[
https://issues.apache.org/jira/browse/GEODE-2459?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15861959#comment-15861959
]
ASF subversion and git services commented on GEODE-2459:
--------------------------------------------------------
Commit dbea592cc96f64c9fbc7abf8bbf597a69c5881cd in geode's branch
refs/heads/develop from [~huynhja]
[ https://git-wip-us.apache.org/repos/asf?p=geode.git;h=dbea592 ]
GEODE-2459: When node fails while renaming, new primary node resumes and may
delete chunks
* Rename will set a flag on the renamed file before continuing
* Deleting possibly renamed files will not delete chunks
* This means we may have lingering chunks if a node crashes but we should no
longer
see EOF exceptions due to missing chunks
> When node fails while renaming, new primary node resumes and may delete chunks
> ------------------------------------------------------------------------------
>
> Key: GEODE-2459
> URL: https://issues.apache.org/jira/browse/GEODE-2459
> Project: Geode
> Issue Type: Bug
> Components: lucene
> Affects Versions: 1.0.0-incubating
> Reporter: Jason Huynh
> Assignee: Jason Huynh
>
> Current order of operations of a rename:
> 1.) Copy "old file" to "new file" File object
> 2.) Put into file region new file object
> 3.) Call remove directly on fileRegion for the "old file"
> When a node closes before step 3, the old file still exists in the file
> region.
> The new primary node will start up and when removing this old file (during
> normal lucene operations), it will erase the chunks. However the "new file"
> is still present and expects the chunks to be there.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
