[
https://issues.apache.org/jira/browse/BIGTOP-1560?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14234864#comment-14234864
]
Konstantin Boudnik commented on BIGTOP-1560:
--------------------------------------------
Patch looks good in a quick glance. A couple of suggestions:
- {{+ assert (numberOfDataNodes >= 3) : "Blocks cannot recover when there
is only one datanode."}}
I would suggest to use Assume.assumeTrue() to simply skip the test with a
message is the condition isn't met
- considering that you are doing a number of operations under {{hdfs}} user, it
would make sense to either skip or fail the test if current user is different.
So, perhaps you can {{USERNAME}} is set as desired.
Thanks for doing this - pretty complex test, BTW!
> Add a test case for performing block corruption recovery
> --------------------------------------------------------
>
> Key: BIGTOP-1560
> URL: https://issues.apache.org/jira/browse/BIGTOP-1560
> Project: Bigtop
> Issue Type: Test
> Components: tests
> Affects Versions: 0.8.0
> Reporter: Dasha Boudnik
> Fix For: 0.9.0
>
> Attachments: BIGTOP-1560.patch
>
>
> Found this issue in internal testing. Roughly:
> create file in HDFS
> find block for the file
> corrupt a block
> trigger recovery by trying to read the file
> check recovery was successful
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
