[
https://issues.apache.org/jira/browse/HADOOP-4885?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13229819#comment-13229819
]
Brandon Li commented on HADOOP-4885:
------------------------------------
The format-addStorageDir solution make the failed directory "active"
immediately even it's not a real active state. The state is visible from the nn
UI and JMX. If the checkpoint fails, the fake "Active" state can be misleading.
The copy-over solution may do some extra work but it sets the recovered storage
directories in the real active state.
I agree those 3 JIRA issues you mentioned should be back ported too to branch
1.02 (the backport patch here is for branch-1 not 1.02).
It's a good point about the network mount problem. :-)
It's also a problem with original patch: the "format-addStorageDir" creates the
storage directory if it doesn't exist. However, if this storage directory is a
mount point, it shouldn't be created automatically. HDFS-3095 is filed for this
issue.
> Try to restore failed replicas of Name Node storage (at checkpoint time)
> ------------------------------------------------------------------------
>
> Key: HADOOP-4885
> URL: https://issues.apache.org/jira/browse/HADOOP-4885
> Project: Hadoop Common
> Issue Type: Improvement
> Reporter: Boris Shkolnik
> Assignee: Boris Shkolnik
> Fix For: 0.21.0
>
> Attachments: HADOOP-4885-1.patch, HADOOP-4885-3.patch,
> HADOOP-4885-3.patch, HADOOP-4885.branch-1.patch,
> HADOOP-4885.branch-1.patch.2, HADOOP-4885.branch-1.patch.3,
> HADOOP-4885.patch, HADOOP-4885.patch
>
>
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
