hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Kihwal Lee (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (HDFS-7046) HA NN can NPE upon transition to active
Date Fri, 12 Sep 2014 14:55:34 GMT

     [ https://issues.apache.org/jira/browse/HDFS-7046?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel

Kihwal Lee updated HDFS-7046:
    Attachment: HDFS-7046_test_reproduce.patch

Attaching a patch with a test case that reproduces the execution path.  On transitioning to
active, the standby can come out of safe mode in the middle of replaying edits.  The patch
also includes partial fix for demonstration purpose.

The test case will fail, because with the partial fix, it does not enter safe mode extension
and the safe mode message is different.  If the partial fix in FSNamesystem.java is removed,
the test will pass since it is already coming out of safe mode.

The complete fix will include {{checkSafeMode()}} call at the end of transitioning, so that
it can come out of same mode even if the cluster is completely idle.

> HA NN can NPE upon transition to active
> ---------------------------------------
>                 Key: HDFS-7046
>                 URL: https://issues.apache.org/jira/browse/HDFS-7046
>             Project: Hadoop HDFS
>          Issue Type: Bug
>          Components: namenode
>    Affects Versions: 3.0.0, 2.5.0
>            Reporter: Daryn Sharp
>            Priority: Critical
>         Attachments: HDFS-7046_test_reproduce.patch
> While processing edits, the NN may decide after adjusting block totals to leave safe
mode - in the middle of the edit.  Going active starts the secret manager which generates
a new secret key, which in turn generates an edit, which NPEs because the edit log is not
> # Transitions should _not_ occur in the middle of an edit.
> # The edit log appears to claim it's open for write when the stream isn't even open

This message was sent by Atlassian JIRA

View raw message