8) & 0xFF) as u8);
-            out.push(((len >> 16) & 0xFF) as u8);
-            out.extend_from_slice(&crc.to_le_bytes());
-            out.extend_from_slice(&uvarint_buf[..int_len]);
-            out.extend_from_slice(plain);
-            out
-        }
-
-        let plain = b"uncompressed passthrough payload";
-
-        // (a) A well-formed uncompressed block decodes to the plaintext verbatim.
-        let mut out = Vec::new();
-        DecompressReader::new(
-            Cursor::new(build_raw_block(COMPRESS_TYPE_UNCOMPRESSED, plain, None)),
-            CompressionAlgorithm::default(),
-        )
-        .read_to_end(&mut out)
-        .await
-        .expect("a well-formed uncompressed block must decode");
-        assert_eq!(out.as_slice(), plain.as_slice());
-
-        // (b) An unknown block type must be rejected instead of being treated as passthrough.
-        let mut out = Vec::new();
-        let err = DecompressReader::new(Cursor::new(build_raw_block(0x7E, plain, None)), CompressionAlgorithm::default())
-            .read_to_end(&mut out)
-            .await
-            .expect_err("unknown compression type must error");
-        assert_eq!(err.kind(), std::io::ErrorKind::InvalidData);
-        assert!(err.to_string().contains("Unknown compression type"), "got: {err}");
-
-        // (c) A zero-length block would stall the decoder, so it must be rejected up front.
-        let mut out = Vec::new();
-        let err = DecompressReader::new(
-            Cursor::new(build_raw_block(COMPRESS_TYPE_UNCOMPRESSED, plain, Some(0))),
-            CompressionAlgorithm::default(),
-        )
-        .read_to_end(&mut out)
-        .await
-        .expect_err("zero-length block must error");
-        assert_eq!(err.kind(), std::io::ErrorKind::InvalidData);
-        assert!(err.to_string().contains("Invalid compressed block length"), "got: {err}");
-
-        // (d) A block that decodes to zero plaintext bytes must be rejected: the
-        // writer never emits empty blocks, and an empty decode surfacing as
-        // Ready(Ok) with no bytes would read as EOF and silently truncate.
-        let mut out = Vec::new();
-        let err = DecompressReader::new(
-            Cursor::new(build_raw_block(COMPRESS_TYPE_UNCOMPRESSED, b"", None)),
-            CompressionAlgorithm::default(),
-        )
-        .read_to_end(&mut out)
-        .await
-        .expect_err("empty block must error");
-        assert_eq!(err.kind(), std::io::ErrorKind::InvalidData);
-        assert!(err.to_string().contains("Empty compressed block"), "got: {err}");
-    }
-
     // Directly exercises the length-prefix guard: an unterminated varint (all continuation bytes)
     // makes `uvarint` return 0, which must be rejected as an invalid length prefix.
     #[tokio::test]

crates/rio/src/hardlimit_reader.rs

@@ -24,17 +24,12 @@ pin_project! {
         #[pin]
         pub inner: R,
         remaining: i64,
-        scratch: Vec<u8>,
     }
 }
 
 impl<R> HardLimitReader<R> {
     pub fn new(inner: R, limit: i64) -> Self {
-        HardLimitReader {
-            inner,
-            remaining: limit,
-            scratch: Vec::new(),
-        }
+        HardLimitReader { inner, remaining: limit }
     }
 }
 
@@ -42,21 +37,19 @@ impl<R> AsyncRead for HardLimitReader<R>
 where
     R: AsyncRead,
 {
-    fn poll_read(self: Pin<&mut Self>, cx: &mut Context<'_>, buf: &mut ReadBuf<'_>) -> Poll<Result<()>> {
-        let mut this = self.project();
-        if *this.remaining < 0 {
+    fn poll_read(mut self: Pin<&mut Self>, cx: &mut Context<'_>, buf: &mut ReadBuf<'_>) -> Poll<Result<()>> {
+        if self.remaining < 0 {
             return Poll::Ready(Err(Error::other("input provided more bytes than specified")));
         }
-        if buf.remaining() == 0 {
-            return Poll::Ready(Ok(()));
-        }
-        if *this.remaining == 0 {
+        let original_filled = buf.filled().len();
+        if self.remaining == 0 {
             let mut discard = [0u8; 8192];
             let mut discard_buf = ReadBuf::new(&mut discard);
-            return match this.inner.as_mut().poll_read(cx, &mut discard_buf) {
+            return match self.as_mut().project().inner.poll_read(cx, &mut discard_buf) {
                 Poll::Pending => Poll::Pending,
                 Poll::Ready(Ok(())) => {
                     if discard_buf.filled().is_empty() {
+                        debug_assert_eq!(buf.filled().len(), original_filled);
                         Poll::Ready(Ok(()))
                     } else {
                         Poll::Ready(Err(Error::other("input provided more bytes than specified")))
@@ -65,46 +58,30 @@ where
                 Poll::Ready(Err(err)) => Poll::Ready(Err(err)),
             };
         }
+        // Save the initial length
+        let before = original_filled;
 
-        let remaining = match usize::try_from(*this.remaining) {
-            Ok(remaining) => remaining,
-            Err(_) => usize::MAX,
-        };
-        let allowed = remaining.min(buf.remaining());
-        let read = if allowed == buf.remaining() {
-            let before = buf.filled().len();
-            match this.inner.as_mut().poll_read(cx, buf) {
-                Poll::Pending => return Poll::Pending,
-                Poll::Ready(Err(err)) => return Poll::Ready(Err(err)),
-                Poll::Ready(Ok(())) => buf.filled().len() - before,
+        // Poll the inner reader
+        let this = self.as_mut().project();
+        let poll = this.inner.poll_read(cx, buf);
+
+        if let Poll::Ready(Ok(())) = &poll {
+            let after = buf.filled().len();
+            let read = (after - before) as i64;
+            if read == 0 && *this.remaining > 0 {
+                return Poll::Ready(Err(Error::new(
+                    std::io::ErrorKind::UnexpectedEof,
+                    IncompleteBody {
+                        remaining: *this.remaining,
+                    },
+                )));
             }
-        } else {
-            this.scratch.resize(allowed, 0);
-            let mut scratch_buf = ReadBuf::new(&mut this.scratch[..allowed]);
-            match this.inner.as_mut().poll_read(cx, &mut scratch_buf) {
-                Poll::Pending => return Poll::Pending,
-                Poll::Ready(Err(err)) => return Poll::Ready(Err(err)),
-                Poll::Ready(Ok(())) => {
-                    let read = scratch_buf.filled().len();
-                    buf.put_slice(scratch_buf.filled());
-                    read
-                }
+            *this.remaining -= read;
+            if *this.remaining < 0 {
+                return Poll::Ready(Err(Error::other("input provided more bytes than specified")));
             }
-        };
-        if read == 0 {
-            return Poll::Ready(Err(Error::new(
-                std::io::ErrorKind::UnexpectedEof,
-                IncompleteBody {
-                    remaining: *this.remaining,
-                },
-            )));
         }
-        let read = match i64::try_from(read) {
-            Ok(read) => read,
-            Err(_) => return Poll::Ready(Err(Error::other("read count exceeds i64::MAX"))),
-        };
-        *this.remaining -= read;
-        Poll::Ready(Ok(()))
+        poll
     }
 }
 
@@ -163,12 +140,7 @@ mod tests {
         assert!(err.is_some());
 
         let err = err.unwrap();
-        assert_eq!(err.kind(), std::io::ErrorKind::UnexpectedEof);
-        assert!(
-            err.get_ref()
-                .and_then(|source| source.downcast_ref::<std::io::Error>())
-                .is_some_and(|source| source.to_string().contains("more bytes than specified"))
-        );
+        assert_eq!(err.kind(), std::io::ErrorKind::Other);
     }
 
     #[tokio::test]
@@ -183,17 +155,6 @@ mod tests {
         assert_eq!(&buf, data);
     }
 
-    #[tokio::test]
-    async fn test_hardlimit_reader_zero_capacity_read_does_not_consume_input() {
-        let mut reader = HardLimitReader::new(BufReader::new(&b"abc"[..]), 3);
-        let mut empty = [];
-
-        assert_eq!(reader.read(&mut empty).await.expect("zero-capacity read should succeed"), 0);
-        let mut out = Vec::new();
-        reader.read_to_end(&mut out).await.expect("input should remain readable");
-        assert_eq!(out, b"abc");
-    }
-
     #[tokio::test]
     async fn test_hardlimit_reader_short_input_returns_unexpected_eof() {
         let data = b"abc";
@@ -234,18 +195,4 @@ mod tests {
         assert_eq!(err.kind(), std::io::ErrorKind::Other);
         assert!(err.to_string().contains("more bytes than specified"));
     }
-
-    #[tokio::test]
-    async fn test_hardlimit_reader_caps_each_read_before_reporting_extra_bytes() {
-        let mut reader = HardLimitReader::new(BufReader::new(&b"abcdef"[..]), 3);
-        let mut out = Vec::new();
-
-        let err = reader
-            .read_to_end(&mut out)
-            .await
-            .expect_err("bytes beyond the declared limit must be rejected");
-
-        assert_eq!(out, b"abc");
-        assert!(err.to_string().contains("more bytes than specified"));
-    }
 }

crates/rio/src/http_reader.rs

@@ -138,12 +138,6 @@ impl std::fmt::Display for InternodeHttpErrorKind {
     }
 }
 
-#[derive(thiserror::Error, Debug, Clone, Copy, Eq, PartialEq)]
-#[error("internode body stalled for {timeout:?}")]
-pub struct BodyStalled {
-    pub timeout: Duration,
-}
-
 #[derive(Debug, Clone, Eq, PartialEq)]
 pub struct InternodeHttpRequestContext {
     method: String,
@@ -277,10 +271,6 @@ pub fn internode_http_timeout_error(method: &Method, url: &str) -> io::Error {
     internode_kind_error(method, url, internode_rpc_operation(url), InternodeHttpErrorKind::ConnectTimeout)
 }
 
-fn body_stalled_error(stall_timeout: Duration) -> io::Error {
-    Error::new(io::ErrorKind::TimedOut, BodyStalled { timeout: stall_timeout })
-}
-
 /// Clone an internode HTTP I/O error while retaining its structured classification.
 ///
 /// The underlying transport source is intentionally omitted because it is not
@@ -1095,7 +1085,10 @@ impl AsyncRead for HttpReader {
                     );
                     record_internode_stall_timeout(*this.track_internode_metrics, *this.internode_operation);
                     record_internode_error(*this.track_internode_metrics, *this.internode_operation);
-                    Poll::Ready(Err(body_stalled_error(stall_timeout)))
+                    Poll::Ready(Err(Error::new(
+                        io::ErrorKind::TimedOut,
+                        "HttpReader stall timeout: no data received before deadline",
+                    )))
                 } else {
                     Poll::Pending
                 }
@@ -1224,7 +1217,10 @@ impl ChunkReader for HttpChunkReader {
                         );
                         record_internode_stall_timeout(*this.track_internode_metrics, *this.internode_operation);
                         record_internode_error(*this.track_internode_metrics, *this.internode_operation);
-                        return Poll::Ready(Err(body_stalled_error(stall_timeout)));
+                        return Poll::Ready(Err(Error::new(
+                            io::ErrorKind::TimedOut,
+                            "HttpReader stall timeout: no data received before deadline",
+                        )));
                     }
                     return Poll::Pending;
                 }
@@ -2383,46 +2379,6 @@ mod tests {
             Err(err) => err,
         };
         assert_eq!(err.kind(), io::ErrorKind::TimedOut);
-        let stalled = err
-            .get_ref()
-            .and_then(|source| source.downcast_ref::<BodyStalled>())
-            .expect("stall timeout should retain typed body-stalled source");
-        assert_eq!(stalled.timeout, Duration::from_millis(20));
-
-        handle.abort();
-    }
-
-    #[tokio::test]
-    async fn http_chunk_reader_stall_timeout_retains_typed_source() {
-        let state = TestState::default();
-        let Some((base_url, handle)) = start_test_server(state).await else {
-            return;
-        };
-        let url = base_url.replace("/stream", "/stall");
-        let mut reader =
-            HttpChunkReader::new_with_stall_timeout(url, Method::GET, HeaderMap::new(), None, Some(Duration::from_millis(20)))
-                .await
-                .expect("chunk reader should open");
-
-        let first = std::future::poll_fn(|cx| Pin::new(&mut reader).poll_read_chunk(cx, 64))
-            .await
-            .expect("initial body chunk should arrive")
-            .expect("initial body chunk should not be EOF");
-        assert_eq!(first, b"hello"[..]);
-
-        let err = tokio::time::timeout(
-            Duration::from_secs(1),
-            std::future::poll_fn(|cx| Pin::new(&mut reader).poll_read_chunk(cx, 64)),
-        )
-        .await
-        .expect("stall timeout should wake chunk reader")
-        .expect_err("chunk reader should return a timeout error");
-        assert_eq!(err.kind(), io::ErrorKind::TimedOut);
-        let stalled = err
-            .get_ref()
-            .and_then(|source| source.downcast_ref::<BodyStalled>())
-            .expect("chunk stall timeout should retain typed body-stalled source");
-        assert_eq!(stalled.timeout, Duration::from_millis(20));
 
         handle.abort();
     }

crates/scanner/src/scanner_io.rs

@@ -4262,7 +4262,6 @@ mod tests {
             .delete_bucket(&bucket, &DeleteBucketOptions::default())
             .await
             .expect("bucket should be removed from the first pool only");
-        init_bucket_metadata_sys_for_scanner_tests(store.clone()).await;
 
         let ctx = CancellationToken::new();
         let budget = ScannerCycleBudget::new(&ctx, ScannerCycleBudgetConfig::default());

docs/architecture/compat-cleanup-register.md

@@ -33,7 +33,6 @@ for later deletion.
 - `tonic-013-status-render` peer RPC failure classification: internode failures that reach a node only as text (a peer's error_info payload, a status flattened through format!) are classified by matching the rendering of an Unavailable gRPC status. Releases up to 1.0.0-alpha.38 shipped tonic 0.13, which rendered that status as "status: Unavailable, message: ..."; tonic 0.14 renders it as "code: 'The service is currently unavailable', message: ...". Both forms are matched so an older peer's relayed text still marks an unreachable peer offline. Remove the tonic 0.13 form after the minimum supported RustFS peer version ships tonic 0.14 or later.
 - `rustfs-5063` pre-beta.9 Local KMS recovery: persisted Local KMS configs from beta.8 and earlier predate the explicit insecure-development flag, and encrypted key files use the legacy SHA-256 KDF. Remove the config fallback after supported upgrades have rewritten or explicitly resaved all pre-beta.9 configs with the development-default field, and remove the legacy KDF after supported upgrades have rewritten all pre-beta.9 Local KMS key files with explicit at-rest protection.
 - `sse-local-dek-json-v1` legacy local SSE DEK decoding: releases before the JSON envelope wrote wrapped DEKs as `base64(nonce):base64(ciphertext)`, so readers retain that decoder while all new writes use the versioned JSON envelope. Remove the colon decoder after the minimum supported direct-upgrade release writes JSON envelopes and migration tooling has rewritten every retained legacy object.
-- `multipart-compression-default-off-window` staged multipart disk-compression rollout: releases before the resumable legacy decompressor fail transient reads of compressed objects under mid-payload suspension, so multipart uploads advertise the compression marker only when RUSTFS_COMPRESSION_MULTIPART_ENABLED is set in addition to RUSTFS_COMPRESSION_ENABLED, keeping rolling upgrades from creating new compressed multipart objects while pre-fix nodes may still serve reads. Flip the default to enabled (and retire the extra switch) after the minimum supported direct-upgrade release ships the resumable decompressor.
 
