DanielLeens commented on issue #12132: URL: https://github.com/apache/seatunnel/issues/12132#issuecomment-5554353694
Correction to the last paragraph above, after reading the code: `KuduSource#createEnumerator` only constructs `KuduSourceSplitEnumerator`, whose constructor stores the config and builds a `KuduInputFormat` without touching Kudu; the client is created in `open()` (`KuduInputFormat#openInputFormat`), which the Flink translation calls from `FlinkSourceEnumerator#start` inside Flink's coordinator executor, not on the JobMaster main thread. So the synchronous-`createEnumerator` explanation does not hold. What remains established is the observable pattern: first job on a freshly started JobManager/TaskManager pair, JobMaster heartbeats stop at job start, JobMaster-side timeouts fire 120 s later, the job neither runs nor fails, and (in the 1.18 variant) the TaskManager also fails to fetch the job BLOB from the JobManager. A JobManager thread dump from a reproducing run is needed to see what the JobMaster main thread and the BlobServer are blocked on. -- 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]
