[
https://issues.apache.org/jira/browse/HDFS-10723?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15421759#comment-15421759
]
Vrushali C commented on HDFS-10723:
-----------------------------------
Yes, the test does seem to require binding to the previously bound port in that
test. My guess is that when the port is set free after the mini cluster is
brought down and the test is trying to bind to the same port again, some other
process grabs that port in the meantime. I am wondering how this can be
controlled. Need to look deeper into how available ports are returned for bind
calls with port 0.
> TestRenameWhileOpen tries to restart NN with the same rpc port
> --------------------------------------------------------------
>
> Key: HDFS-10723
> URL: https://issues.apache.org/jira/browse/HDFS-10723
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: test
> Reporter: Kihwal Lee
> Attachments: testWhileOpenRenameParentToNonexistentDir.txt
>
>
> In {{testWhileOpenRenameParentToNonexistentDir}},
> {code}
> // restart cluster with the same namenode port as before.
> // This ensures that leases are persisted in fsimage.
> cluster.shutdown();
> {code}
> And we are getting test failures in precommit builds.
> {noformat}
> java.net.BindException: Problem binding to [localhost:47395]
> java.net.BindException: Address already in use;
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
