Jackie-Jiang commented on code in PR #19468:
URL: https://github.com/apache/pinot/pull/19468#discussion_r4087665815
##########
pinot-core/src/main/java/org/apache/pinot/core/data/manager/realtime/RealtimeTableDataManager.java:
##########
@@ -542,7 +550,29 @@ private void doAddConsumingSegment(String segmentName)
return;
}
IndexLoadingConfig indexLoadingConfig = fetchIndexLoadingConfig();
- handleSegmentPreload(zkMetadata, indexLoadingConfig);
+ LLCSegmentName llcSegmentName = new LLCSegmentName(segmentName);
+ int partitionGroupId = llcSegmentName.getPartitionGroupId();
+ PartitionUpsertMetadataManager partitionUpsertMetadataManager;
+ PartitionDedupMetadataManager partitionDedupMetadataManager;
+ synchronized (_segmentDataManagerMap) {
Review Comment:
An admitted CONSUMING callback is not included in
BaseTableDataManager.shutDown()'s wait. It can pass this check, pause, let
deleteTable() finish and a same-name table start, then resume through
FileUtils.deleteQuietly(_indexDir/segmentName) and the MmapMemoryManager
constructor before the later shutdown check. LLC names use minute precision and
a recreated table starts its partition sequence at 0, so a same-minute
recreation can reuse the name and the old callback can delete the new
consumer's files. Could shutdown drain admitted consuming callbacks before
releasing the table lifecycle lock? A regression test should pause the old
callback before filesystem cleanup while the replacement uses the shared data
directory.
--
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]
