accumulo-notifications mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Josh Elser (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (ACCUMULO-3600) ShellServerIT.deleteRows might fail
Date Tue, 17 Feb 2015 23:32:11 GMT

    [ https://issues.apache.org/jira/browse/ACCUMULO-3600?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14325115#comment-14325115
] 

Josh Elser commented on ACCUMULO-3600:
--------------------------------------

Trying to reproduce this, I think it's even more subtle that I initially described. Increasing
the time between each update doesn't the test to fail, so I'm guessing that the previous test
cases happened to skew in such a way that one tabletserver was *just about* out of balance
and the table created by deleteRows pushed it over the edge?

> ShellServerIT.deleteRows might fail
> -----------------------------------
>
>                 Key: ACCUMULO-3600
>                 URL: https://issues.apache.org/jira/browse/ACCUMULO-3600
>             Project: Accumulo
>          Issue Type: Bug
>          Components: test
>    Affects Versions: 1.6.1
>            Reporter: Josh Elser
>            Assignee: Josh Elser
>            Priority: Minor
>             Fix For: 1.7.0, 1.6.3
>
>
> Saw a failure in {{ShellServerIT#deleteRows}} where one tablet had 2 files instead of
the expected one. The test roughly does the following:
> 1. Create a table
> 2. Add 2 splits to that table (3 tablets)
> 3. Add 10 records (split "evenly" across the tablets)
> 4. Flush the table (spill memory to disk)
> 5. Check that each tablet has one file.
> The problem with the test is that while step 3 is happening, one of the new tablets could
be migrated to the other tablet server. This would cause a minc (and write a file). Step3
would eventually finish, and then step4 would cause another minc (and file) on the tablet
that was migrated. This would result in two rfiles for the tablet instead of the expected
one.
> Need to update the test so that it accounts for this possibility.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message