DanielLeens commented on issue #11816: URL: https://github.com/apache/seatunnel/issues/11816#issuecomment-5307586271
Thanks for picking up this follow-up. I re-checked the current `dev` connector path before replying here. The gap is still real: `PrometheusWriter` still relies on three flush triggers only: `batch_size`, the engine-level `registerFlushAction(...)`, and `close()`. In the same code path it still documents that Spark/Flink keep the sink-writer context's default no-op `registerFlushAction`, so below `batch_size` there is still no periodic flush there. By contrast, the sibling buffered sinks already flush in `prepareCommit()`, so bounding this by checkpoint is a reasonable connector-level first slice. Your proposed scope looks right for v1: 1. add a `prepareCommit()` flush to `PrometheusWriter`; 2. add a regression test for the checkpoint-flush path; 3. update the Prometheus docs / compatibility note so Spark/Flink are documented as checkpoint-bounded rather than no periodic flush at all. I also tried to assign this issue to you, but GitHub rejected the assignee target from my side, so that part did not stick. Please go ahead with the PR path and link it here once it is open. -- 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]
