gaoyunhaii commented on pull request #16800: URL: https://github.com/apache/flink/pull/16800#issuecomment-898365742
Hi @pnowojski , very thanks for opening the PR! It seems to me the current implementation get the result whether the trigger is successful only after it actually execute the synchronous part? It seems we might need to be that strict: currently as long as `invokable.triggerCheckpointAsync` is called without exception, we could ensure the checkpoint must be performed. This is due to if the method is successful, it ensures the mailbox is not `prepareClose()` and `isRunning = true`. Since `isRunning` is set to `false` after the mailbox is drained, thus we ensures when this mail is processed and the checkpoint is triggered, isRunning must be `true`. Although logically we only care about the "false" result, but since Akka has a timeout, thus if the mail queued for a long time or the synchronous part takes long time, we might meet with AkkaAskTimeout and cancel the checkpoint wrongly? -- 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]
