[
https://issues.apache.org/jira/browse/ZOOKEEPER-2236?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Rakesh R updated ZOOKEEPER-2236:
--------------------------------
Fix Version/s: (was: 3.4.6)
> Zookeeper truncates file to 0bytes
> ----------------------------------
>
> Key: ZOOKEEPER-2236
> URL: https://issues.apache.org/jira/browse/ZOOKEEPER-2236
> Project: ZooKeeper
> Issue Type: Bug
> Components: contrib-zkfuse, server, tests
> Affects Versions: 3.4.6
> Environment: Ubuntu14.04, Standalone Zookeeper server, Zkfuse
> Reporter: Bharat Singh
>
> I am facing a rename issue with Zkfuse.
> I am trying to test file atomic updates. After some iterations the file size
> becomes 0.
> This is easily reproducible, just running the below script for ~5mins.
> Setup:
> zookeeper-3.4.6 with Zkfuse mounted, size of testFile = 1k
> while [ 1 ]
> do
> cp /root/testFile /mnt/zk/testFile.tmp
> mv /mnt/zk/testFile.tmp /mnt/zk/testFile
> ls -larth /mnt/zk/
> sleep 1
> done
> Zkfuse debug logs doesn't show any suspicious activity. Looks like
> zookeeper/zkfuse RENAME is not atomic.
> But code browsing and log messages show that update have issues:
> 1) update is not able to pull data from zookeeper due to the _refCnt > 1,
> so rename get an empty ZkfuseFile object.
> 2) I also hit an assert in update,
> assert(newFile == false || _isOnlyRegOpen());
> Now I have suspicion on the refcount logic. Have any one faced similar issues
> or have used Zkfuse in production environment.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