 ## Review Checklist
 

docs/architecture/global-state-inventory.md

@@ -111,7 +111,7 @@ inventory. Generic function-local names such as `CACHE`, `LOCK`, `INIT`, and
 | `GET_OBJECT_BUFFER_THRESHOLD_WARNED`, `GET_READER_STREAM_BUFFER_SIZE_OVERRIDE`, function-local `ENABLED`, `OBJECT_SEEK_SUPPORT_THRESHOLD`, `OBJECT_SEEK_SUPPORT_CONCURRENCY_THRESHOLDS` | `rustfs/src/app/object_usecase.rs` | Cache or constant / owner-local cache | Object GET/seek tuning caches and warning guards stay private to object usecase helpers. |
 | `SUPPORTED_HEADERS` | `rustfs/src/storage/options.rs` | Cache or constant / owner-local constant | Supported-header lookup state stays private to storage option parsing. |
 | `AUDIT_TARGET_SPECS`, `NOTIFICATION_TARGET_SPECS` | `rustfs/src/admin/handlers/audit.rs`, `rustfs/src/admin/handlers/event.rs`, `rustfs/src/admin/handlers/plugins_instances.rs` | Cache or constant / owner-local constant | Admin target descriptor tables stay private to their handler owners. |
-| `SITE_REPLICATION_PEER_CLIENT` | `rustfs/src/admin/handlers/site_replication.rs` | Process-global owner-local cache | Site-replication peer client cache stays private to site-replication handlers. The state RMW transaction holds no process-local mutex — see `rustfs/src/admin/site_replication_state.rs`. |
+| `SITE_REPLICATION_PEER_CLIENT`, `SITE_REPLICATION_STATE_LOCK` | `rustfs/src/admin/handlers/site_replication.rs` | Process-global owner-local cache / guard | Site-replication peer client cache and state lock stay private to site-replication handlers. |
 | `AUDIT_MODULE_ENABLED`, `NOTIFY_MODULE_ENABLED`, `PERSISTED_NOTIFY_MODULE_ENABLED`, `PERSISTED_AUDIT_MODULE_ENABLED`, `PERSISTED_MODULE_SWITCH_CONFIGURED` | `rustfs/src/server/audit.rs`, `rustfs/src/server/event.rs`, `rustfs/src/server/module_switch.rs` | Process-global owner-local toggles | Audit/notify module snapshots stay private to the server module switch owners. |
 | `DELETE_TAIL_TOTAL`, `DELETE_CLEANUP_TOTAL`, `DELETE_REPLICATION_TOTAL`, `DELETE_NOTIFY_TOTAL` | `rustfs/src/delete_tail_activity.rs` | Process-global owner-local counters | Delete-tail activity counters stay private behind delete-tail activity helpers. |
 | `EMBEDDED_SERVER_STARTED` | `rustfs/src/startup_lifecycle.rs` | Process-global owner-local guard | Embedded startup single-start protection stays private to startup lifecycle. |

rustfs/src/admin/handlers/table_catalog/tests.rs

@@ -12,8 +12,8 @@ use datafusion::{
 use std::sync::Arc;
 
 use crate::table_catalog::test_support::{
-    TestCatalogObjectBackend as TestTableCatalogObjectBackend, TestCatalogObjectRecord, TestCatalogPublishPause,
-    TestTableCatalogStore, manifest_avro_bytes as test_manifest_avro_bytes,
+    TestCatalogObjectBackend as TestTableCatalogObjectBackend, TestCatalogObjectRecord,
+    manifest_avro_bytes as test_manifest_avro_bytes,
     manifest_avro_bytes_with_nullable_sequences as test_manifest_avro_bytes_with_nullable_sequences,
     manifest_list_avro_bytes as test_manifest_list_avro_bytes, manifest_list_avro_entries as test_manifest_list_avro_entries,
     table_metadata_json as test_table_metadata_json,
@@ -6328,96 +6328,181 @@ async fn row_level_conflict_rejects_changed_inherited_manifest_identity() {
     assert_eq!(unchanged.generation, current.generation);
 }
 
-/// Table-driven fold of the three commit-rejection cases whose bodies were
-/// identical apart from four literals (backlog#1837 PR3). Each row keeps its
-/// original manifest-list sequence, data-file name, manifest-entry snapshot
-/// id, and failure message, so no poison combination is lost.
 #[tokio::test]
-async fn row_level_conflict_rejects_stale_or_historical_manifest_sequences() {
-    // (case, manifest-list sequence, data-file suffix, manifest-entry snapshot id, expected failure)
-    let cases: &[(&str, i64, &str, i64, &str)] = &[
-        (
-            "stale-new-manifest-sequence",
-            1,
-            "11",
-            11,
-            "new manifest sequence must match the committed snapshot",
-        ),
-        (
-            "stale-added-entry-sequence",
-            2,
-            "11",
-            11,
-            "added file sequence must match the new manifest",
-        ),
-        (
-            "historical-change-in-new-manifest",
-            2,
-            "10",
-            10,
-            "new manifest must not claim a historical changed entry",
-        ),
-    ];
-
-    for (case, manifest_list_sequence, data_file_suffix, entry_snapshot_id, failure) in cases {
-        let store = TestTableCatalogStore::default();
-        let metadata_backend = TestTableCatalogObjectBackend::content_addressed();
-        let namespace = crate::table_catalog::Namespace::parse("analytics").expect("namespace should parse");
-        let created = create_standard_events_table(&store, &metadata_backend, &namespace).await;
-        let table_location = created.metadata["location"]
-            .as_str()
-            .expect("created metadata should have table location");
-        let current = store
-            .load_table("warehouse", "analytics", "events")
-            .await
-            .expect("table lookup should succeed")
-            .expect("table should exist");
-        let manifest_list = format!("{table_location}/metadata/snap-11.avro");
-        let manifest = format!("{table_location}/metadata/manifest-11.avro");
-        let data_file = format!("{table_location}/data/part-{data_file_suffix}.parquet");
-        seed_test_manifest_list(&metadata_backend, "warehouse", &manifest_list, &[&manifest], *manifest_list_sequence, 11).await;
-        seed_test_manifest(&metadata_backend, "warehouse", &manifest, &[(&data_file, 0, 1, *entry_snapshot_id, 1)]).await;
-        let append_request: RestCommitTableRequest = serde_json::from_value(serde_json::json!({
-            "updates": [
-                {
-                    "action": "add-snapshot",
-                    "snapshot": {
-                        "snapshot-id": 11,
-                        "sequence-number": 2,
-                        "timestamp-ms": 2234,
-                        "manifest-list": manifest_list,
-                        "summary": {
-                            "operation": "append"
-                        }
+async fn row_level_conflict_rejects_stale_new_manifest_sequence() {
+    let store = TestTableCatalogStore::default();
+    let metadata_backend = TestTableCatalogObjectBackend::content_addressed();
+    let namespace = crate::table_catalog::Namespace::parse("analytics").expect("namespace should parse");
+    let created = create_standard_events_table(&store, &metadata_backend, &namespace).await;
+    let table_location = created.metadata["location"]
+        .as_str()
+        .expect("created metadata should have table location");
+    let current = store
+        .load_table("warehouse", "analytics", "events")
+        .await
+        .expect("table lookup should succeed")
+        .expect("table should exist");
+    let manifest_list = format!("{table_location}/metadata/snap-11.avro");
+    let manifest = format!("{table_location}/metadata/manifest-11.avro");
+    let data_file = format!("{table_location}/data/part-11.parquet");
+    seed_test_manifest_list(&metadata_backend, "warehouse", &manifest_list, &[&manifest], 1, 11).await;
+    seed_test_manifest(&metadata_backend, "warehouse", &manifest, &[(&data_file, 0, 1, 11, 1)]).await;
+    let append_request: RestCommitTableRequest = serde_json::from_value(serde_json::json!({
+        "updates": [
+            {
+                "action": "add-snapshot",
+                "snapshot": {
+                    "snapshot-id": 11,
+                    "sequence-number": 2,
+                    "timestamp-ms": 2234,
+                    "manifest-list": manifest_list,
+                    "summary": {
+                        "operation": "append"
                     }
                 }
-            ]
-        }))
-        .expect("append request should parse");
+            }
+        ]
+    }))
+    .expect("append request should parse");
 
-        let Err(error) = commit_table_response(
-            &store,
-            &trusted_table_commit_backend(&metadata_backend),
-            "warehouse",
-            &namespace,
-            "events",
-            append_request,
-        )
+    let error = commit_table_response(
+        &store,
+        &trusted_table_commit_backend(&metadata_backend),
+        "warehouse",
+        &namespace,
+        "events",
+        append_request,
+    )
+    .await
+    .expect_err("new manifest sequence must match the committed snapshot");
+
+    assert_eq!(error.code(), &s3s::S3ErrorCode::InvalidRequest);
+    let unchanged = store
+        .load_table("warehouse", "analytics", "events")
         .await
-        else {
-            panic!("[{case}] {failure}");
-        };
+        .expect("table lookup should succeed")
+        .expect("table should still exist");
+    assert_eq!(unchanged.metadata_location, current.metadata_location);
+    assert_eq!(unchanged.version_token, current.version_token);
+    assert_eq!(unchanged.generation, current.generation);
+}
 
-        assert_eq!(error.code(), &s3s::S3ErrorCode::InvalidRequest, "[{case}] {failure}");
-        let unchanged = store
-            .load_table("warehouse", "analytics", "events")
-            .await
-            .expect("table lookup should succeed")
-            .expect("table should still exist");
-        assert_eq!(unchanged.metadata_location, current.metadata_location, "[{case}] {failure}");
-        assert_eq!(unchanged.version_token, current.version_token, "[{case}] {failure}");
-        assert_eq!(unchanged.generation, current.generation, "[{case}] {failure}");
-    }
+#[tokio::test]
+async fn row_level_conflict_rejects_stale_added_entry_sequence() {
+    let store = TestTableCatalogStore::default();
+    let metadata_backend = TestTableCatalogObjectBackend::content_addressed();
+    let namespace = crate::table_catalog::Namespace::parse("analytics").expect("namespace should parse");
+    let created = create_standard_events_table(&store, &metadata_backend, &namespace).await;
+    let table_location = created.metadata["location"]
+        .as_str()
+        .expect("created metadata should have table location");
+    let current = store
+        .load_table("warehouse", "analytics", "events")
+        .await
+        .expect("table lookup should succeed")
+        .expect("table should exist");
+    let manifest_list = format!("{table_location}/metadata/snap-11.avro");
+    let manifest = format!("{table_location}/metadata/manifest-11.avro");
+    let data_file = format!("{table_location}/data/part-11.parquet");
+    seed_test_manifest_list(&metadata_backend, "warehouse", &manifest_list, &[&manifest], 2, 11).await;
+    seed_test_manifest(&metadata_backend, "warehouse", &manifest, &[(&data_file, 0, 1, 11, 1)]).await;
+    let append_request: RestCommitTableRequest = serde_json::from_value(serde_json::json!({
+        "updates": [
+            {
+                "action": "add-snapshot",
+                "snapshot": {
+                    "snapshot-id": 11,
+                    "sequence-number": 2,
+                    "timestamp-ms": 2234,
+                    "manifest-list": manifest_list,
+                    "summary": {
+                        "operation": "append"
+                    }
+                }
+            }
+        ]
+    }))
+    .expect("append request should parse");
+
+    let error = commit_table_response(
+        &store,
+        &trusted_table_commit_backend(&metadata_backend),
+        "warehouse",
+        &namespace,
+        "events",
+        append_request,
+    )
+    .await
+    .expect_err("added file sequence must match the new manifest");
+
+    assert_eq!(error.code(), &s3s::S3ErrorCode::InvalidRequest);
+    let unchanged = store
+        .load_table("warehouse", "analytics", "events")
+        .await
+        .expect("table lookup should succeed")
+        .expect("table should still exist");
+    assert_eq!(unchanged.metadata_location, current.metadata_location);
+    assert_eq!(unchanged.version_token, current.version_token);
+    assert_eq!(unchanged.generation, current.generation);
+}
+
+#[tokio::test]
+async fn row_level_conflict_rejects_historical_change_in_new_manifest() {
+    let store = TestTableCatalogStore::default();
+    let metadata_backend = TestTableCatalogObjectBackend::content_addressed();
+    let namespace = crate::table_catalog::Namespace::parse("analytics").expect("namespace should parse");
+    let created = create_standard_events_table(&store, &metadata_backend, &namespace).await;
+    let table_location = created.metadata["location"]
+        .as_str()
+        .expect("created metadata should have table location");
+    let current = store
+        .load_table("warehouse", "analytics", "events")
+        .await
+        .expect("table lookup should succeed")
+        .expect("table should exist");
+    let manifest_list = format!("{table_location}/metadata/snap-11.avro");
+    let manifest = format!("{table_location}/metadata/manifest-11.avro");
+    let data_file = format!("{table_location}/data/part-10.parquet");
+    seed_test_manifest_list(&metadata_backend, "warehouse", &manifest_list, &[&manifest], 2, 11).await;
+    seed_test_manifest(&metadata_backend, "warehouse", &manifest, &[(&data_file, 0, 1, 10, 1)]).await;
+    let append_request: RestCommitTableRequest = serde_json::from_value(serde_json::json!({
+        "updates": [
+            {
+                "action": "add-snapshot",
+                "snapshot": {
+                    "snapshot-id": 11,
+                    "sequence-number": 2,
+                    "timestamp-ms": 2234,
+                    "manifest-list": manifest_list,
+                    "summary": {
+                        "operation": "append"
+                    }
+                }
+            }
+        ]
+    }))
+    .expect("append request should parse");
+
+    let error = commit_table_response(
+        &store,
+        &trusted_table_commit_backend(&metadata_backend),
+        "warehouse",
+        &namespace,
+        "events",
+        append_request,
+    )
+    .await
+    .expect_err("new manifest must not claim a historical changed entry");
+
+    assert_eq!(error.code(), &s3s::S3ErrorCode::InvalidRequest);
+    let unchanged = store
+        .load_table("warehouse", "analytics", "events")
+        .await
+        .expect("table lookup should succeed")
+        .expect("table should still exist");
+    assert_eq!(unchanged.metadata_location, current.metadata_location);
+    assert_eq!(unchanged.version_token, current.version_token);
+    assert_eq!(unchanged.generation, current.generation);
 }
 
