AHeise commented on pull request #15252: URL: https://github.com/apache/flink/pull/15252#issuecomment-805987468
> The approach to setting the log level programmatically looks good to me. > > However I am a bit concerned that with the `TRACE` level enabled for `UnalignedCheckpointTestBase` the log files become huge. For the azure run for this PR, the size of mvn-2.log reaches ~2GB, whereas the latest successful run on master produces log files of around ~100MB. I am slightly concerned if we should enable it for all test runs... Thanks for the feedback. I investigated two reasons why the logs grew so big: - Buffers were logged even though respective checkpoint was cancelled and effectively NOT written (so a bug in tracing) - The test setup caused a long recovery time where many checkpoints were rapidly triggered and subsequently canceled. That could easily lead to 300 checkpoints in 1 min each leading to quite a few buffers being written before checkpoint was cancelled downstream. I modified the test to pump data on the source and sink in bursts equal to the checkpointing interval (so causing full buffers during checkpointing but not elsewhere). -- 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. For queries about this service, please contact Infrastructure at: us...@infra.apache.org
