j7nhai commented on code in PR #10249:
URL:
https://github.com/apache/incubator-gluten/pull/10249#discussion_r2227210687
##########
cpp/velox/memory/VeloxMemoryManager.cc:
##########
@@ -444,6 +444,13 @@ bool VeloxMemoryManager::tryDestructSafe() {
return true;
}
+void VeloxMemoryManager::finalDestructUnsafe() {
+ heldVeloxPools_.clear();
+ veloxLeafPool_.reset();
+ veloxAggregatePool_.reset();
+ veloxMemoryManager_.reset();
Review Comment:
I cannot reproduce this problem locally, it is triggered by our actual
production tasks. After such a small optimization, the stability can be
improved a lot. Extreme tasks also need to increase
`spark.gluten.sql.columnar.backend.velox.asyncTimeoutOnTaskStopping`, so I
think it is necessary to limit the interval to a maximum of 2s, otherwise the
spark task will have to wait for a long time after several attempts.
--
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]