 #[tokio::test]
@@ -7837,6 +7922,34 @@ fn commit_table_request_uses_rest_commit_fields() {
     assert_eq!(request.writer.as_deref(), Some("pyiceberg"));
 }
 
+#[derive(Default)]
+struct TestTableCatalogStore {
+    table_buckets: tokio::sync::Mutex<Vec<crate::table_catalog::TableBucketEntry>>,
+    namespaces: tokio::sync::Mutex<Vec<crate::table_catalog::NamespaceEntry>>,
+    tables: tokio::sync::Mutex<Vec<crate::table_catalog::TableEntry>>,
+    views: tokio::sync::Mutex<Vec<crate::table_catalog::ViewEntry>>,
+    commits: tokio::sync::Mutex<Vec<crate::table_catalog::CommitLogEntry>>,
+    fail_put_table_bucket: tokio::sync::Mutex<bool>,
+    register_table_pause: Option<TestCatalogPublishPause>,
+    commit_table_pause: Option<TestCatalogPublishPause>,
+}
+
+#[derive(Clone, Default)]
+struct TestCatalogPublishPause {
+    started: Arc<tokio::sync::Notify>,
+    release: Arc<tokio::sync::Notify>,
+}
+
+impl TestCatalogPublishPause {
+    async fn wait_started(&self) {
+        self.started.notified().await;
+    }
+
+    fn release(&self) {
+        self.release.notify_one();
+    }
+}
+
 fn trusted_table_commit_backend(
     backend: &TestTableCatalogObjectBackend,
 ) -> TableCommitObjectBackend<TestTableCatalogObjectBackend> {
@@ -8205,6 +8318,410 @@ async fn seed_object_table_for_metadata_maintenance(
         .await;
 }
 
+#[async_trait::async_trait]
+impl crate::table_catalog::TableCatalogStore for TestTableCatalogStore {
+    async fn get_table_bucket(
+        &self,
+        table_bucket: &str,
+    ) -> crate::table_catalog::TableCatalogStoreResult<Option<crate::table_catalog::TableBucketEntry>> {
+        Ok(self
+            .table_buckets
+            .lock()
+            .await
+            .iter()
+            .find(|entry| entry.table_bucket == table_bucket)
+            .cloned())
+    }
+
+    async fn put_table_bucket(
+        &self,
+        entry: crate::table_catalog::TableBucketEntry,
+    ) -> crate::table_catalog::TableCatalogStoreResult<()> {
+        let mut fail_put_table_bucket = self.fail_put_table_bucket.lock().await;
+        if *fail_put_table_bucket {
+            *fail_put_table_bucket = false;
+            return Err(crate::table_catalog::TableCatalogStoreError::Internal(
+                "injected table bucket write failure".to_string(),
+            ));
+        }
+        drop(fail_put_table_bucket);
+
+        let mut table_buckets = self.table_buckets.lock().await;
+        table_buckets.retain(|existing| existing.table_bucket != entry.table_bucket);
+        table_buckets.push(entry);
+        Ok(())
+    }
+
+    async fn create_namespace(
+        &self,
+        entry: crate::table_catalog::NamespaceEntry,
+    ) -> crate::table_catalog::TableCatalogStoreResult<()> {
+        if self.get_table_bucket(&entry.table_bucket).await?.is_none() {
+            return Err(crate::table_catalog::TableCatalogStoreError::NotFound(format!(
+                "table bucket {}",
+                entry.table_bucket
+            )));
+        }
+        self.namespaces.lock().await.push(entry);
+        Ok(())
+    }
+
+    async fn list_namespaces(
+        &self,
+        table_bucket: &str,
+    ) -> crate::table_catalog::TableCatalogStoreResult<Vec<crate::table_catalog::NamespaceEntry>> {
+        Ok(self
+            .namespaces
+            .lock()
+            .await
+            .iter()
+            .filter(|entry| entry.table_bucket == table_bucket)
+            .cloned()
+            .collect())
+    }
+
+    async fn get_namespace(
+        &self,
+        table_bucket: &str,
+        namespace: &str,
+    ) -> crate::table_catalog::TableCatalogStoreResult<Option<crate::table_catalog::NamespaceEntry>> {
+        Ok(self
+            .namespaces
+            .lock()
+            .await
+            .iter()
+            .find(|entry| entry.table_bucket == table_bucket && entry.namespace == namespace)
+            .cloned())
+    }
+
+    async fn update_namespace_properties(
+        &self,
+        table_bucket: &str,
+        namespace: &str,
+        update: crate::table_catalog::NamespacePropertiesUpdate,
+    ) -> crate::table_catalog::TableCatalogStoreResult<crate::table_catalog::NamespacePropertiesUpdateResult> {
+        let mut namespaces = self.namespaces.lock().await;
+        let entry = namespaces
+            .iter_mut()
+            .find(|entry| entry.table_bucket == table_bucket && entry.namespace == namespace)
+            .ok_or_else(|| {
+                crate::table_catalog::TableCatalogStoreError::NotFound(format!("namespace {table_bucket}/{namespace}"))
+            })?;
+        Ok(update.apply_to(entry))
+    }
+
+    async fn drop_namespace(&self, table_bucket: &str, namespace: &str) -> crate::table_catalog::TableCatalogStoreResult<()> {
+        self.namespaces
+            .lock()
+            .await
+            .retain(|entry| !(entry.table_bucket == table_bucket && entry.namespace == namespace));
+        Ok(())
+    }
+
+    async fn create_table(&self, entry: crate::table_catalog::TableEntry) -> crate::table_catalog::TableCatalogStoreResult<()> {
+        if self.get_table_bucket(&entry.table_bucket).await?.is_none() {
+            return Err(crate::table_catalog::TableCatalogStoreError::NotFound(format!(
+                "table bucket {}",
+                entry.table_bucket
+            )));
+        }
+        if self.get_namespace(&entry.table_bucket, &entry.namespace).await?.is_none() {
+            return Err(crate::table_catalog::TableCatalogStoreError::NotFound(format!(
+                "namespace {}/{}",
+                entry.table_bucket, entry.namespace
+            )));
+        }
+        self.tables.lock().await.push(entry);
+        Ok(())
+    }
+
+    async fn register_table(&self, entry: crate::table_catalog::TableEntry) -> crate::table_catalog::TableCatalogStoreResult<()> {
+        if self.get_table_bucket(&entry.table_bucket).await?.is_none() {
+            return Err(crate::table_catalog::TableCatalogStoreError::NotFound(format!(
+                "table bucket {}",
+                entry.table_bucket
+            )));
+        }
+        if self.get_namespace(&entry.table_bucket, &entry.namespace).await?.is_none() {
+            return Err(crate::table_catalog::TableCatalogStoreError::NotFound(format!(
+                "namespace {}/{}",
+                entry.table_bucket, entry.namespace
+            )));
+        }
+        if let Some(pause) = &self.register_table_pause {
+            pause.started.notify_one();
+            pause.release.notified().await;
+        }
+        self.tables.lock().await.push(entry);
+        Ok(())
+    }
+
+    async fn register_table_with_publication(
+        &self,
+        entry: crate::table_catalog::TableEntry,
+        publication: &(dyn crate::table_catalog::TableCommitPublication + Sync),
+    ) -> crate::table_catalog::TableCatalogStoreResult<()> {
+        publication.begin_table_bucket(&entry.table_bucket).await?;
+        if !publication.holds_table_bucket(&entry.table_bucket) {
+            return Err(crate::table_catalog::TableCatalogStoreError::Internal(
+                "table registration requires a table-bucket publication fence".to_string(),
+            ));
+        }
+        let _publication_completion = crate::table_catalog::TableCommitPublicationCompletion::new(publication);
+        publication
+            .prepare(&entry.table_bucket, &entry.namespace, &entry.table)
+            .await?;
+        if !publication.holds_table(&entry.table_bucket, &entry.namespace, &entry.table) {
+            return Err(crate::table_catalog::TableCatalogStoreError::Internal(
+                "table registration requires a table publication fence".to_string(),
+            ));
+        }
+        self.register_table(entry).await
+    }
+
+    async fn list_tables(
+        &self,
+        table_bucket: &str,
+        namespace: &str,
+    ) -> crate::table_catalog::TableCatalogStoreResult<Vec<crate::table_catalog::TableEntry>> {
+        Ok(self
+            .tables
+            .lock()
+            .await
+            .iter()
+            .filter(|entry| entry.table_bucket == table_bucket && entry.namespace == namespace)
+            .cloned()
+            .collect())
+    }
+
+    async fn list_all_tables(
+        &self,
+        table_bucket: &str,
+    ) -> crate::table_catalog::TableCatalogStoreResult<Vec<crate::table_catalog::TableEntry>> {
+        Ok(self
+            .tables
+            .lock()
+            .await
+            .iter()
+            .filter(|entry| entry.table_bucket == table_bucket)
+            .cloned()
+            .collect())
+    }
+
+    async fn load_table(
+        &self,
+        table_bucket: &str,
+        namespace: &str,
+        table: &str,
+    ) -> crate::table_catalog::TableCatalogStoreResult<Option<crate::table_catalog::TableEntry>> {
+        Ok(self
+            .tables
+            .lock()
+            .await
+            .iter()
+            .find(|entry| entry.table_bucket == table_bucket && entry.namespace == namespace && entry.table == table)
+            .cloned())
+    }
+
+    async fn commit_table(
+        &self,
+        request: crate::table_catalog::TableCommitRequest,
+    ) -> crate::table_catalog::TableCatalogStoreResult<crate::table_catalog::TableCommitResult> {
+        let mut tables = self.tables.lock().await;
+        let Some(index) = tables.iter().position(|entry| {
+            entry.table_bucket == request.table_bucket && entry.namespace == request.namespace && entry.table == request.table
+        }) else {
+            return Err(crate::table_catalog::TableCatalogStoreError::NotFound(format!(
+                "table {}/{}/{}",
+                request.table_bucket, request.namespace, request.table
+            )));
+        };
+
+        let current = tables[index].clone();
+        if current.version_token != request.expected_version_token {
+            return Err(crate::table_catalog::TableCatalogStoreError::Conflict(
+                "current table version token does not match expected token".to_string(),
+            ));
+        }
+        if current.metadata_location != request.expected_metadata_location {
+            return Err(crate::table_catalog::TableCatalogStoreError::Conflict(
+                "current table metadata location does not match expected location".to_string(),
+            ));
+        }
+        if let Some(pause) = &self.commit_table_pause {
+            pause.started.notify_one();
+            pause.release.notified().await;
+        }
+
+        let mut next = current.clone();
+        next.metadata_location = request.new_metadata_location.clone();
+        next.version_token = "token-committed".to_string();
+        next.generation = next.generation.saturating_add(1);
+        tables[index] = next.clone();
+        drop(tables);
+
+        let commit_log = crate::table_catalog::CommitLogEntry {
+            version: crate::table_catalog::TABLE_CATALOG_ENTRY_VERSION,
+            commit_id: request.commit_id,
+            idempotency_key: request.idempotency_key,
+            table_id: current.table_id,
+            operation: request.operation,
+            expected_version_token: request.expected_version_token,
+            new_version_token: next.version_token.clone(),
+            previous_metadata_location: request.expected_metadata_location,
+            new_metadata_location: request.new_metadata_location,
+            requirements: request.requirements,
+            status: crate::table_catalog::CommitLogStatus::Committed,
+            writer: request.writer,
+            created_at: None,
+            updated_at: None,
+        };
+        self.commits.lock().await.push(commit_log.clone());
+
+        Ok(crate::table_catalog::TableCommitResult { table: next, commit_log })
+    }
+
+    async fn commit_table_with_publication(
+        &self,
+        request: crate::table_catalog::TableCommitRequest,
+        publication: &(dyn crate::table_catalog::TableCommitPublication + Sync),
+    ) -> crate::table_catalog::TableCatalogStoreResult<crate::table_catalog::TableCommitResult> {
+        publication
+            .prepare(&request.table_bucket, &request.namespace, &request.table)
+            .await?;
+        if !publication.holds_table(&request.table_bucket, &request.namespace, &request.table) {
+            return Err(crate::table_catalog::TableCatalogStoreError::Internal(
+                "table commit requires a table publication fence".to_string(),
+            ));
+        }
+        let _publication_completion = crate::table_catalog::TableCommitPublicationCompletion::new(publication);
+        self.commit_table(request).await
+    }
+
+    async fn drop_table(
+        &self,
+        table_bucket: &str,
+        namespace: &str,
+        table: &str,
+    ) -> crate::table_catalog::TableCatalogStoreResult<()> {
+        self.tables
+            .lock()
+            .await
+            .retain(|entry| !(entry.table_bucket == table_bucket && entry.namespace == namespace && entry.table == table));
+        Ok(())
+    }
+
+    async fn create_view(&self, entry: crate::table_catalog::ViewEntry) -> crate::table_catalog::TableCatalogStoreResult<()> {
+        if self.get_table_bucket(&entry.table_bucket).await?.is_none() {
+            return Err(crate::table_catalog::TableCatalogStoreError::NotFound(format!(
+                "table bucket {}",
+                entry.table_bucket
+            )));
+        }
+        if self.get_namespace(&entry.table_bucket, &entry.namespace).await?.is_none() {
+            return Err(crate::table_catalog::TableCatalogStoreError::NotFound(format!(
+                "namespace {}/{}",
+                entry.table_bucket, entry.namespace
+            )));
+        }
+        self.views.lock().await.push(entry);
+        Ok(())
+    }
+
+    async fn list_views(
+        &self,
+        table_bucket: &str,
+        namespace: &str,
+    ) -> crate::table_catalog::TableCatalogStoreResult<Vec<crate::table_catalog::ViewEntry>> {
+        Ok(self
+            .views
+            .lock()
+            .await
+            .iter()
+            .filter(|entry| entry.table_bucket == table_bucket && entry.namespace == namespace)
+            .cloned()
+            .collect())
+    }
+
+    async fn load_view(
+        &self,
+        table_bucket: &str,
+        namespace: &str,
+        view: &str,
+    ) -> crate::table_catalog::TableCatalogStoreResult<Option<crate::table_catalog::ViewEntry>> {
+        Ok(self
+            .views
+            .lock()
+            .await
+            .iter()
+            .find(|entry| entry.table_bucket == table_bucket && entry.namespace == namespace && entry.view == view)
+            .cloned())
+    }
+
+    async fn replace_view(
+        &self,
+        request: crate::table_catalog::ViewCommitRequest,
+    ) -> crate::table_catalog::TableCatalogStoreResult<crate::table_catalog::ViewCommitResult> {
+        let mut views = self.views.lock().await;
+        let Some(index) = views.iter().position(|entry| {
+            entry.table_bucket == request.table_bucket && entry.namespace == request.namespace && entry.view == request.view
+        }) else {
+            return Err(crate::table_catalog::TableCatalogStoreError::NotFound(format!(
+                "view {}/{}/{}",
+                request.table_bucket, request.namespace, request.view
+            )));
+        };
+        let current = views[index].clone();
+        if current.version_token != request.expected_version_token {
+            return Err(crate::table_catalog::TableCatalogStoreError::Conflict(
+                "current view version token does not match expected token".to_string(),
+            ));
+        }
+        if current.metadata_location != request.expected_metadata_location {
+            return Err(crate::table_catalog::TableCatalogStoreError::Conflict(
+                "current view metadata location does not match expected location".to_string(),
+            ));
+        }
+        let mut next = current;
+        next.metadata_location = request.new_metadata_location;
+        next.version_token = "token-view-committed".to_string();
+        next.generation = next.generation.saturating_add(1);
+        views[index] = next.clone();
+        Ok(crate::table_catalog::ViewCommitResult { view: next })
+    }
+
+    async fn drop_view(
+        &self,
+        table_bucket: &str,
+        namespace: &str,
+        view: &str,
+    ) -> crate::table_catalog::TableCatalogStoreResult<()> {
+        self.views
+            .lock()
+            .await
+            .retain(|entry| !(entry.table_bucket == table_bucket && entry.namespace == namespace && entry.view == view));
+        Ok(())
+    }
+
+    async fn get_commit_by_id(
+        &self,
+        _table_bucket: &str,
+        _table_id: &str,
+        _commit_id: &str,
+    ) -> crate::table_catalog::TableCatalogStoreResult<Option<crate::table_catalog::CommitLogEntry>> {
+        Ok(None)
+    }
+
+    async fn get_commit_by_idempotency_key(
+        &self,
+        _table_bucket: &str,
+        _table_id: &str,
+        _idempotency_key: &str,
+    ) -> crate::table_catalog::TableCatalogStoreResult<Option<crate::table_catalog::CommitLogEntry>> {
+        Ok(None)
+    }
+}
+
 #[tokio::test]
 async fn ensure_table_bucket_entry_seeds_enabled_bucket_before_namespace_create() {
     let store = TestTableCatalogStore::default();

