mini666 commented on PR #8380: URL: https://github.com/apache/hbase/pull/8380#issuecomment-4988881493
I pushed an update addressing the review feedback. This revision focuses on the correctness issue in this PR: preventing duplicate execution of replicated bulk-load events on retry/timeout paths. I split the event tracker into an interface and kept the current implementation backed by ZooKeeper. That keeps the sink-side lifecycle independent from the backend, while leaving a table-backed tracker as a possible follow-up without expanding this PR into a larger storage change. The important behavior change is table-level completion. Events for a table are marked done as soon as that table is loaded successfully, and if a later table fails, already-loaded events are not released as retryable work. I also tightened the ZK claim path so it re-checks the done marker after creating the in-progress marker, and best-effort cleans up the marker this call created on the relevant post-create failure paths. Local verification passed: * Focused tests: 9 tests, 0 failures, 0 errors. * Cross-RS duplicate bulk-load regression test: 1 test, 0 failures, 0 errors. * `git diff --cached --check` before commit. -- 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]
