davidzollo commented on PR #11569: URL: https://github.com/apache/seatunnel/pull/11569#issuecomment-5339594060
Current head `03f7fd4347f7` has addressed the previous XA restore blocker and is ready for human re-review from my side. What changed after the 2026-08-06 review: - `restoreCommit()` no longer treats every `XAER_NOTA` as idempotent success. - Restore now runs an XA `recover()` scan, normalizes driver-specific `Xid` values by `formatId` + GTRID + BQUAL, and commits only checkpoint-owned XIDs that are still present in the resource manager recovery scan. - A missing checkpoint XID is only skipped when it is an already-resolved prefix before the first still-prepared transaction in the same commit batch; missing gaps/tails fail closed instead of being reported as successful. - Added regression coverage for recovered-by-value matching, unrelated recovered XIDs, already-resolved prefixes, no-evidence missing XIDs, missing gaps after a recovered transaction, transient recovery-scan retry, and bounded commit retry behavior. - Docs now describe the JDBC XA restore/recovery behavior and incompatible behavior change. Current CI note: the latest required Build on this head failed in `paimon-connector-it`, specifically `PaimonSinkCDCIT.testSinkWithIncompatibleSchema`, with a Paimon schema mismatch / container thread timeout. That job is outside this PR's JDBC XA source/test scope, so I do not see it as a code blocker for this PR. Could reviewers please re-check the latest head and clear the stale review gate if the restore reconciliation looks good? -- 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]
