[
https://issues.apache.org/jira/browse/HDFS-1505?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13036376#comment-13036376
]
Matt Foley commented on HDFS-1505:
----------------------------------
Reading the HDFS-1073 spec, I infer that fsimage files will have a tag
identifying the last txn included in the image, and edits logs will have tags
for the first and last txn included in them. And you're referring to the
resulting fact that one could take an image ending with txn 100, jump into the
middle of a log file that went from txn 50 to 170, and successfully generate
the in-memory structures current as of txn 170. Is that right?
If the above understanding is correct, then I agree it seems that
saveNamespace() should just save the fsimage file. Although it doesn't hurt to
also clear the edits logs, once you have multiple copies of the fsimage. Does
your log-rolling logic automatically delete log chunk files older than
available fsimage files? That would be sufficient edits file management.
> saveNamespace appears to succeed even if all directories fail to save
> ---------------------------------------------------------------------
>
> Key: HDFS-1505
> URL: https://issues.apache.org/jira/browse/HDFS-1505
> Project: Hadoop HDFS
> Issue Type: Bug
> Affects Versions: 0.22.0, 0.23.0
> Reporter: Todd Lipcon
> Assignee: Aaron T. Myers
> Priority: Blocker
> Fix For: 0.22.0
>
> Attachments: hdfs-1505-1-test.txt, hdfs-1505-22.0.patch,
> hdfs-1505-22.1.patch, hdfs-1505-22.2.patch, hdfs-1505-test.txt,
> hdfs-1505-trunk.0.patch, hdfs-1505-trunk.1.patch, hdfs-1505-trunk.2.patch,
> hdfs-1505-trunk.3.patch
>
>
> After HDFS-1071, saveNamespace now appears to "succeed" even if all of the
> individual directories failed to save.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
