[
https://issues.apache.org/jira/browse/HBASE-13590?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15098527#comment-15098527
]
Yu Li commented on HBASE-13590:
-------------------------------
[~stack] Thanks for checking this.
These lines follow the same logic as the code in trunk (or say the original
patch in HBASE-13041). I think they are doing the same thing as the original
{{m.getAssignmentManager().assign(admin.getTableRegions(tableName));}}. I'm ok
to roll back these lines to the single line one, but just would like to mention
that the 100 times run I executed was with current patch, although it should be
the same in theory. Just let me know if you insist. :-)
According to our investigation, regions would get assigned during regionserver
start, and the same regions assigned again by the later codes. The two assign
plan conflict somehow thus cause the hang. And I think the real fix in the
patch is the {{cluster.waitForRegionServerToStart}} line
Honestly, I have no idea why we should assign the regions again after
regionserver is up. IMHO design of this test case is to make sure that table
enabled when no RS up could get assigned after RS come back, so the second
assign-action seems useless to me.
> TestEnableTableHandler.testEnableTableWithNoRegionServers is flakey
> -------------------------------------------------------------------
>
> Key: HBASE-13590
> URL: https://issues.apache.org/jira/browse/HBASE-13590
> Project: HBase
> Issue Type: Test
> Components: master
> Reporter: Nick Dimiduk
> Assignee: Yu Li
> Fix For: 2.0.0, 1.3.0, 1.2.1, 1.1.4
>
> Attachments: HBASE-13590.branch-1.patch
>
>
> Looking at our [build
> history|https://builds.apache.org/job/HBase-1.1/buildTimeTrend], it seems
> this test is flakey. See builds 429, 431, 439.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
