ferenc-csaky commented on PR #25955: URL: https://github.com/apache/flink/pull/25955#issuecomment-2605094044
After I rethink it again, it should probably be not enabled here, because the test should verify the `flink-runtime` shuffle memory allocation, which was on Netty4 back in 2019 as well when it was originally implemented, so it used the `pooled` allocator since ever. Enabling reflection only seemed to stabilize the test on my machine, but since this test was flaky on the CI failed sporadically, and on my machine consistently, it may not add much. My concern about the backported commit from `master` is that it gives 90MB memory instead of 7. It states `one chunk (4M * 20) plus some margins`, which might be true for `master` if there were some other changes how we (did not follow this up), but definitely not true for 1.19 and 1.20, so IMO it invalidates this test case. -- 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: issues-unsubscr...@flink.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
