test(ilm): make active cancellation deterministic (#5403)

* ci: preserve timeout evidence for workspace tests

* test(ilm): make active cancellation deterministic

* ci: allow cold doctest compilation to finish

---------

Co-authored-by: houseme <housemecn@gmail.com>

.github/workflows/ci.yml

@@ -372,7 +372,7 @@ jobs:
           github-token: ${{ secrets.GITHUB_TOKEN }}
 
       - name: Build debug binary
-        run: cargo build -p rustfs --bins
+        run: cargo build -p rustfs --bins --features e2e-test-hooks
 
       - name: Upload debug binary
         uses: actions/upload-artifact@b7c566a772e6b6bfb58ed0dc250532a479d7789f # v6
@@ -402,7 +402,7 @@ jobs:
           github-token: ${{ secrets.GITHUB_TOKEN }}
 
       - name: Build debug binary with rio-v2
-        run: cargo build -p rustfs --bins --features rio-v2
+        run: cargo build -p rustfs --bins --features rio-v2,e2e-test-hooks
 
       - name: Upload debug binary
         uses: actions/upload-artifact@b7c566a772e6b6bfb58ed0dc250532a479d7789f # v6

crates/e2e_test/src/reliant/tiering.rs

@@ -95,6 +95,7 @@ const MANUAL_ASYNC_PARALLEL_OBJECTS: usize = 64;
 const MANUAL_ACTIVE_CANCEL_OBJECTS: usize = 512;
 const MANUAL_RESTART_CANCEL_OBJECTS: usize = 512;
 const MANUAL_ACTIVE_CANCEL_RUNNING_TIMEOUT: StdDuration = StdDuration::from_secs(15);
+const MANUAL_TRANSITION_CANCEL_BARRIER_ENV: &str = "RUSTFS_E2E_MANUAL_TRANSITION_CANCEL_BARRIER";
 const MANUAL_ASYNC_CONFLICT_TERMINAL_TIMEOUT: StdDuration = StdDuration::from_secs(90);
 const MANUAL_RESTART_RECOVERY_TIMEOUT: StdDuration = StdDuration::from_secs(80);
 const OBJECT_KEY: &str = "tier/鲁A12345/report.bin";
@@ -1684,8 +1685,15 @@ async fn test_manual_transition_async_active_cancel_reports_terminal_cancelled()
     cold_client.create_bucket().bucket(TIER_BUCKET).send().await?;
 
     let mut hot = RustFSTestEnvironment::new().await?;
-    hot.start_rustfs_server_with_env(vec![], &[("RUSTFS_SCANNER_ENABLED", "false"), ("RUSTFS_SCANNER_CYCLE", "3600")])
-        .await?;
+    hot.start_rustfs_server_with_env(
+        vec![],
+        &[
+            ("RUSTFS_SCANNER_ENABLED", "false"),
+            ("RUSTFS_SCANNER_CYCLE", "3600"),
+            (MANUAL_TRANSITION_CANCEL_BARRIER_ENV, "1"),
+        ],
+    )
+    .await?;
     let hot_client = hot.create_s3_client();
     add_rustfs_tier(&hot, &cold).await?;
 

rustfs/Cargo.toml

@@ -47,6 +47,7 @@ io-scheduler-debug = []  # Enable debug information in I/O scheduler
 tracing-chunk-debug = []  # Enable per-chunk tracing in data plane (high noise, for debugging only)
 full = ["metrics-gpu", "ftps", "swift", "webdav", "sftp", "pyroscope"]
 manual-test-runners = []
+e2e-test-hooks = []
 rio-v2 = ["rustfs-ecstore/rio-v2"]
 pyroscope = ["rustfs-obs/pyroscope"]
 # Tokio runtime telemetry. Requires `--cfg tokio_unstable`; use `make build-profiling`.

rustfs/src/admin/handlers/ilm_transition.rs

@@ -58,6 +58,8 @@ const LOG_SUBSYSTEM_ILM_TRANSITION: &str = "ilm_transition";
 const EVENT_ADMIN_ILM_TRANSITION_STATE: &str = "admin_ilm_transition_state";
 
 static ACTIVE_MANUAL_TRANSITION_SCOPES: OnceLock<Mutex<Vec<ManualTransitionRunScope>>> = OnceLock::new();
+#[cfg(feature = "e2e-test-hooks")]
+const E2E_MANUAL_TRANSITION_CANCEL_BARRIER_ENV: &str = "RUSTFS_E2E_MANUAL_TRANSITION_CANCEL_BARRIER";
 static ACTIVE_MANUAL_TRANSITION_JOBS: OnceLock<Mutex<HashMap<Uuid, CancellationToken>>> = OnceLock::new();
 static MANUAL_TRANSITION_OWNER_ID: OnceLock<String> = OnceLock::new();
 
@@ -748,6 +750,10 @@ async fn start_manual_transition_job(
     let job_heartbeat_shutdown_token = heartbeat_shutdown_token.clone();
     spawn_manual_transition_job_heartbeat(store, job_id, scan_cancel_token, heartbeat_shutdown_token);
     tokio::spawn(async move {
+        #[cfg(feature = "e2e-test-hooks")]
+        if std::env::var_os(E2E_MANUAL_TRANSITION_CANCEL_BARRIER_ENV).is_some() {
+            job_scan_cancel_token.cancelled().await;
+        }
         let result = enqueue_transition_for_existing_objects_scoped(run_store.clone(), &bucket, run_options).await;
         if let Some(final_record) = finalize_manual_transition_job(run_store.clone(), job_id, result).await
             && final_record.is_terminal()