The GitHub Actions job "Flink CDC CI" on flink-cdc.git/FLINK-40538 has 
succeeded.
Run started by GitHub user eskabetxe (triggered by loserwang1024).

Head commit for run:
4f2645fc863e575617bf56a94af091b6154dba8e / Joao Boto <[email protected]>
[FLINK-40538][postgres] Skip WAL position search on idle publication for a 
fresh start

PostgresSourceFetchTaskContext.loadStartingOffsetState always returns a
non-null PostgresOffsetContext built from the stream split's starting offset,
so in the forked PostgresStreamingChangeEventSource#execute the WAL position
search branch was taken unconditionally. On an idle publication that search
loops forever waiting for a decoded message: no WAL is produced, and the
heartbeat action query that would generate some only runs from the main
streaming loop, which the search precedes. The job stays RUNNING while the
slot's confirmed_flush_lsn never advances and WAL grows without bound.

Guard the search with offsetContext.hasCompletelyProcessedPosition(), mirroring
the fix Debezium shipped in 2.7. On a fresh start (nothing processed yet) the
search is skipped; streaming still starts from the stored LSN, so no events are
missed. On a resumed offset the search still runs.

Report URL: https://github.com/apache/flink-cdc/actions/runs/36489059064

With regards,
GitHub Actions via GitBox

Reply via email to