yidaimi commented on PR #13200: URL: https://github.com/apache/seatunnel/pull/13200#issuecomment-7032541536
@DanielLeens Thanks, both blockers are addressed: 1. **Upgrade note scoped to non-transactional writers** (`ab6d2250b`, en and zh). The behavior change now names `transaction = false` (the default). It also states that `transaction = true` sinks are not affected, because they keep the rows until the next checkpoint and the committer writes them in one transaction, using neither the buffer options nor `retry.*`. The migration guide now says that a smaller `buffer-flush.max-rows` above 1 only reduces how many rows are buffered and does not bound how long a partial batch waits. For a time bound, rely on checkpoints or `sink.flush.interval` (Zeta). For immediate writes, use `buffer-flush.max-rows = 1`. The PR description is aligned. 2. **Conflict with `dev`.** Rebased onto the latest `dev`. The MongoDB entry stays at the top of `## dev`, and the branch merges cleanly again. From the non-blocking suggestions, `MongodbIT` now cleans up both collections in `finally`. I kept the reflective last-send time in the unit test: a controllable time source would need a clock hook in `MongodbWriter`, and I'd rather not widen the production change of this fix. Happy to do it in a follow-up if you prefer. Local checks on the new head: `spotless:check`, `seatunnel-ci-tools`, mongodb unit tests (71), and `MongodbIT#testSinkWithOmittedBufferAndRetryOptions` on Zeta (3 insert commands for 1234 rows). The fork Build on `ab6d2250b` is green: https://github.com/yidaimi/seatunnel/actions/runs/44678509159. `MongodbIT` passed with 96 tests on Flink, Spark and Zeta, and the new test logged `3 insert commands for 1234 rows` on all seven containers. -- 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]
