Rangsh commented on PR #12173: URL: https://github.com/apache/seatunnel/pull/12173#issuecomment-5595572198
Thanks @SEZ9 for the careful re-review on `991e67b`. Addressed in `1de359ef9` on this PR branch: ### Issue 1 (Critical) — durability coverage Agreed: `FileMapStore.store` is `void` and does not fail the measured put when the underlying WAL `boolean` is false, so “write-through put returned ⇒ durable” was the wrong premise for dropping every iteration reload. Updated fixture: - every iteration still checks **resident** growth (now the **whole batch**, not only the last key); - **sampled** MapStore reload on the first iteration and every 4th iteration (interval chosen because `loadAll` always replays the full WAL); - trial tear-down still reloads the **full last growth batch**; - cleanup / fixture-close failures are attached with `addSuppressed` and no longer replace a durability failure. ### Issue 2 — merge atomicity `storeFinishedPipelineMetrics` now takes the IMap key lock around get → merge → single TTL `put`, so concurrent callers for the same `jobId` are serialized without referencing `JobMaster#metricsLock`. ### Issue 3 — trial sample scope Trial (and sampled) durability now reloads / verifies **every key in the growth batch**. Same WAL cost as a single-key `loadAll`, better coverage. ### Issue 4 — docs `docs/en|zh/engines/zeta/benchmark.md` now describe unique keys, resident checks, and light durable sampling vs per-sample full WAL replay. ### Issue 5 — Javadoc Rewrote the method contract (current IMap value, TTL put, `@param`); removed PR-history / `computeIfAbsent` wording and the private-field reference. ### Issue 6 — tests Dropped `never().computeIfAbsent(...)`. Added null-metrics rejection and distinct-tag merge coverage. `JobHistoryServiceFinishedMetricsTest`: 4 tests green locally. ### Evidence still in flight Fork Diagnostics for `completedJobHistoryGrowth` @ `initialStoredJobCount=1000` on this head: https://github.com/Rangsh/seatunnel/actions/runs/34309225279 Will follow with the matching `dev` before run (one-at-a-time on the fork runner), then post the before/after frames on #12063. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
