Details
-
Bug
-
Status: Closed
-
Major
-
Resolution: Fixed
-
None
-
None
Description
When write task in a Flink job want to flush data, there is a condition that listened a new instant which have been start. If there is no new instant, the TM will wait for timeout.
We can see the code at https://github.com/apache/hudi/blob/master/hudi-flink-datasource/hudi-flink/src/main/java/org/apache/hudi/sink/common/AbstractStreamWriteFunction.java#L252.
Now there is a case that when the JM start new instant fail, JM won't retry this work. So how all the write tasks will hang.
Attachments
Issue Links
- links to