rustfs/src/admin/site_replication_state.rs

@@ -18,21 +18,28 @@
 //! `config/site-replication/state.json` is mutated by read-modify-write
 //! sequences spread over many call sites: admin handlers, the retry-event
 //! writers on every hook broadcast path, and the service-side reload driven
-//! over node RPC.
+//! over node RPC. Historically only some of them held the process-local
+//! mutex and none held a distributed lock across the whole RMW, so
+//! concurrent writers overwrote each other (single-process for the unlocked
+//! writers, cross-node for everyone).
 //!
 //! `with_site_replication_state_lock` is the single transaction boundary:
-//! it holds the distributed config-object write lock (the pattern proven by
-//! the repair state, `update_site_replication_repair_state`) for the
-//! duration of the caller's closure. The object lock is the sole mechanism —
-//! it is the only thing that can serialize two nodes of the same site, so a
-//! process-local lock must never be reintroduced in front of it as if it
-//! added protection. All IO inside the closure must use the `*_no_lock`
-//! config helpers — the locked variants would self-deadlock on the same
-//! object lock. Do not perform peer network calls or take other config locks
+//! it holds the process-local mutex AND the distributed config-object write
+//! lock (the pattern proven by the repair state,
+//! `update_site_replication_repair_state`) for the duration of the caller's
+//! closure. All IO inside the closure must use the `*_no_lock` config
+//! helpers — the locked variants would self-deadlock on the same object
+//! lock. Do not perform peer network calls or take other config locks
 //! inside the closure.
 //!
-//! Lock order: lifecycle -> bucket operation -> repair admission
-//! -> state object lock -> per-bucket metadata.
+//! The process-local mutex is transitional: call sites still outside this
+//! primitive serialize against migrated ones through it. Once every RMW
+//! call site goes through here (P1-15 PR2) it will be removed, leaving the
+//! object lock as the only mechanism.
+//!
+//! Lock order (unchanged from the historical comment next to the mutex):
+//! lifecycle -> bucket operation -> repair admission -> state (process
+//! mutex, then state object lock) -> per-bucket metadata.
 
 use crate::admin::storage_api::runtime::ECStore;
 use crate::admin::storage_api::s3::{S3Error, S3ErrorCode, S3Result};
@@ -46,8 +53,24 @@ use super::runtime_sources::current_object_store_handle;
 /// byte-level tolerant reload on the service side.
 pub(crate) const SITE_REPLICATION_STATE_PATH: &str = "config/site-replication/state.json";
 
+/// Transitional process-local mutex — see the module docs. Stays private to
+/// this module (owner-local static, enforced by
+/// `scripts/check_architecture_migration_rules.sh`); callers go through
+/// [`site_replication_state_process_guard`].
+static SITE_REPLICATION_STATE_LOCK: std::sync::LazyLock<tokio::sync::Mutex<()>> =
+    std::sync::LazyLock::new(|| tokio::sync::Mutex::new(()));
+
+/// Owner helper for the transitional process mutex: the RMW call sites in
+/// `handlers::site_replication` that PR2 has not migrated to
+/// [`with_site_replication_state_lock`] yet hold this guard so they stay
+/// mutually exclusive with the migrated ones. Removed together with the
+/// mutex once every call site runs inside the transaction boundary.
+pub(crate) async fn site_replication_state_process_guard() -> tokio::sync::MutexGuard<'static, ()> {
+    SITE_REPLICATION_STATE_LOCK.lock().await
+}
+
 /// Run `operation` under the site-replication state transaction boundary:
-/// the distributed state-object write lock.
+/// process mutex first, then the distributed state-object write lock.
 pub(crate) async fn with_site_replication_state_lock<T, F, Fut>(operation: F) -> S3Result<T>
 where
     T: Send + 'static,
@@ -60,11 +83,21 @@ where
 
 /// Context-store variant for callers that resolve their store from an
 /// explicit [`AppContext`] (the service-side reload driven over node RPC).
-///
-/// This is the whole boundary: a state-object write lock serializes writers
-/// in *different* processes, which is what two nodes of one site are and
-/// what a process mutex could never cover.
 pub(crate) async fn with_site_replication_state_lock_on<T, F, Fut>(store: Arc<ECStore>, operation: F) -> S3Result<T>
+where
+    T: Send + 'static,
+    F: FnOnce() -> Fut + Send + 'static,
+    Fut: std::future::Future<Output = S3Result<T>> + Send + 'static,
+{
+    let _process_guard = SITE_REPLICATION_STATE_LOCK.lock().await;
+    with_site_replication_state_object_lock(store, operation).await
+}
+
+/// The distributed half of the boundary on its own: the state-object write
+/// lock, without the process mutex. This is the only thing that serializes
+/// writers in *different* processes (the mutex cannot), so it is also what
+/// the separate-nodes regression test drives.
+pub(crate) async fn with_site_replication_state_object_lock<T, F, Fut>(store: Arc<ECStore>, operation: F) -> S3Result<T>
 where
     T: Send + 'static,
     F: FnOnce() -> Fut + Send + 'static,

rustfs/src/app/multipart_usecase.rs

@@ -27,7 +27,6 @@ use super::storage_api::multipart_usecase::bucket::{
     replication::{must_replicate_object, schedule_object_replication},
     versioning_sys::BucketVersioningSys,
 };
-use super::storage_api::multipart_usecase::compression::{is_disk_compressible, is_multipart_disk_compression_enabled};
 #[cfg(test)]
 use super::storage_api::multipart_usecase::contract::http::HTTPPreconditions;
 use super::storage_api::multipart_usecase::contract::multipart::{CompletePart, MultipartOperations as _, MultipartUploadResult};
@@ -40,7 +39,7 @@ use super::storage_api::multipart_usecase::error::{StorageError, is_err_object_n
 use super::storage_api::multipart_usecase::helper::OperationHelper;
 #[cfg(test)]
 use super::storage_api::multipart_usecase::io::{DecryptReader, EncryptReader, HardLimitReader, boxed_reader, wrap_reader};
-use super::storage_api::multipart_usecase::io::{HashReader, WriteEncryption, WritePlan, compression_metadata_value};
+use super::storage_api::multipart_usecase::io::{HashReader, WriteEncryption, WritePlan};
 use super::storage_api::multipart_usecase::object_utils::to_s3s_etag;
 use super::storage_api::multipart_usecase::options::{
     copy_src_opts, extract_metadata_from_mime, get_complete_multipart_upload_opts_with_replication_authorization,
@@ -211,28 +210,6 @@ fn create_multipart_upload_metadata(
     metadata
 }
 
-/// A multipart session advertises disk compression only when the staged-rollout
-/// switch (`RUSTFS_COMPRESSION_MULTIPART_ENABLED`) is on, the object key/headers
-/// qualify, AND the session is not an SSE-C ciphertext-passthrough replication
-/// session, which must preserve source bytes verbatim.
-///
-/// The rollout switch defaults to off so a rolling upgrade never creates new
-/// compressed multipart objects while pre-fix nodes (whose decompressor is not
-/// resumable) may still serve reads. Enable it once the fleet has converged on a
-/// fixed build; the default flips per the `multipart-compression-default-off-window`
-/// entry in docs/architecture/compat-cleanup-register.md.
-///
-/// Each part is compressed as an independent stream; the GET path decodes across part
-/// boundaries (see `ReadTransform::Compressed`), so the session may advertise
-/// object-level compression again.
-///
-/// Unlike single PUT there is no `MIN_DISK_COMPRESSIBLE_SIZE` floor here: the total
-/// object size is unknown at CreateMultipartUpload time, so tiny multipart objects pay
-/// the (harmless) framing overhead. This is a deliberate trade-off, not a bug.
-fn should_advertise_session_compression(multipart_enabled: bool, ciphertext_passthrough: bool, disk_compressible: bool) -> bool {
-    multipart_enabled && !ciphertext_passthrough && disk_compressible
-}
-
 async fn validate_table_catalog_object_mutation(bucket: &str, key: &str) -> S3Result<()> {
     table_catalog::validate_bucket_object_mutation(bucket, key)
         .await
@@ -860,17 +837,8 @@ impl DefaultMultipartUsecase {
             None => (None, None),
         };
 
