DanielLeens commented on PR #12789: URL: https://github.com/apache/seatunnel/pull/12789#issuecomment-6266743341
Thanks @dybyte for the careful look, and +1 to your points. They line up with what I raised in my earlier review on this head (a5aad37): - Upper bound or WARN for large values: this is my Issue 1. Beyond the heap, `HazelcastCheckpointOverviewStateStore.updateOverview` rewrites the whole `CheckpointOverview` through `IMap.compute` on every checkpoint event, so a very large value multiplies serialization and backup traffic on the master. A bounded setter or a startup WARN would address it. - Adding the key to `hybrid-cluster-deployment.md` (en/zh) next to `min-pause`: this is my Issue 3. On time-based retention: I agree it matches the stated motivation better, since `checkpoint.interval` is per job while this option is cluster-wide. I think a count-based option is still a reasonable first step because the default is unchanged and the wiring is tiny, so I would be fine with time-based retention as a follow-up, as you suggested. @SEZ9 it would help to note this limitation in the option description so operators know that the covered time window depends on the job's checkpoint interval. No new commit has landed since my review, so the review conclusion is unchanged. -- 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]
