[
https://issues.apache.org/jira/browse/PHOENIX-5348?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16905366#comment-16905366
]
Gokcen Iskender commented on PHOENIX-5348:
------------------------------------------
[~lhofhansl], while I was working on this patch, I realized that we have some
bugs in the code itself. One of the bugs is we did not take into account where
a table name and index name and task type can be same but timestamp is
different. Originally the code was picking just the first task (that task could
be completed but there might be another task with a newer timestamp which is
running). I found that issue while looking at Chinmay's comments. This fix also
queries with timestamp to eliminate the scenario I described.
The second thing I addressed was this issue: In the test, I was disabling index
table and truncating it. Index tool returns success but no rows get populated.
The reason for that is, phoenix is not aware of hbase's disable and thinks that
the index table did not change.
I hope this explains a bit why it got bigger.
> IndexRebuildTaskIT.testIndexRebuildTask flaps (Jenkins only)
> ------------------------------------------------------------
>
> Key: PHOENIX-5348
> URL: https://issues.apache.org/jira/browse/PHOENIX-5348
> Project: Phoenix
> Issue Type: Bug
> Reporter: Lars Hofhansl
> Assignee: Gokcen Iskender
> Priority: Critical
> Fix For: 4.15.0, 5.1.0
>
> Attachments: 5348-disable.txt, 5348-v2.txt, 5348.txt,
> PHOENIX-5348.patch
>
> Time Spent: 7h 40m
> Remaining Estimate: 0h
>
> This failed in the past four builds:
> {code}
> [ERROR] testIndexRebuildTask(org.apache.phoenix.end2end.IndexRebuildTaskIT)
> Time elapsed: 20.672 s <<< FAILURE!
> java.lang.AssertionError: expected:<1> but was:<0>
> at
> org.apache.phoenix.end2end.IndexRebuildTaskIT.testIndexRebuildTask(IndexRebuildTaskIT.java:164)
> {code}
> But... It does not fail locally for me.
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