-        if should_advertise_session_compression(
-            is_multipart_disk_compression_enabled(),
-            ciphertext_passthrough,
-            is_disk_compressible(&req.headers, &key),
-        ) {
-            rustfs_utils::http::insert_str(
-                &mut metadata,
-                rustfs_utils::http::SUFFIX_COMPRESSION,
-                compression_metadata_value(CompressionAlgorithm::default()),
-            );
-        }
+        // Multipart parts are independent physical streams. Advertising object-level
+        // compression here would make GET decode the completed object as one stream.
 
         let mt2 = metadata.clone();
         let mut opts: ObjectOptions =
@@ -1664,31 +1632,6 @@ mod tests {
         DefaultMultipartUsecase::without_context()
     }
 
-    #[test]
-    fn session_compression_is_advertised_only_for_non_passthrough_compressible_uploads() {
-        // (multipart_enabled, ciphertext_passthrough, disk_compressible, expected)
-        let cases = [
-            (true, false, false, false),
-            (true, false, true, true),
-            (true, true, false, false),
-            (true, true, true, false),
-            // The staged-rollout switch keeps multipart compression dark by
-            // default regardless of the other gates.
-            (false, false, true, false),
-            (false, false, false, false),
-            (false, true, true, false),
-            (false, true, false, false),
-        ];
-
-        for (multipart_enabled, ciphertext_passthrough, disk_compressible, expected) in cases {
-            assert_eq!(
-                should_advertise_session_compression(multipart_enabled, ciphertext_passthrough, disk_compressible),
-                expected,
-                "multipart_enabled={multipart_enabled} ciphertext_passthrough={ciphertext_passthrough} disk_compressible={disk_compressible}"
-            );
-        }
-    }
-
     #[test]
     fn quota_accounting_uses_logical_size_when_available() {
         let mut metadata = HashMap::new();

rustfs/src/app/object_usecase.rs

@@ -2241,11 +2241,10 @@ fn get_object_resume_control(ctx: GetObjectResumeContext) -> GetObjectResumeCont
 /// disks" failures keep the existing fail-loud behavior.
 fn is_object_relocation_error(err: &std::io::Error) -> bool {
     let Some(inner) = err.get_ref() else { return false };
-    match inner.downcast_ref::<StorageError>() {
-        Some(StorageError::FileNotFound | StorageError::ObjectNotFound(..) | StorageError::InsufficientReadQuorum(..)) => true,
-        Some(StorageError::Io(source)) => source.kind() == std::io::ErrorKind::NotFound,
-        _ => false,
-    }
+    matches!(
+        inner.downcast_ref::<StorageError>(),
+        Some(StorageError::FileNotFound | StorageError::ObjectNotFound(..) | StorageError::InsufficientReadQuorum(..))
+    )
 }
 
 /// Resolve the S3 request-body inter-chunk read timeout from the environment.
@@ -13164,7 +13163,6 @@ mod tests {
             StorageError::FileNotFound,
             StorageError::ObjectNotFound("test-bucket".to_string(), "relocated-object".to_string()),
             StorageError::InsufficientReadQuorum("test-bucket".to_string(), "relocated-object".to_string()),
-            StorageError::Io(std::io::Error::new(std::io::ErrorKind::NotFound, "relocated shard disappeared")),
         ] {
             let reopen_count = Arc::new(AtomicUsize::new(0));
             let control = counting_resume_control(Arc::clone(&reopen_count), |emitted| {

rustfs/src/app/storage_api.rs

@@ -942,9 +942,7 @@ pub(crate) mod concurrency {
 }
 
 pub(crate) mod compression {
-    pub(crate) use crate::storage::storage_api::ecstore_compression::{
-        MIN_DISK_COMPRESSIBLE_SIZE, is_disk_compressible, is_multipart_disk_compression_enabled,
-    };
+    pub(crate) use crate::storage::storage_api::ecstore_compression::{MIN_DISK_COMPRESSIBLE_SIZE, is_disk_compressible};
 }
 
 pub(crate) mod deadlock_detector {
@@ -1155,7 +1153,7 @@ pub(crate) mod multipart_usecase {
     }
 
     pub(crate) use super::{
-        access, bucket, compression, data_usage, error, helper, io, object_utils, options, request_context, s3_api, set_disk, sse,
+        access, bucket, data_usage, error, helper, io, object_utils, options, request_context, s3_api, set_disk, sse,
     };
     pub(crate) use crate::storage::storage_api::{ECStore, StorageObjectInfo, StorageObjectOptions, StoragePutObjReader};
 }

rustfs/src/error.rs

@@ -326,11 +326,7 @@ impl From<StorageError> for ApiError {
             _ => S3ErrorCode::InternalError,
         };
 
-        let message = if matches!(&err, StorageError::QuotaExceeded { .. }) {
-            err.to_string()
-        } else if code == S3ErrorCode::InternalError && matches!(&err, StorageError::Io(_)) {
-            ApiError::error_code_to_message(&code)
-        } else if code == S3ErrorCode::InternalError {
+        let message = if matches!(&err, StorageError::QuotaExceeded { .. }) || code == S3ErrorCode::InternalError {
             err.to_string()
         } else if let StorageError::InvalidArgument(_, _, reason) = &err
             && !reason.is_empty()
@@ -529,25 +525,6 @@ mod tests {
         assert!(api_error.source.is_some());
     }
 
-    #[test]
-    fn storage_io_internal_error_redacts_public_message_and_retains_source() {
-        let sensitive_path = "/sensitive/storage/path";
-        let api_error = ApiError::from(StorageError::Io(IoError::new(
-            ErrorKind::PermissionDenied,
-            format!("permission denied: {sensitive_path}"),
-        )));
-
-        assert_eq!(api_error.code, S3ErrorCode::InternalError);
-        assert_eq!(api_error.message, ApiError::error_code_to_message(&S3ErrorCode::InternalError));
-        assert!(!api_error.message.contains(sensitive_path));
-        let source = api_error
-            .source
-            .as_deref()
-            .and_then(|source| source.downcast_ref::<StorageError>())
-            .expect("API error should retain the storage error source");
-        assert!(matches!(source, StorageError::Io(io_error) if io_error.to_string().contains(sensitive_path)));
-    }
-
     #[test]
     fn test_kms_service_unavailable_maps_to_retryable_error() {
         let api_error = ApiError::from(StorageError::other(KmsUnavailableError));
@@ -692,36 +669,14 @@ mod tests {
         assert!(api_error.source.is_some());
     }
 
-    #[test]
-    fn test_api_error_from_storage_io_copy_object_terminal_error_stays_internal() {
-        let io_error = IoError::other(StorageError::FileCorrupt);
-        let storage_error: StorageError = io_error.into();
-        assert!(matches!(storage_error, StorageError::FileCorrupt));
-
-        let api_error: ApiError = storage_error.into();
-
-        assert_eq!(api_error.code, S3ErrorCode::InternalError);
-        let source = api_error
-            .source
-            .as_deref()
-            .and_then(|source| source.downcast_ref::<StorageError>())
-            .expect("API error should retain the storage error source");
-        assert!(matches!(source, StorageError::FileCorrupt));
-    }
-
     #[test]
     fn test_api_error_from_iam_error() {
         let iam_error = rustfs_iam::error::Error::other("IAM test error");
         let api_error: ApiError = iam_error.into();
 
-        assert_eq!(api_error.code, S3ErrorCode::InternalError);
-        assert_eq!(api_error.message, ApiError::error_code_to_message(&S3ErrorCode::InternalError));
-        let source = api_error
-            .source
-            .as_deref()
-            .and_then(|source| source.downcast_ref::<StorageError>())
-            .expect("API error should retain the storage error source");
-        assert!(matches!(source, StorageError::Io(io_error) if io_error.to_string().contains("IAM test error")));
+        // IAM error is first converted to StorageError, then to ApiError
+        assert!(api_error.source.is_some());
+        assert!(api_error.message.contains("test error"));
     }
 
     #[test]

rustfs/src/storage/s3_api/multipart.rs

@@ -39,11 +39,6 @@ pub(crate) struct ListMultipartUploadsParams {
 pub(crate) fn build_list_parts_output(res: ListPartsInfo) -> ListPartsOutput {
     let owner = rustfs_owner();
     let initiator = rustfs_initiator();
-    let transformed_parts = rustfs_utils::http::contains_key_str(&res.user_defined, rustfs_utils::http::SUFFIX_COMPRESSION)
-        || res
-            .user_defined
-            .keys()
-            .any(|key| rustfs_utils::http::is_object_encryption_marker(key));
 
     ListPartsOutput {
         bucket: Some(res.bucket),
@@ -56,14 +51,7 @@ pub(crate) fn build_list_parts_output(res: ListPartsInfo) -> ListPartsOutput {
                     e_tag: p.etag.map(|etag| to_s3s_etag(&etag)),
                     last_modified: p.last_mod.map(Timestamp::from),
                     part_number: p.part_num.try_into().ok(),
-                    // Compressed parts store fewer bytes than the client sent; S3
-                    // semantics report the uploaded (logical) size, matching
-                    // GetObjectAttributes ObjectParts.
-                    size: if p.actual_size > 0 || (transformed_parts && p.actual_size == 0) {
-                        Some(p.actual_size)
-                    } else {
-                        p.size.try_into().ok()
-                    },
+                    size: p.size.try_into().ok(),
                     ..Default::default()
                 })
                 .collect(),
@@ -259,116 +247,6 @@ mod tests {
         assert_eq!(output.initiator, Some(rustfs_initiator()));
     }
 
-    #[test]
-    fn test_list_parts_output_reports_logical_size_for_compressed_parts() {
-        let input = ListPartsInfo {
-            bucket: "bucket-a".to_string(),
-            object: "obj-a".to_string(),
-            upload_id: "upload-a".to_string(),
-            parts: vec![PartInfo {
-                part_num: 1,
-                // Stored (compressed) bytes on disk vs. the logical size the client uploaded.
-                size: 1_024,
-                actual_size: 8_388_608,
-                ..Default::default()
-            }],
-            ..Default::default()
-        };
-
-        let output = build_list_parts_output(input);
-        let parts = output.parts.as_ref().expect("parts should be present");
-
-        assert_eq!(parts.len(), 1);
-        assert_eq!(
-            parts[0].size,
-            Some(8_388_608),
-            "compressed parts must report the uploaded logical size, not the stored size"
-        );
-    }
-
-    #[test]
-    fn test_list_parts_output_reports_zero_logical_size_for_compressed_parts() {
-        let mut user_defined = std::collections::HashMap::new();
-        rustfs_utils::http::insert_str(&mut user_defined, rustfs_utils::http::SUFFIX_COMPRESSION, "S2".to_string());
-        let input = ListPartsInfo {
-            user_defined,
-            parts: vec![PartInfo {
-                part_num: 1,
-                // Legacy SSE writes an 8-byte end record for an empty part.
-                size: 8,
-                actual_size: 0,
-                ..Default::default()
-            }],
-            ..Default::default()
-        };
-
-        let output = build_list_parts_output(input);
-        let parts = output.parts.as_ref().expect("parts should be present");
-
-        assert_eq!(parts[0].size, Some(0));
-    }
-
-    #[test]
-    fn test_list_parts_output_reports_zero_logical_size_for_encrypted_parts() {
-        let input = ListPartsInfo {
-            user_defined: std::collections::HashMap::from([(
-                rustfs_utils::http::AMZ_SERVER_SIDE_ENCRYPTION.to_string(),
-                "AES256".to_string(),
-            )]),
-            parts: vec![
-                PartInfo {
-                    part_num: 1,
-                    size: 8,
-                    actual_size: 0,
-                    ..Default::default()
-                },
-                PartInfo {
-                    part_num: 2,
-                    size: 8,
-                    actual_size: -1,
-                    ..Default::default()
-                },
-            ],
-            ..Default::default()
-        };
-
-        let output = build_list_parts_output(input);
-        let parts = output.parts.as_ref().expect("parts should be present");
-
-        assert_eq!(parts[0].size, Some(0));
-        assert_eq!(parts[1].size, Some(8));
-    }
-
-    #[test]
-    fn test_list_parts_output_falls_back_to_stored_size_when_actual_size_unknown() {
-        let input = ListPartsInfo {
-            parts: vec![
-                PartInfo {
-                    part_num: 1,
-                    size: 1_024,
-                    // Uncompressed parts leave actual_size unset.
-                    actual_size: 0,
-                    ..Default::default()
-                },
-                PartInfo {
-                    part_num: 2,
-                    size: 1_024,
-                    // Legacy/unknown sentinel must not leak a negative size to clients.
-                    actual_size: -1,
-                    ..Default::default()
-                },
-            ],
-            ..Default::default()
-        };
-
-        let output = build_list_parts_output(input);
-        let parts = output.parts.as_ref().expect("parts should be present");
-
-        assert_eq!(parts.len(), 2);
-        assert_eq!(parts[0].size, Some(1024));
-        assert_eq!(parts[1].size, Some(1024));
-    }
-
     #[test]
     fn test_list_parts_output_normalizes_legacy_storage_class_and_handles_overflow_markers() {
         let input = ListPartsInfo {

rustfs/src/storage/sse.rs

@@ -4565,9 +4565,11 @@ mod tests {
         })
         .await
         .expect_err("mismatched kms context should fail");
-        assert_eq!(err.code, S3ErrorCode::InternalError);
-        assert_eq!(err.message, ApiError::error_code_to_message(&S3ErrorCode::InternalError));
-        assert_eq!(super::kms_data_plane_error_class(&err), "context_mismatch");
+        assert!(
+            err.message.contains("context") || err.message.contains("Context"),
+            "unexpected error for mismatched kms context: {}",
+            err.message
+        );
 
         manager.stop().await.expect("kms service should stop cleanly");
         reset_sse_dek_provider();
@@ -5334,16 +5336,7 @@ mod tests {
 
         let error = TestSseDekProvider::decrypt_dek(&envelope, [0x55u8; 32])
             .expect_err("unknown JSON envelope versions must fail closed");
-        assert_eq!(error.code, S3ErrorCode::InternalError);
-        assert_eq!(error.message, ApiError::error_code_to_message(&S3ErrorCode::InternalError));
-        let source = error
-            .source
-            .as_deref()
-            .and_then(|source| source.downcast_ref::<StorageError>())
-            .expect("API error should retain the storage error source");
-        assert!(matches!(source, StorageError::Io(io_error) if io_error
-            .to_string()
-            .contains("Unsupported encrypted DEK format version")));
+        assert!(error.message.contains("Unsupported encrypted DEK format version"));
     }
 
