nzw921rx opened a new pull request, #13201: URL: https://github.com/apache/seatunnel/pull/13201
### Purpose of this pull request Multi-table timer flush currently flushes writer queues one at a time. In a 45-second recording of a JDBC sink with five writer queues and a 200 ms flush interval, the dispatcher spent 39.58 seconds waiting for INSERT responses inside `aggregatedFlush`. Draining the queues accounted for 0.30 seconds. This change reuses the existing executor to flush independent queues concurrently after draining pending writes. Child writers sharing a queue still flush sequentially under that queue's writer lock. The dispatcher waits for all flush tasks before processing more rows. Failed-table isolation runs after releasing the current writer lock because it also acquires other queue locks. Related to #10801 and #11537. ### Does this PR introduce _any_ user-facing change? Yes. Multi-table timer flush can now use multiple writer queues concurrently. `FAIL_FAST` propagates a flush error after the submitted tasks finish. `CONTINUE_OTHER_TABLES` isolates failed tables and continues flushing healthy tables. The English and Chinese timer flush documentation describes the new behavior. ### How was this patch tested? - Two regression cases cover concurrent flush execution, error propagation, and failed-table isolation. Both timed out with serial flushes and passed with this change. - All 44 MultiTable tests passed on JDK 11 against `dev` at `e1fe5c5e`. - API module `verify`, Spotless, and `git diff --check` passed. - Full offline `verify` stopped because Flink 1.15.3 artifacts were missing from the Maven cache. Post-change OceanBase throughput is pending validation. -- 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]
