[
https://issues.apache.org/jira/browse/HDFS-5014?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13749948#comment-13749948
]
Vinay commented on HDFS-5014:
-----------------------------
Updated patch to solve the concurrency issues, illustrated by chris. Please
review.
In both examples, adding a double check in case of state changes would solve
the issue.
Removed the lock for processing standby commands to avoid the following case in
prev patch.
1. Standby NN was unstable and asked DN to register.
2. DN acquired the readlock and started registering, but by this time Standby
NN went down. So DN will keep trying in this point with readlock held.
3. Now any state changes from ANN, will block in getting writeLock().
I feel lock not required while processing commands from standby as no data
modification commands will be processed here. Please correct me if I am wrong.
:)
> BPOfferService#processCommandFromActor() synchronization on namenode RPC call
> delays IBR to Active NN, if Stanby NN is unstable
> -------------------------------------------------------------------------------------------------------------------------------
>
> Key: HDFS-5014
> URL: https://issues.apache.org/jira/browse/HDFS-5014
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode, ha
> Affects Versions: 3.0.0, 2.0.4-alpha
> Reporter: Vinay
> Assignee: Vinay
> Attachments: HDFS-5014.patch, HDFS-5014.patch, HDFS-5014.patch,
> HDFS-5014.patch, HDFS-5014.patch
>
>
> In one of our cluster, following has happened which failed HDFS write.
> 1. Standby NN was unstable and continously restarting due to some errors. But
> Active NN was stable.
> 2. MR Job was writing files.
> 3. At some point SNN went down again while datanode processing the REGISTER
> command for SNN.
> 4. Datanodes started retrying to connect to SNN to register at the following
> code in BPServiceActor#retrieveNamespaceInfo() which will be called under
> synchronization.
> {code} try {
> nsInfo = bpNamenode.versionRequest();
> LOG.debug(this + " received versionRequest response: " + nsInfo);
> break;{code}
> Unfortunately in all datanodes at same point this happened.
> 5. For next 7-8 min standby was down, and no blocks were reported to active
> NN at this point and writes have failed.
> So culprit is {{BPOfferService#processCommandFromActor()}} is completely
> synchronized which is not required.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