     #[tokio::test]
@@ -5901,16 +5894,10 @@ mod tests {
     }
 
     #[test]
-    fn test_map_get_object_reader_error_redacts_non_ssec_internal_errors() {
+    fn test_map_get_object_reader_error_leaves_non_ssec_errors_unchanged() {
         let err = map_get_object_reader_error(StorageError::other("plain io failure"));
         assert_eq!(err.code, S3ErrorCode::InternalError);
-        assert_eq!(err.message, ApiError::error_code_to_message(&S3ErrorCode::InternalError));
-        let source = err
-            .source
-            .as_deref()
-            .and_then(|source| source.downcast_ref::<StorageError>())
-            .expect("API error should retain the storage error source");
-        assert!(matches!(source, StorageError::Io(io_error) if io_error.to_string().contains("plain io failure")));
+        assert_eq!(err.message, "Io error: plain io failure");
     }
 
     #[test]

rustfs/src/storage/storage_api.rs

@@ -409,9 +409,7 @@ pub(crate) mod ecstore_client {
 }
 
 pub(crate) mod ecstore_compression {
-    pub(crate) use rustfs_ecstore::api::compression::{
-        MIN_DISK_COMPRESSIBLE_SIZE, is_disk_compressible, is_multipart_disk_compression_enabled,
-    };
+    pub(crate) use rustfs_ecstore::api::compression::{MIN_DISK_COMPRESSIBLE_SIZE, is_disk_compressible};
 }
 
 pub(crate) mod ecstore_cluster {

rustfs/src/table_catalog/test_support.rs

@@ -26,7 +26,10 @@ use std::collections::{BTreeMap, BTreeSet};
 use std::sync::Arc;
 use time::OffsetDateTime;
 
-use super::*;
+use super::{
+    StrongTableCatalogRuntime, TableCatalogObject, TableCatalogObjectBackend, TableCatalogObjectMetadata,
+    TableCatalogPutPrecondition, TableCatalogStoreError, TableCatalogStoreResult, TableCommitPublication,
+};
 
