vahmed-hamdy commented on a change in pull request #18945:
URL: https://github.com/apache/flink/pull/18945#discussion_r816889853
##########
File path:
flink-connectors/flink-connector-base/src/main/java/org/apache/flink/connector/base/sink/writer/AsyncSinkWriter.java
##########
@@ -51,7 +52,8 @@
*/
@PublicEvolving
public abstract class AsyncSinkWriter<InputT, RequestEntryT extends
Serializable>
- implements StatefulSink.StatefulSinkWriter<InputT,
BufferedRequestState<RequestEntryT>> {
+ implements StatefulSink.StatefulSinkWriter<InputT,
BufferedRequestState<RequestEntryT>>,
+ TwoPhaseCommittingSink.PrecommittingSinkWriter<InputT, Void> {
Review comment:
After a second look.
I believe this issue should be resolved, as per this
[line](https://github.com/apache/flink/blob/dcd215620d96c5e648cea777f0b7594f7b5a1e26/flink-streaming-java/src/main/java/org/apache/flink/streaming/runtime/operators/sink/SinkWriterOperator.java#L147)
the `flush` method is used to notify the writer with the upcoming snapshot.
It appears that the `prepareCommit` interface is only used in 2PC cases and
for operators that would separate building the immutable state object and
returning it in `snapshotState`. Since `AsyncSinkWriter` creates the state
object on returning it, I think this PR is not needed.
--
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]
