mynameborat commented on a change in pull request #1385:
URL: https://github.com/apache/samza/pull/1385#discussion_r464526206
##########
File path:
samza-core/src/main/scala/org/apache/samza/storage/ContainerStorageManager.java
##########
@@ -604,6 +597,9 @@ private StorageEngine createStore(String storeName,
TaskName taskName, TaskModel
Map<String, StorageEngine> sideInputStores =
getSideInputStores(taskName);
Map<String, Set<SystemStreamPartition>> sideInputStoresToSSPs = new
HashMap<>();
+ CountDownLatch taskCountDownLatch = new CountDownLatch(1);
+ this.sideInputTaskLatches.put(taskName, taskCountDownLatch);
Review comment:
Now that I think about it a bit more, I think you can do away without
latches here. Since you will anyways need some form of callback to the runloop
within CSM to denote completion of a message, can you actually not piggy back
on the callback for particular end offset to count down the latches? With this,
you can reuse the same approach for transactional state support for standbys.
Unless you have a rationale of handling that part differently and not tie
that into the initial catch up scenario.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
