The GitHub Actions job "Flink CDC CI" on flink-cdc.git/fix_flaky_tests has failed. Run started by GitHub user leonardBang (triggered by leonardBang).
Head commit for run: 41147f2b080ea08df012c577081c79d45130d01a / Leonard Xu <[email protected]> [test][connector/oceanbase] Drain replayed no-PK snapshot rows synchronously in binlog phase to avoid an in-flight hasNext race The previous fix drained trailing at-least-once snapshot duplicates between the snapshot and binlog phases via hasNextData(iterator). On timeout that helper only calls executor.shutdown(), leaving the background iterator.hasNext() blocked on the shared CollectResultIterator. Once the binlog phase wrote data, that orphaned hasNext() could fetch and cache a record concurrently with the main thread's read, dropping the first binlog event and reintroducing flakiness. Remove the async drain. The snapshot phase now only reads (blocking, on the calling thread) until every expected row is observed with at least its expected multiplicity. Because the incremental source finishes the entire snapshot before reading binlog, any replayed snapshot +I rows strictly precede the changelog, so the binlog phase skips those leading snapshot rows synchronously before collecting the expected changelog. No background hasNext() is left in flight. Co-Authored-By: Claude Opus 4.7 <[email protected]> Report URL: https://github.com/apache/flink-cdc/actions/runs/27961381839 With regards, GitHub Actions via GitBox
