Open rafael81 opened 1 year ago
I had a similar problem and I wanted to not keep retrying after a failed write. And I set the maximum number of retries does not take effect.
DorisExecutionOptions.Builder executionBuilder = DorisExecutionOptions.builder();
executionBuilder.setMaxRetries(2);
DorisSink.sink(
DorisReadOptions.builder().build(),
DorisExecutionOptions.builder()
//.setBatchSize(1)
//.setBatchIntervalMs(0L)
.setMaxRetries(1)
.setStreamLoadProp(properties)
.build(),
DorisOptions.builder()
.setFenodes(feNodes)
.setTableIdentifier(tableIdentifier)
.setUsername(username)
.setPassword(password)
.build()
);
Hi, we ran into the same issue. Anything new?
This happens during the Commit phase. The transaction ID recorded in the checkpoint has expired on the FE side. If you commit again, the above error will occur. You cannot start from the checkpoint at this time. You can extend the expiration time by modifying the streaming_label_keep_max_second configuration in fe.conf. The default is 12 hours.
BTW,The latest version also supports the sink.ignore.commit-error property. If you confirm that the txn has succeeded, you can skip the error.
Search before asking
Version
1.15 sink.max-retries = 1
What's Wrong?
retry infinite 2pc
What You Expected?
don't retry infinite 2pc which not found transaction in doris be.
How to Reproduce?
You can reproduce If you killed doris backend during sinking data through flink.
Anything Else?
No response
Are you willing to submit PR?
Code of Conduct