Vikas Basavaraj Mallapura
10/21/2024, 5:13 PMInvocation Guarantees
It is NOT guaranteed that the implementation will receive a notification for each completed or aborted checkpoint. While these notifications come in most cases, notifications might not happen, for example, when a failure/restore happens directly after a checkpoint completed.
Is there a guarantee that this listener will be invoked exactly-once for completed checkpoint though ? In my Flink job, it seems like if a job failure occurs soon after checkpoint completion, it gets called multiple times for the same checkpoint id.. Is that expected ?