Will Berkeley has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/13264 )

Change subject: KUDU-2807 Crash when flush or compaction overlaps with another 
compaction
......................................................................


Patch Set 1:

(1 comment)

> Would be great to test this. I am surprised mt-tablet-test didn't
 > trigger the crash; could you take a look?

I think we are testing for it as the mt-tablet-test has threads flushing and 
compacting and GCing undos. Presumably it's just really unlikely because the 
compact_flush_lock_ is held so briefly for each rowset when recomputing the 
average height. What's interesting is that the reporter reports running into it 
on multiple servers consistently. Maybe the most efficient thing to do it 
investigate that report to understand why they are seeing this, if this patch 
fixes them, and if we can write a more targeted test based off what we learn 
from them.

http://gerrit.cloudera.org:8080/#/c/13264/1/src/kudu/tablet/tablet.cc
File src/kudu/tablet/tablet.cc:

http://gerrit.cloudera.org:8080/#/c/13264/1/src/kudu/tablet/tablet.cc@1753
PS1, Line 1753: Status Tablet::HandleEmptyCompactionOrFlush(const RowSetVector& 
rowsets,
> Shouldn't we recompute the rowset height in the event of an empty compact/f
My bad. Missed this.



--
To view, visit http://gerrit.cloudera.org:8080/13264
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: Ic255f0466aa2c158fa32e8e38428eddfcf901b99
Gerrit-Change-Number: 13264
Gerrit-PatchSet: 1
Gerrit-Owner: Will Berkeley <[email protected]>
Gerrit-Reviewer: Adar Dembo <[email protected]>
Gerrit-Reviewer: Kudu Jenkins (120)
Gerrit-Reviewer: Will Berkeley <[email protected]>
Gerrit-Comment-Date: Wed, 08 May 2019 19:50:25 +0000
Gerrit-HasComments: Yes

Reply via email to