Yao Xu has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/12323 )

Change subject: KUDU-2670: Part 1: Build ScanToken by KeyRange
......................................................................


Patch Set 33:

> > The reason for the problem should be that KeyRange is duplicate.
 > > In the following log, the tablet(79e027fcb6bd4f67b85028ab7d08352c)
 > primary key range is duplicate.
 > >
 > > 13:07:31.784 [DEBUG - New I/O worker #158] (AsyncKuduClient.java:2039)
 > Add key range [<start>, <end>), 40603, "79e027fcb6bd4f67b85028ab7d08352c"
 > [0x80000032, <end>)
 > > 13:07:31.784 [DEBUG - New I/O worker #157] (AsyncKuduClient.java:2039)
 > Add key range [<start>, <end>), 35286, "453294f2eb9341cfb4e59964404fc4e0"
 > [<start>, 0x80000032)
 > > 13:07:31.785 [DEBUG - New I/O worker #158] (AsyncKuduClient.java:2039)
 > Add key range [<start>, <end>), 40603, "79e027fcb6bd4f67b85028ab7d08352c"
 > [0x80000032, <end>)
 >
 > That sounds like it might be an interaction between split key range
 > requests and the DuplicatingRowSet, a special kind of rowset that
 > is used during a MRS flush. See Tablet::DoMergeCompactionOrFlush
 > for details.
 >
 > In any case, the problem is likely to be server-side, so I suggest
 > you add some debugging to Tablet and RowSetInfo to figure out
 > what's going on.

I found the root cause of this problem. Because Deferred.group no guarantee on 
the order of the results in the deferred list returned, but we rely on the 
return order when we call PartitionPruner.removePartitionKeyRange, This will 
cause some KeyRange to be created repeatedly. It's my fault. :(

Thank you again for your help. :)
Btw, dist-test seems to be a good test tools. Is there any doc for it?


--
To view, visit http://gerrit.cloudera.org:8080/12323
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: I0502f5d64569e8b1d45e88de3cb36aa2e01234d0
Gerrit-Change-Number: 12323
Gerrit-PatchSet: 33
Gerrit-Owner: yangz <zhe...@gmail.com>
Gerrit-Reviewer: Adar Dembo <a...@cloudera.com>
Gerrit-Reviewer: Grant Henke <granthe...@apache.org>
Gerrit-Reviewer: Kudu Jenkins (120)
Gerrit-Reviewer: Yao Xu <oclarms....@gmail.com>
Gerrit-Reviewer: yangz <zhe...@gmail.com>
Gerrit-Comment-Date: Thu, 23 May 2019 12:01:25 +0000
Gerrit-HasComments: No

Reply via email to