SEZ9 commented on PR #12163: URL: https://github.com/apache/seatunnel/pull/12163#issuecomment-5691904377
Following up on the rerun `34791705815` — the summary above matches what I see, so here's where I land on each failure and what I'd like before we merge: - `OpengaussCDCIT.testAddFieldWithRestore` and `SplitClusterFaultToleranceIT.testStreamJobCancelResolvesWhenWorkerCrashesBeforeCancelAck` (`all-connectors-it-2`, `engine-v2-it (8)`): identical signatures to the earlier diagnosis and no overlap with `connector-cdc-base` or the Zeta enumerator-context change. I'm comfortable treating these as unrelated. - `S3FileConnectDryRunIT` on `all-connectors-it-7` (`pull access denied for minio/minio`): this is the Docker Hub image issue already fixed in `dev` by #12302. Your branch is 5 commits behind `dev`, so this will keep failing until you sync. - `BackpressureSlowSinkIT.testCheckpointsKeepCompletingUnderSustainedBackpressure` on `engine-v2-it (8, ubuntu-latest)`: I agree the call chain doesn't touch the CDC snapshot-split ack path or `SeaTunnelSplitEnumeratorContext`, and the runner-contention explanation is plausible, but it's still unconfirmed. I don't want to wave it through on reasoning alone. Concrete asks: 1. Rebase/merge the latest `dev` so #12302 is picked up, then push and let CI run again. 2. After that run, if `BackpressureSlowSinkIT` passes we're good; if it fails again with the same "expected at least 3 additional checkpoints ... only observed 0" message, please link a `dev` or sibling-PR run showing the same failure, or do a quick local repro so we can rule out a regression from the `sendEventToSourceReader` change on Zeta. 3. On the change itself: the description covers the re-report-until-acked loop in `reportFinishedSnapshotSplitsIfNeed`, the new `handleSourceEvent` handling of `CompletedSnapshotSplitsAckEvent`, and restoring `finishedUnackedSplits` via `addSplits` on failover. Could you point me to (or add) a test that exercises that restore-and-re-report path end to end? That's the exact gap this PR closes, and I'd like it locked down so the `TODO` doesn't quietly come back. Once `dev` is synced and the rerun is green (or the backpressure failure is shown to be pre-existing), I'll do a final pass and approve. <!-- streview-comment:1089 --> -- 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]