 pub(crate) fn table_metadata_json(table_uuid: &str, location: &str) -> serde_json::Value {
     serde_json::json!({
@@ -936,637 +939,3 @@ impl TestCatalogObjectBackend {
         .expect("lock acquisition attempts should be observable");
     }
 }
-
-#[derive(Clone, Default)]
-pub(crate) struct TestCatalogPublishPause {
-    started: Arc<tokio::sync::Notify>,
-    release: Arc<tokio::sync::Notify>,
-}
-
-impl TestCatalogPublishPause {
-    pub(crate) async fn wait_started(&self) {
-        self.started.notified().await;
-    }
-
-    pub(crate) fn release(&self) {
-        self.release.notify_one();
-    }
-}
-
-// --- TableCatalogStore test doubles (backlog#1837 PR3) ---
-//
-// Two deliberately different shapes, per the issue's ruling: NoopTableCatalogStore
-// is a pure stub whose methods answer "nothing here", used where a store must
-// exist but never matter; TestTableCatalogStore is a stateful fake with commit
-// pauses and failure injection. Both live here so a TableCatalogStore trait
-// change is one file to update instead of two.
-
-pub(crate) struct NoopTableCatalogStore;
-
-#[async_trait::async_trait]
-impl TableCatalogStore for NoopTableCatalogStore {
-    async fn get_table_bucket(&self, _table_bucket: &str) -> TableCatalogStoreResult<Option<TableBucketEntry>> {
-        Ok(None)
-    }
-
-    async fn put_table_bucket(&self, _entry: TableBucketEntry) -> TableCatalogStoreResult<()> {
-        Ok(())
-    }
-
-    async fn create_namespace(&self, _entry: NamespaceEntry) -> TableCatalogStoreResult<()> {
-        Ok(())
-    }
-
-    async fn list_namespaces(&self, _table_bucket: &str) -> TableCatalogStoreResult<Vec<NamespaceEntry>> {
-        Ok(Vec::new())
-    }
-
-    async fn get_namespace(&self, _table_bucket: &str, _namespace: &str) -> TableCatalogStoreResult<Option<NamespaceEntry>> {
-        Ok(None)
-    }
-
-    async fn drop_namespace(&self, _table_bucket: &str, _namespace: &str) -> TableCatalogStoreResult<()> {
-        Ok(())
-    }
-
-    async fn create_table(&self, _entry: TableEntry) -> TableCatalogStoreResult<()> {
-        Ok(())
-    }
-
-    async fn register_table(&self, _entry: TableEntry) -> TableCatalogStoreResult<()> {
-        Ok(())
-    }
-
-    async fn register_table_with_publication(
-        &self,
-        entry: TableEntry,
-        publication: &(dyn TableCommitPublication + Sync),
-    ) -> TableCatalogStoreResult<()> {
-        publication.begin_table_bucket(&entry.table_bucket).await?;
-        if !publication.holds_table_bucket(&entry.table_bucket) {
-            return Err(TableCatalogStoreError::Internal(
-                "table registration requires a table-bucket publication fence".to_string(),
-            ));
-        }
-        let _publication_completion = TableCommitPublicationCompletion::new(publication);
-        publication
-            .prepare(&entry.table_bucket, &entry.namespace, &entry.table)
-            .await?;
-        if !publication.holds_table(&entry.table_bucket, &entry.namespace, &entry.table) {
-            return Err(TableCatalogStoreError::Internal(
-                "table registration requires a table publication fence".to_string(),
-            ));
-        }
-        self.register_table(entry).await
-    }
-
-    async fn list_tables(&self, _table_bucket: &str, _namespace: &str) -> TableCatalogStoreResult<Vec<TableEntry>> {
-        Ok(Vec::new())
-    }
-
-    async fn list_all_tables(&self, _table_bucket: &str) -> TableCatalogStoreResult<Vec<TableEntry>> {
-        Ok(Vec::new())
-    }
-
-    async fn load_table(
-        &self,
-        _table_bucket: &str,
-        _namespace: &str,
-        _table: &str,
-    ) -> TableCatalogStoreResult<Option<TableEntry>> {
-        Ok(None)
-    }
-
-    async fn commit_table(&self, request: TableCommitRequest) -> TableCatalogStoreResult<TableCommitResult> {
-        let table = TableEntry {
-            version: TABLE_CATALOG_ENTRY_VERSION,
-            table_bucket: request.table_bucket,
-            namespace: request.namespace,
-            table: request.table,
-            table_id: "table-id".to_string(),
-            table_uuid: "table-uuid".to_string(),
-            format: "ICEBERG".to_string(),
-            format_version: 2,
-            warehouse_location: "s3://analytics/tables/table-id".to_string(),
-            metadata_location: request.new_metadata_location.clone(),
-            version_token: "token-v2".to_string(),
-            generation: 2,
-            state: TableCatalogEntryState::Active,
-            properties: BTreeMap::new(),
-            created_at: None,
-            updated_at: None,
-        };
-        let commit_log = CommitLogEntry {
-            version: TABLE_CATALOG_ENTRY_VERSION,
-            commit_id: request.commit_id,
-            idempotency_key: request.idempotency_key,
-            table_id: table.table_id.clone(),
-            operation: request.operation,
-            expected_version_token: request.expected_version_token,
-            new_version_token: table.version_token.clone(),
-            previous_metadata_location: request.expected_metadata_location,
-            new_metadata_location: table.metadata_location.clone(),
-            requirements: request.requirements,
-            status: CommitLogStatus::Committed,
-            writer: request.writer,
-            created_at: None,
-            updated_at: None,
-        };
-
-        Ok(TableCommitResult { table, commit_log })
-    }
-
-    async fn commit_table_with_publication(
-        &self,
-        request: TableCommitRequest,
-        publication: &(dyn TableCommitPublication + Sync),
-    ) -> TableCatalogStoreResult<TableCommitResult> {
-        publication
-            .prepare(&request.table_bucket, &request.namespace, &request.table)
-            .await?;
-        if !publication.holds_table(&request.table_bucket, &request.namespace, &request.table) {
-            return Err(TableCatalogStoreError::Internal(
-                "table commit requires a table publication fence".to_string(),
-            ));
-        }
-        let _publication_completion = TableCommitPublicationCompletion::new(publication);
-        self.commit_table(request).await
-    }
-
-    async fn drop_table(&self, _table_bucket: &str, _namespace: &str, _table: &str) -> TableCatalogStoreResult<()> {
-        Ok(())
-    }
-
-    async fn create_view(&self, _entry: ViewEntry) -> TableCatalogStoreResult<()> {
-        Ok(())
-    }
-
-    async fn list_views(&self, _table_bucket: &str, _namespace: &str) -> TableCatalogStoreResult<Vec<ViewEntry>> {
-        Ok(Vec::new())
-    }
-
-    async fn load_view(&self, _table_bucket: &str, _namespace: &str, _view: &str) -> TableCatalogStoreResult<Option<ViewEntry>> {
-        Ok(None)
-    }
-
-    async fn replace_view(&self, request: ViewCommitRequest) -> TableCatalogStoreResult<ViewCommitResult> {
-        Ok(ViewCommitResult {
-            view: ViewEntry {
-                version: TABLE_CATALOG_ENTRY_VERSION,
-                table_bucket: request.table_bucket,
-                namespace: request.namespace,
-                view: request.view,
-                view_id: "view-id".to_string(),
-                view_uuid: "view-uuid".to_string(),
-                format: "ICEBERG_VIEW".to_string(),
-                format_version: 1,
-                warehouse_location: "s3://analytics/views/view-id".to_string(),
-                metadata_location: request.new_metadata_location,
-                version_token: "token-v2".to_string(),
-                generation: 2,
-                state: TableCatalogEntryState::Active,
-                properties: BTreeMap::new(),
-                created_at: None,
-                updated_at: None,
-            },
-        })
-    }
-
-    async fn drop_view(&self, _table_bucket: &str, _namespace: &str, _view: &str) -> TableCatalogStoreResult<()> {
-        Ok(())
-    }
-
-    async fn get_commit_by_id(
-        &self,
-        _table_bucket: &str,
-        _table_id: &str,
-        _commit_id: &str,
-    ) -> TableCatalogStoreResult<Option<CommitLogEntry>> {
-        Ok(None)
-    }
-
-    async fn get_commit_by_idempotency_key(
-        &self,
-        _table_bucket: &str,
-        _table_id: &str,
-        _idempotency_key: &str,
-    ) -> TableCatalogStoreResult<Option<CommitLogEntry>> {
-        Ok(None)
-    }
-}
-
-#[derive(Default)]
-pub(crate) struct TestTableCatalogStore {
-    pub(crate) table_buckets: tokio::sync::Mutex<Vec<crate::table_catalog::TableBucketEntry>>,
-    pub(crate) namespaces: tokio::sync::Mutex<Vec<crate::table_catalog::NamespaceEntry>>,
-    pub(crate) tables: tokio::sync::Mutex<Vec<crate::table_catalog::TableEntry>>,
-    pub(crate) views: tokio::sync::Mutex<Vec<crate::table_catalog::ViewEntry>>,
-    pub(crate) commits: tokio::sync::Mutex<Vec<crate::table_catalog::CommitLogEntry>>,
-    pub(crate) fail_put_table_bucket: tokio::sync::Mutex<bool>,
-    pub(crate) register_table_pause: Option<TestCatalogPublishPause>,
-    pub(crate) commit_table_pause: Option<TestCatalogPublishPause>,
-}
-
-#[async_trait::async_trait]
-impl crate::table_catalog::TableCatalogStore for TestTableCatalogStore {
-    async fn get_table_bucket(
-        &self,
-        table_bucket: &str,
-    ) -> crate::table_catalog::TableCatalogStoreResult<Option<crate::table_catalog::TableBucketEntry>> {
-        Ok(self
-            .table_buckets
-            .lock()
-            .await
-            .iter()
-            .find(|entry| entry.table_bucket == table_bucket)
-            .cloned())
-    }
-
-    async fn put_table_bucket(
-        &self,
-        entry: crate::table_catalog::TableBucketEntry,
-    ) -> crate::table_catalog::TableCatalogStoreResult<()> {
-        let mut fail_put_table_bucket = self.fail_put_table_bucket.lock().await;
-        if *fail_put_table_bucket {
-            *fail_put_table_bucket = false;
-            return Err(crate::table_catalog::TableCatalogStoreError::Internal(
-                "injected table bucket write failure".to_string(),
-            ));
-        }
-        drop(fail_put_table_bucket);
-
-        let mut table_buckets = self.table_buckets.lock().await;
-        table_buckets.retain(|existing| existing.table_bucket != entry.table_bucket);
-        table_buckets.push(entry);
-        Ok(())
-    }
-
-    async fn create_namespace(
-        &self,
-        entry: crate::table_catalog::NamespaceEntry,
-    ) -> crate::table_catalog::TableCatalogStoreResult<()> {
-        if self.get_table_bucket(&entry.table_bucket).await?.is_none() {
-            return Err(crate::table_catalog::TableCatalogStoreError::NotFound(format!(
-                "table bucket {}",
-                entry.table_bucket
-            )));
-        }
-        self.namespaces.lock().await.push(entry);
-        Ok(())
-    }
-
-    async fn list_namespaces(
-        &self,
-        table_bucket: &str,
-    ) -> crate::table_catalog::TableCatalogStoreResult<Vec<crate::table_catalog::NamespaceEntry>> {
-        Ok(self
-            .namespaces
-            .lock()
-            .await
-            .iter()
-            .filter(|entry| entry.table_bucket == table_bucket)
-            .cloned()
-            .collect())
-    }
-
-    async fn get_namespace(
-        &self,
-        table_bucket: &str,
-        namespace: &str,
-    ) -> crate::table_catalog::TableCatalogStoreResult<Option<crate::table_catalog::NamespaceEntry>> {
-        Ok(self
-            .namespaces
-            .lock()
-            .await
-            .iter()
-            .find(|entry| entry.table_bucket == table_bucket && entry.namespace == namespace)
-            .cloned())
-    }
-
-    async fn update_namespace_properties(
-        &self,
-        table_bucket: &str,
-        namespace: &str,
-        update: crate::table_catalog::NamespacePropertiesUpdate,
-    ) -> crate::table_catalog::TableCatalogStoreResult<crate::table_catalog::NamespacePropertiesUpdateResult> {
-        let mut namespaces = self.namespaces.lock().await;
-        let entry = namespaces
-            .iter_mut()
-            .find(|entry| entry.table_bucket == table_bucket && entry.namespace == namespace)
-            .ok_or_else(|| {
-                crate::table_catalog::TableCatalogStoreError::NotFound(format!("namespace {table_bucket}/{namespace}"))
-            })?;
-        Ok(update.apply_to(entry))
-    }
-
-    async fn drop_namespace(&self, table_bucket: &str, namespace: &str) -> crate::table_catalog::TableCatalogStoreResult<()> {
-        self.namespaces
-            .lock()
-            .await
-            .retain(|entry| !(entry.table_bucket == table_bucket && entry.namespace == namespace));
-        Ok(())
-    }
-
-    async fn create_table(&self, entry: crate::table_catalog::TableEntry) -> crate::table_catalog::TableCatalogStoreResult<()> {
-        if self.get_table_bucket(&entry.table_bucket).await?.is_none() {
-            return Err(crate::table_catalog::TableCatalogStoreError::NotFound(format!(
-                "table bucket {}",
-                entry.table_bucket
-            )));
-        }
-        if self.get_namespace(&entry.table_bucket, &entry.namespace).await?.is_none() {
-            return Err(crate::table_catalog::TableCatalogStoreError::NotFound(format!(
-                "namespace {}/{}",
-                entry.table_bucket, entry.namespace
-            )));
-        }
-        self.tables.lock().await.push(entry);
-        Ok(())
-    }
-
-    async fn register_table(&self, entry: crate::table_catalog::TableEntry) -> crate::table_catalog::TableCatalogStoreResult<()> {
-        if self.get_table_bucket(&entry.table_bucket).await?.is_none() {
-            return Err(crate::table_catalog::TableCatalogStoreError::NotFound(format!(
-                "table bucket {}",
-                entry.table_bucket
-            )));
-        }
-        if self.get_namespace(&entry.table_bucket, &entry.namespace).await?.is_none() {
-            return Err(crate::table_catalog::TableCatalogStoreError::NotFound(format!(
-                "namespace {}/{}",
-                entry.table_bucket, entry.namespace
-            )));
-        }
-        if let Some(pause) = &self.register_table_pause {
-            pause.started.notify_one();
-            pause.release.notified().await;
-        }
-        self.tables.lock().await.push(entry);
-        Ok(())
-    }
-
-    async fn register_table_with_publication(
-        &self,
-        entry: crate::table_catalog::TableEntry,
-        publication: &(dyn crate::table_catalog::TableCommitPublication + Sync),
-    ) -> crate::table_catalog::TableCatalogStoreResult<()> {
-        publication.begin_table_bucket(&entry.table_bucket).await?;
-        if !publication.holds_table_bucket(&entry.table_bucket) {
-            return Err(crate::table_catalog::TableCatalogStoreError::Internal(
-                "table registration requires a table-bucket publication fence".to_string(),
-            ));
-        }
-        let _publication_completion = crate::table_catalog::TableCommitPublicationCompletion::new(publication);
-        publication
-            .prepare(&entry.table_bucket, &entry.namespace, &entry.table)
-            .await?;
-        if !publication.holds_table(&entry.table_bucket, &entry.namespace, &entry.table) {
-            return Err(crate::table_catalog::TableCatalogStoreError::Internal(
-                "table registration requires a table publication fence".to_string(),
-            ));
-        }
-        self.register_table(entry).await
-    }
-
-    async fn list_tables(
-        &self,
-        table_bucket: &str,
-        namespace: &str,
-    ) -> crate::table_catalog::TableCatalogStoreResult<Vec<crate::table_catalog::TableEntry>> {
-        Ok(self
-            .tables
-            .lock()
-            .await
-            .iter()
-            .filter(|entry| entry.table_bucket == table_bucket && entry.namespace == namespace)
-            .cloned()
-            .collect())
-    }
-
-    async fn list_all_tables(
-        &self,
-        table_bucket: &str,
-    ) -> crate::table_catalog::TableCatalogStoreResult<Vec<crate::table_catalog::TableEntry>> {
-        Ok(self
-            .tables
-            .lock()
-            .await
-            .iter()
-            .filter(|entry| entry.table_bucket == table_bucket)
-            .cloned()
-            .collect())
-    }
-
-    async fn load_table(
-        &self,
-        table_bucket: &str,
-        namespace: &str,
-        table: &str,
-    ) -> crate::table_catalog::TableCatalogStoreResult<Option<crate::table_catalog::TableEntry>> {
-        Ok(self
-            .tables
-            .lock()
-            .await
-            .iter()
-            .find(|entry| entry.table_bucket == table_bucket && entry.namespace == namespace && entry.table == table)
-            .cloned())
-    }
-
-    async fn commit_table(
-        &self,
-        request: crate::table_catalog::TableCommitRequest,
-    ) -> crate::table_catalog::TableCatalogStoreResult<crate::table_catalog::TableCommitResult> {
-        let mut tables = self.tables.lock().await;
-        let Some(index) = tables.iter().position(|entry| {
-            entry.table_bucket == request.table_bucket && entry.namespace == request.namespace && entry.table == request.table
-        }) else {
-            return Err(crate::table_catalog::TableCatalogStoreError::NotFound(format!(
-                "table {}/{}/{}",
-                request.table_bucket, request.namespace, request.table
-            )));
-        };
-
-        let current = tables[index].clone();
-        if current.version_token != request.expected_version_token {
-            return Err(crate::table_catalog::TableCatalogStoreError::Conflict(
-                "current table version token does not match expected token".to_string(),
-            ));
-        }
-        if current.metadata_location != request.expected_metadata_location {
-            return Err(crate::table_catalog::TableCatalogStoreError::Conflict(
-                "current table metadata location does not match expected location".to_string(),
-            ));
-        }
-        if let Some(pause) = &self.commit_table_pause {
-            pause.started.notify_one();
-            pause.release.notified().await;
-        }
-
-        let mut next = current.clone();
-        next.metadata_location = request.new_metadata_location.clone();
-        next.version_token = "token-committed".to_string();
-        next.generation = next.generation.saturating_add(1);
-        tables[index] = next.clone();
-        drop(tables);
-
-        let commit_log = crate::table_catalog::CommitLogEntry {
-            version: crate::table_catalog::TABLE_CATALOG_ENTRY_VERSION,
-            commit_id: request.commit_id,
-            idempotency_key: request.idempotency_key,
-            table_id: current.table_id,
-            operation: request.operation,
-            expected_version_token: request.expected_version_token,
-            new_version_token: next.version_token.clone(),
-            previous_metadata_location: request.expected_metadata_location,
-            new_metadata_location: request.new_metadata_location,
-            requirements: request.requirements,
-            status: crate::table_catalog::CommitLogStatus::Committed,
-            writer: request.writer,
-            created_at: None,
-            updated_at: None,
-        };
-        self.commits.lock().await.push(commit_log.clone());
-
-        Ok(crate::table_catalog::TableCommitResult { table: next, commit_log })
-    }
-
-    async fn commit_table_with_publication(
-        &self,
-        request: crate::table_catalog::TableCommitRequest,
-        publication: &(dyn crate::table_catalog::TableCommitPublication + Sync),
-    ) -> crate::table_catalog::TableCatalogStoreResult<crate::table_catalog::TableCommitResult> {
-        publication
-            .prepare(&request.table_bucket, &request.namespace, &request.table)
-            .await?;
-        if !publication.holds_table(&request.table_bucket, &request.namespace, &request.table) {
-            return Err(crate::table_catalog::TableCatalogStoreError::Internal(
-                "table commit requires a table publication fence".to_string(),
-            ));
-        }
-        let _publication_completion = crate::table_catalog::TableCommitPublicationCompletion::new(publication);
-        self.commit_table(request).await
-    }
-
-    async fn drop_table(
-        &self,
-        table_bucket: &str,
-        namespace: &str,
-        table: &str,
-    ) -> crate::table_catalog::TableCatalogStoreResult<()> {
-        self.tables
-            .lock()
-            .await
-            .retain(|entry| !(entry.table_bucket == table_bucket && entry.namespace == namespace && entry.table == table));
-        Ok(())
-    }
-
-    async fn create_view(&self, entry: crate::table_catalog::ViewEntry) -> crate::table_catalog::TableCatalogStoreResult<()> {
-        if self.get_table_bucket(&entry.table_bucket).await?.is_none() {
-            return Err(crate::table_catalog::TableCatalogStoreError::NotFound(format!(
-                "table bucket {}",
-                entry.table_bucket
-            )));
-        }
-        if self.get_namespace(&entry.table_bucket, &entry.namespace).await?.is_none() {
-            return Err(crate::table_catalog::TableCatalogStoreError::NotFound(format!(
-                "namespace {}/{}",
-                entry.table_bucket, entry.namespace
-            )));
-        }
-        self.views.lock().await.push(entry);
-        Ok(())
-    }
-
-    async fn list_views(
-        &self,
-        table_bucket: &str,
-        namespace: &str,
-    ) -> crate::table_catalog::TableCatalogStoreResult<Vec<crate::table_catalog::ViewEntry>> {
-        Ok(self
-            .views
-            .lock()
-            .await
-            .iter()
-            .filter(|entry| entry.table_bucket == table_bucket && entry.namespace == namespace)
-            .cloned()
-            .collect())
-    }
-
-    async fn load_view(
-        &self,
-        table_bucket: &str,
-        namespace: &str,
-        view: &str,
-    ) -> crate::table_catalog::TableCatalogStoreResult<Option<crate::table_catalog::ViewEntry>> {
-        Ok(self
-            .views
-            .lock()
-            .await
-            .iter()
-            .find(|entry| entry.table_bucket == table_bucket && entry.namespace == namespace && entry.view == view)
-            .cloned())
-    }
-
-    async fn replace_view(
-        &self,
-        request: crate::table_catalog::ViewCommitRequest,
-    ) -> crate::table_catalog::TableCatalogStoreResult<crate::table_catalog::ViewCommitResult> {
-        let mut views = self.views.lock().await;
-        let Some(index) = views.iter().position(|entry| {
-            entry.table_bucket == request.table_bucket && entry.namespace == request.namespace && entry.view == request.view
-        }) else {
-            return Err(crate::table_catalog::TableCatalogStoreError::NotFound(format!(
-                "view {}/{}/{}",
-                request.table_bucket, request.namespace, request.view
-            )));
-        };
-        let current = views[index].clone();
-        if current.version_token != request.expected_version_token {
-            return Err(crate::table_catalog::TableCatalogStoreError::Conflict(
-                "current view version token does not match expected token".to_string(),
-            ));
-        }
-        if current.metadata_location != request.expected_metadata_location {
-            return Err(crate::table_catalog::TableCatalogStoreError::Conflict(
-                "current view metadata location does not match expected location".to_string(),
-            ));
-        }
-        let mut next = current;
-        next.metadata_location = request.new_metadata_location;
-        next.version_token = "token-view-committed".to_string();
-        next.generation = next.generation.saturating_add(1);
-        views[index] = next.clone();
-        Ok(crate::table_catalog::ViewCommitResult { view: next })
-    }
-
-    async fn drop_view(
-        &self,
-        table_bucket: &str,
-        namespace: &str,
-        view: &str,
-    ) -> crate::table_catalog::TableCatalogStoreResult<()> {
-        self.views
-            .lock()
-            .await
-            .retain(|entry| !(entry.table_bucket == table_bucket && entry.namespace == namespace && entry.view == view));
-        Ok(())
-    }
-
-    async fn get_commit_by_id(
-        &self,
-        _table_bucket: &str,
-        _table_id: &str,
-        _commit_id: &str,
-    ) -> crate::table_catalog::TableCatalogStoreResult<Option<crate::table_catalog::CommitLogEntry>> {
-        Ok(None)
-    }
-
-    async fn get_commit_by_idempotency_key(
-        &self,
-        _table_bucket: &str,
-        _table_id: &str,
-        _idempotency_key: &str,
-    ) -> crate::table_catalog::TableCatalogStoreResult<Option<crate::table_catalog::CommitLogEntry>> {
-        Ok(None)
-    }
-}

rustfs/src/table_catalog/tests.rs

@@ -3,9 +3,7 @@ use super::identifier::{
     default_table_lifecycle_path, default_table_marker_path, default_table_root_prefix, is_valid_table_metadata_file_name,
     namespace_name_from_marker_path, table_name_from_marker_path, validate_object_mutation,
 };
-use super::test_support::{
-    BlockingObjectPublication, NoopTableCatalogStore, TestCatalogObjectBackend, UnserializedTestPublication,
-};
+use super::test_support::{BlockingObjectPublication, TestCatalogObjectBackend, UnserializedTestPublication};
 use super::*;
 use datafusion::{
     arrow::{
@@ -224,6 +222,200 @@ fn catalog_object_listing_rejects_missing_or_stalled_continuation_tokens() {
     );
 }
 
