hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Boris Shkolnik (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HDFS-457) better handling of volume failure in Data Node storage
Date Fri, 07 Aug 2009 18:59:14 GMT

    [ https://issues.apache.org/jira/browse/HDFS-457?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12740680#action_12740680

Boris Shkolnik commented on HDFS-457:

thanks for the comments.

>>    *  Also, would the new checkDataDir() implementation take a long time to execute?
We may need some performance test on this.). 
This function makes one loop over ALL the blocks. I believe this is what causes the concern.
I've looked at it and ran some tests.
1. This loop will happen only in case when there is an actual failure. I think it is expected
to have some extra work to be done in this case.
2. This extra work shouldn't be so bad. I've ran test for 1500 blocks it took about 600mls,
for 5000 about 800mls. For huge nodes with hundreds of thousands blocks it still should be

> better handling of volume failure in Data Node storage
> ------------------------------------------------------
>                 Key: HDFS-457
>                 URL: https://issues.apache.org/jira/browse/HDFS-457
>             Project: Hadoop HDFS
>          Issue Type: Bug
>          Components: data-node
>            Reporter: Boris Shkolnik
>            Assignee: Boris Shkolnik
>         Attachments: HDFS-457.patch
> Current implementation shuts DataNode down completely when one of the configured volumes
of the storage fails.
> This is rather wasteful behavior because it  decreases utilization (good storage becomes
unavailable) and imposes extra load on the system (replication of the blocks from the good
volumes). These problems will become even more prominent when we move to mixed (heterogeneous)
clusters with many more volumes per Data Node.

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

View raw message