SEZ9 commented on PR #12290: URL: https://github.com/apache/seatunnel/pull/12290#issuecomment-5788012366
Thanks @SEPURI-SAI-KRISHNA, that clears it up. Agreed that there is nothing further to add on the `SubPlan.java:219` / CANCELING point since it was already raised in the changes-requested review, and that review attention rather than more analysis is what moves #12311, #12313 and #12377. On the backpressure side, the split you laid out is the one I'll go with: #12316 is the fix to review (the `awaitInjectors()` handoff in `SourceFlowLifeCycle` applies to every source, not just `FakeSource`), and #12313 is a test-only resize from `row.num = 10000000` / `split.num = 1` to `row.num = 2000000` / `split.num = 500`. Your correction that `BackpressureSlowSinkIT` fails at the first-checkpoint gate rather than the sustained window also matches the starvation reading. Good catch on the "once per second" comment in #12313 too — with 4000 rows per split at `write_delay_ms = 2` it is once per 8 seconds, and since that margin is the justification for the numbers, it would be worth leaving that as a comment on #12313 so the javadoc gets fixed before it lands. For this PR: the head is still `678c754a7bd6`, so nothing on the diff changes and it stays parked. Concrete remaining asks: 1. Rebase onto dev once #12311 and #12316 have merged. 2. Rerun `engine-v2-it` and post the link to one clean run here. 3. If either engine fix stalls, ping me on this thread and we can decide whether to merge on a targeted rerun instead of waiting. I'll pick up the review of #12316 from here. <!-- streview-comment:1248 --> -- 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]