+struct NoopTableCatalogStore;
+
+#[async_trait::async_trait]
+impl TableCatalogStore for NoopTableCatalogStore {
+    async fn get_table_bucket(&self, _table_bucket: &str) -> TableCatalogStoreResult<Option<TableBucketEntry>> {
+        Ok(None)
+    }
+
+    async fn put_table_bucket(&self, _entry: TableBucketEntry) -> TableCatalogStoreResult<()> {
+        Ok(())
+    }
+
+    async fn create_namespace(&self, _entry: NamespaceEntry) -> TableCatalogStoreResult<()> {
+        Ok(())
+    }
+
+    async fn list_namespaces(&self, _table_bucket: &str) -> TableCatalogStoreResult<Vec<NamespaceEntry>> {
+        Ok(Vec::new())
+    }
+
+    async fn get_namespace(&self, _table_bucket: &str, _namespace: &str) -> TableCatalogStoreResult<Option<NamespaceEntry>> {
+        Ok(None)
+    }
+
+    async fn drop_namespace(&self, _table_bucket: &str, _namespace: &str) -> TableCatalogStoreResult<()> {
+        Ok(())
+    }
+
+    async fn create_table(&self, _entry: TableEntry) -> TableCatalogStoreResult<()> {
+        Ok(())
+    }
+
+    async fn register_table(&self, _entry: TableEntry) -> TableCatalogStoreResult<()> {
+        Ok(())
+    }
+
+    async fn register_table_with_publication(
+        &self,
+        entry: TableEntry,
+        publication: &(dyn TableCommitPublication + Sync),
+    ) -> TableCatalogStoreResult<()> {
+        publication.begin_table_bucket(&entry.table_bucket).await?;
+        if !publication.holds_table_bucket(&entry.table_bucket) {
+            return Err(TableCatalogStoreError::Internal(
+                "table registration requires a table-bucket publication fence".to_string(),
+            ));
+        }
+        let _publication_completion = TableCommitPublicationCompletion::new(publication);
+        publication
+            .prepare(&entry.table_bucket, &entry.namespace, &entry.table)
+            .await?;
+        if !publication.holds_table(&entry.table_bucket, &entry.namespace, &entry.table) {
+            return Err(TableCatalogStoreError::Internal(
+                "table registration requires a table publication fence".to_string(),
+            ));
+        }
+        self.register_table(entry).await
+    }
+
+    async fn list_tables(&self, _table_bucket: &str, _namespace: &str) -> TableCatalogStoreResult<Vec<TableEntry>> {
+        Ok(Vec::new())
+    }
+
+    async fn list_all_tables(&self, _table_bucket: &str) -> TableCatalogStoreResult<Vec<TableEntry>> {
+        Ok(Vec::new())
+    }
+
+    async fn load_table(
+        &self,
+        _table_bucket: &str,
+        _namespace: &str,
+        _table: &str,
+    ) -> TableCatalogStoreResult<Option<TableEntry>> {
+        Ok(None)
+    }
+
+    async fn commit_table(&self, request: TableCommitRequest) -> TableCatalogStoreResult<TableCommitResult> {
+        let table = TableEntry {
+            version: TABLE_CATALOG_ENTRY_VERSION,
+            table_bucket: request.table_bucket,
+            namespace: request.namespace,
+            table: request.table,
+            table_id: "table-id".to_string(),
+            table_uuid: "table-uuid".to_string(),
+            format: "ICEBERG".to_string(),
+            format_version: 2,
+            warehouse_location: "s3://analytics/tables/table-id".to_string(),
+            metadata_location: request.new_metadata_location.clone(),
+            version_token: "token-v2".to_string(),
+            generation: 2,
+            state: TableCatalogEntryState::Active,
+            properties: BTreeMap::new(),
+            created_at: None,
+            updated_at: None,
+        };
+        let commit_log = CommitLogEntry {
+            version: TABLE_CATALOG_ENTRY_VERSION,
+            commit_id: request.commit_id,
+            idempotency_key: request.idempotency_key,
+            table_id: table.table_id.clone(),
+            operation: request.operation,
+            expected_version_token: request.expected_version_token,
+            new_version_token: table.version_token.clone(),
+            previous_metadata_location: request.expected_metadata_location,
+            new_metadata_location: table.metadata_location.clone(),
+            requirements: request.requirements,
+            status: CommitLogStatus::Committed,
+            writer: request.writer,
+            created_at: None,
+            updated_at: None,
+        };
+
+        Ok(TableCommitResult { table, commit_log })
+    }
+
+    async fn commit_table_with_publication(
+        &self,
+        request: TableCommitRequest,
+        publication: &(dyn TableCommitPublication + Sync),
+    ) -> TableCatalogStoreResult<TableCommitResult> {
+        publication
+            .prepare(&request.table_bucket, &request.namespace, &request.table)
+            .await?;
+        if !publication.holds_table(&request.table_bucket, &request.namespace, &request.table) {
+            return Err(TableCatalogStoreError::Internal(
+                "table commit requires a table publication fence".to_string(),
+            ));
+        }
+        let _publication_completion = TableCommitPublicationCompletion::new(publication);
+        self.commit_table(request).await
+    }
+
+    async fn drop_table(&self, _table_bucket: &str, _namespace: &str, _table: &str) -> TableCatalogStoreResult<()> {
+        Ok(())
+    }
+
+    async fn create_view(&self, _entry: ViewEntry) -> TableCatalogStoreResult<()> {
+        Ok(())
+    }
+
+    async fn list_views(&self, _table_bucket: &str, _namespace: &str) -> TableCatalogStoreResult<Vec<ViewEntry>> {
+        Ok(Vec::new())
+    }
+
+    async fn load_view(&self, _table_bucket: &str, _namespace: &str, _view: &str) -> TableCatalogStoreResult<Option<ViewEntry>> {
+        Ok(None)
+    }
+
+    async fn replace_view(&self, request: ViewCommitRequest) -> TableCatalogStoreResult<ViewCommitResult> {
+        Ok(ViewCommitResult {
+            view: ViewEntry {
+                version: TABLE_CATALOG_ENTRY_VERSION,
+                table_bucket: request.table_bucket,
+                namespace: request.namespace,
+                view: request.view,
+                view_id: "view-id".to_string(),
+                view_uuid: "view-uuid".to_string(),
+                format: "ICEBERG_VIEW".to_string(),
+                format_version: 1,
+                warehouse_location: "s3://analytics/views/view-id".to_string(),
+                metadata_location: request.new_metadata_location,
+                version_token: "token-v2".to_string(),
+                generation: 2,
+                state: TableCatalogEntryState::Active,
+                properties: BTreeMap::new(),
+                created_at: None,
+                updated_at: None,
+            },
+        })
+    }
+
+    async fn drop_view(&self, _table_bucket: &str, _namespace: &str, _view: &str) -> TableCatalogStoreResult<()> {
+        Ok(())
+    }
+
+    async fn get_commit_by_id(
+        &self,
+        _table_bucket: &str,
+        _table_id: &str,
+        _commit_id: &str,
+    ) -> TableCatalogStoreResult<Option<CommitLogEntry>> {
+        Ok(None)
+    }
+
+    async fn get_commit_by_idempotency_key(
+        &self,
+        _table_bucket: &str,
+        _table_id: &str,
+        _idempotency_key: &str,
+    ) -> TableCatalogStoreResult<Option<CommitLogEntry>> {
+        Ok(None)
+    }
+}
+
 #[tokio::test]
 async fn table_catalog_store_trait_covers_entry_read_write_shapes() {
     let store: &dyn TableCatalogStore = &NoopTableCatalogStore;

scripts/check_architecture_migration_rules.sh

@@ -4033,7 +4033,7 @@ if [[ -s "$ECSTORE_REMOTE_TIER_DELETE_STATE_BYPASS_HITS_FILE" ]]; then
   report_failure "remote tier delete state access must stay behind ECStore tier sweeper owner helpers: $(paste -sd '; ' "$ECSTORE_REMOTE_TIER_DELETE_STATE_BYPASS_HITS_FILE")"
 fi
 
-RUSTFS_OWNER_LOCAL_STATIC_NAMES='(KEYSTONE_AUTH|KEYSTONE_MAPPER|KEYSTONE_CONFIG|LICENSE_STATE|LICENSE_VERIFIER|CPU_CONT_GUARD|PROFILING_CANCEL_TOKEN|MEMORY_SYSTEM|DIAL9_TELEMETRY_GUARD|DISPLAY_CONFIG_SNAPSHOT|GLOBAL_CONFIG_SNAPSHOT|BUFFER_CONFIG_SINGLETON|BUFFER_PROFILE_ENABLED|LEGACY_CREDENTIAL_WARNED_KEYS|CONSOLE_CONFIG|ACTIVE_HTTP_REQUESTS|USE_STARSHARD_CACHE|BUCKET_CACHE_SMALL|BUCKET_CACHE_LARGE|GLOBAL_SSE_DEK_PROVIDER|SSE_TEST_LOCK|AUTH_FS|LOCK_STATS|DEADLOCK_DETECTOR|GET_OBJECT_BUFFER_THRESHOLD_WARNED|GET_READER_STREAM_BUFFER_SIZE_OVERRIDE|OBJECT_SEEK_SUPPORT_THRESHOLD|OBJECT_SEEK_SUPPORT_CONCURRENCY_THRESHOLDS|SUPPORTED_HEADERS|SITE_REPLICATION_PEER_CLIENT|AUDIT_MODULE_ENABLED|NOTIFY_MODULE_ENABLED|PERSISTED_NOTIFY_MODULE_ENABLED|PERSISTED_AUDIT_MODULE_ENABLED|PERSISTED_MODULE_SWITCH_CONFIGURED|DELETE_TAIL_TOTAL|DELETE_CLEANUP_TOTAL|DELETE_REPLICATION_TOTAL|DELETE_NOTIFY_TOTAL|EMBEDDED_SERVER_STARTED|TEST_OUTBOUND_TLS_GENERATION|TEST_REMAINING_FAILURES|CAPACITY_DIRTY_SCOPE_ENV|CAPACITY_DIRTY_SCOPE_INIT|GLOBAL_ENV)'
+RUSTFS_OWNER_LOCAL_STATIC_NAMES='(KEYSTONE_AUTH|KEYSTONE_MAPPER|KEYSTONE_CONFIG|LICENSE_STATE|LICENSE_VERIFIER|CPU_CONT_GUARD|PROFILING_CANCEL_TOKEN|MEMORY_SYSTEM|DIAL9_TELEMETRY_GUARD|DISPLAY_CONFIG_SNAPSHOT|GLOBAL_CONFIG_SNAPSHOT|BUFFER_CONFIG_SINGLETON|BUFFER_PROFILE_ENABLED|LEGACY_CREDENTIAL_WARNED_KEYS|CONSOLE_CONFIG|ACTIVE_HTTP_REQUESTS|USE_STARSHARD_CACHE|BUCKET_CACHE_SMALL|BUCKET_CACHE_LARGE|GLOBAL_SSE_DEK_PROVIDER|SSE_TEST_LOCK|AUTH_FS|LOCK_STATS|DEADLOCK_DETECTOR|GET_OBJECT_BUFFER_THRESHOLD_WARNED|GET_READER_STREAM_BUFFER_SIZE_OVERRIDE|OBJECT_SEEK_SUPPORT_THRESHOLD|OBJECT_SEEK_SUPPORT_CONCURRENCY_THRESHOLDS|SUPPORTED_HEADERS|SITE_REPLICATION_PEER_CLIENT|SITE_REPLICATION_STATE_LOCK|AUDIT_MODULE_ENABLED|NOTIFY_MODULE_ENABLED|PERSISTED_NOTIFY_MODULE_ENABLED|PERSISTED_AUDIT_MODULE_ENABLED|PERSISTED_MODULE_SWITCH_CONFIGURED|DELETE_TAIL_TOTAL|DELETE_CLEANUP_TOTAL|DELETE_REPLICATION_TOTAL|DELETE_NOTIFY_TOTAL|EMBEDDED_SERVER_STARTED|TEST_OUTBOUND_TLS_GENERATION|TEST_REMAINING_FAILURES|CAPACITY_DIRTY_SCOPE_ENV|CAPACITY_DIRTY_SCOPE_INIT|GLOBAL_ENV)'
 
 (
   cd "$ROOT_DIR"

scripts/check_logging_guardrails.sh

@@ -984,9 +984,7 @@ trace_hot_spans=(
   "crates/ecstore/src/store/object.rs:handle_get_object_info"
   "crates/ecstore/src/set_disk/ops/object.rs:get_object_info"
   "crates/ecstore/src/store/mod.rs:list_objects_v2"
-  # The ECStore handle_list_objects_v2 forwarder was folded into the trait impl
-  # above, so store/mod.rs now carries this hot path's TRACE requirement
-  # directly (backlog#1821).
+  "crates/ecstore/src/store/list.rs:handle_list_objects_v2"
   "crates/ecstore/src/core/sets.rs:list_objects_v2"
   "crates/ecstore/src/set_disk/ops/list.rs:list_objects_v2"
   "rustfs/src/app/bucket_usecase.rs:execute_list_objects_v2"

scripts/run.sh

@@ -209,7 +209,6 @@ export RUSTFS_NS_SCANNER_INTERVAL=60  # Object scanning interval in seconds
 
 # Storage level compression (compression at object storage level)
 # export RUSTFS_COMPRESSION_ENABLED=true # Whether to enable storage-level compression for objects
-# export RUSTFS_COMPRESSION_MULTIPART_ENABLED=true # Additionally compress multipart uploads (staged rollout switch: enable only after the whole fleet runs a build with the resumable decompressor; see docs/architecture/compat-cleanup-register.md)
 
 # HTTP Response Compression (whitelist-based, aligned with MinIO)
 # By default, HTTP response compression is DISABLED (aligned with MinIO behavior)