zuston commented on code in PR #494:
URL: https://github.com/apache/incubator-uniffle/pull/494#discussion_r1073250262
##########
server/src/main/java/org/apache/uniffle/server/ShuffleTaskManager.java:
##########
@@ -574,11 +584,20 @@ public void removeResources(String appId) {
new AppPurgeEvent(appId, getUserByAppId(appId), new
ArrayList<>(shuffleToCachedBlockIds.keySet()))
);
}
+ if (shuffleTaskInfo.hasHugePartition()) {
+ ShuffleServerMetrics.gaugeAppWithHugePartitionNum.dec();
+
ShuffleServerMetrics.gaugeHugePartitionNum.dec(shuffleTaskInfo.getHugePartitionSize());
+ }
LOG.info("Finish remove resource for appId[" + appId + "] cost " +
(System.currentTimeMillis() - start) + " ms");
}
public void refreshAppId(String appId) {
- shuffleTaskInfos.computeIfAbsent(appId, x -> new
ShuffleTaskInfo()).setCurrentTimes(System.currentTimeMillis());
+ shuffleTaskInfos.computeIfAbsent(
+ appId,
+ x -> {
+ ShuffleServerMetrics.counterTotalAppNum.inc();
Review Comment:
> My whole point is that shuffleTaskInfos cannot be fully trusted to
deduplicate appId.
Yes, the problem you mentioned is a corner case. But for metric calculating,
I don't think we should make too much effort and introduce too much complexity
to improve accuracy for 0.1 percent possibility.
For this case, I could delete this metric of total_app_num, leaving someone
having better solution to support
--
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]
