[
https://issues.apache.org/jira/browse/HDFS-8664?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14609253#comment-14609253
]
Arpit Agarwal edited comment on HDFS-8664 at 6/30/15 11:36 PM:
---------------------------------------------------------------
Hi [~patrickwhite], likely one of the test cases in {{TestBPOfferService}}
fails to clean up after itself so subsequent tests fail. One issue we used to
see was failing to shut down the MiniDFSCluster. Junit now randomizes the
execution order of individual test cases so that adds some debugging complexity
but it also increases the chances of catching such spurious dependencies.
Your best bet would be code inspection of tests for incorrect cleanup.
was (Author: arpitagarwal):
Hi [~patrickwhite], likely one of the test cases fails to clean up after itself
so subsequent tests fail. One issue we used to see was failing to shut down the
MiniDFSCluster. Junit now randomizes the execution order of individual test
cases so that adds some debugging complexity but it also increases the chances
of catching such spurious dependencies.
Your best bet would be code inspection of tests for incorrect cleanup.
> Allow wildcards in dfs.datanode.data.dir
> ----------------------------------------
>
> Key: HDFS-8664
> URL: https://issues.apache.org/jira/browse/HDFS-8664
> Project: Hadoop HDFS
> Issue Type: New Feature
> Components: datanode, HDFS
> Affects Versions: 3.0.0
> Reporter: Patrick White
> Assignee: Patrick White
> Attachments: HDFS-8664.001.patch, HDFS-8664.002.patch,
> HDFS-8664.003.patch, TestBPOfferService-output.txt
>
>
> We have many disks per machine (12+) that don't always have the same
> numbering when they come back from provisioning, but they're always in the
> same tree following the same pattern.
> It would greatly reduce our config complexity to be able to specify a
> wildcard for all the data directories.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
