hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Todd Lipcon (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HDFS-1766) Datanode is marked dead, but datanode process is alive and verifying blocks
Date Thu, 17 Mar 2011 22:24:29 GMT

    [ https://issues.apache.org/jira/browse/HDFS-1766?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13008193#comment-13008193

Todd Lipcon commented on HDFS-1766:

Another way to go about this would be to set up an UncaughtExceptionHandler on the main threads
that does System.exit(1) -- that would also deal with cases like OutOfMemoryError, which you're
currently not catching.

I think we should either do that, or modify your patch to catch all Throwable

> Datanode is marked dead, but datanode process is alive and verifying blocks
> ---------------------------------------------------------------------------
>                 Key: HDFS-1766
>                 URL: https://issues.apache.org/jira/browse/HDFS-1766
>             Project: Hadoop HDFS
>          Issue Type: Bug
>          Components: data-node
>    Affects Versions: 0.23.0
>            Reporter: Hairong Kuang
>            Assignee: Hairong Kuang
>             Fix For: 0.23.0
>         Attachments: killDN.patch
> We have a datanode marked dead in the namenode, and it is not taking any traffic. But
it is verifying blocks continuously, so the DataNode process is definitely not dead. Jstack
shows that the main thread and the offerService thread are gone but the JVM stuck at waiting
for other threads to die. It seems to me that the offerService thread has died abnormally,
for example, by a runtime exception and it did not shut down other threads before exiting.

This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira

View raw message