[GitHub] [hudi] danny0405 commented on a diff in pull request #7626: [HUDI-5516] Reduce memory footprint on workload with thousand active partitions

[GitBox <gi...@apache.org>]

danny0405 commented on code in PR #7626:
URL: https://github.com/apache/hudi/pull/7626#discussion_r1064456753


##########
hudi-client/hudi-flink-client/src/main/java/org/apache/hudi/client/HoodieFlinkWriteClient.java:
##########
@@ -94,7 +94,7 @@
    * FileID to write handle mapping in order to record the write handles for each file group,
    * so that we can append the mini-batch data buffer incrementally.
    */
-  private final Map<String, HoodieWriteHandle<?, ?, ?, ?>> bucketToHandles;
+  private final Map<String, MiniBatchHandle> bucketToHandles;
 

Review Comment:
   Can we just cache the write path instead of write handles to not introduce too many unnecessary changes.



##########
hudi-flink-datasource/hudi-flink/src/main/java/org/apache/hudi/sink/StreamWriteFunction.java:
##########
@@ -449,6 +450,7 @@ private boolean flushBucket(DataBucket bucket) {
 
     this.eventGateway.sendEventToCoordinator(event);
     writeStatuses.addAll(writeStatus);
+    writeClient.cleanHandle(bucket.fileID);
     return true;
   }

Review Comment:
   Do we need to clean the handles eagerly for `#flushRemaining` ?



-- 
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: commits-unsubscribe@hudi.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org