advancedxy commented on code in PR #494:
URL: https://github.com/apache/incubator-uniffle/pull/494#discussion_r1073273322
##########
server/src/main/java/org/apache/uniffle/server/ShuffleTaskManager.java:
##########
@@ -322,16 +322,22 @@ public void updateCachedBlockIds(String appId, int
shuffleId, int partitionId, S
size += spb.getSize();
}
}
- shuffleTaskInfo.addPartitionDataSize(
+ long partitionSize = shuffleTaskInfo.addPartitionDataSize(
shuffleId,
partitionId,
size
);
+ if (shuffleBufferManager.isHugePartition(partitionSize)) {
Review Comment:
> One appId only has one shuffleTaskInfo.
My fault. The class name is kind of misleading, I used to think this class
corresponding to one shuffle process/task, mapping to one shuffleId in spark
app.
> How to maintain `hasSeenAppIdSet`'s lifecycle?
If we are going to this way. `hasSeenAppIdSet` would be a LRU with maximum
size(let's say 10W) set. It should be independent with `shuffleTaskInfos`.
But that may be a little bit complex..
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
