curcur commented on PR #20091: URL: https://github.com/apache/flink/pull/20091#issuecomment-1174569765
> Got it, thanks. My concern is that probably we should fix the error handling in `CheckpointFailureManager`, instead of `mkdirs` call: > > ``` > case FINALIZE_CHECKPOINT_FAILURE: > // ignore > break; > ``` > > The [documentation](execution.checkpointing.tolerable-failed-checkpoints) for `execution.checkpointing.tolerable-failed-checkpoints` says that it applies to any `IOException on the Job Manager`. This is the case here (`IOException` gets wrapped into `CheckpointException` when I try it locally). > > By adding a check of `mkdirs`, we do fix this particular problem; but any other `IOException` (e.g. intermittent failure when writing the `_metadata` file) will be ignored by `CheckpointFailureManager`. > > So how about counting `FINALIZE_CHECKPOINT_FAILURE` as a real failure in `CheckpointFailureManager`? We could still fail fast if the directories could not be created, but that would be just an optimization then. @pnowojski Do you have an idea why `FINALIZE_CHECKPOINT_FAILURE` is not counted as a checkpoint failure? Looks strange to me as well. -- 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