[
https://issues.apache.org/jira/browse/SOLR-7503?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Timothy Potter updated SOLR-7503:
---------------------------------
Attachment: SOLR-7503.patch
Simple patch that registers cores in the background after ZK session
expiration. I had to add some getter methods for the ExecutionService in the
ZkContainer so that it is available to the ZkController when needed (iff cc is
not null). I didn't want to use a new ExecutionService since the one setup by
ZkContainer seemed most appropriate for this work, but you can't expose
ZkContainer directly in ZkController because it's only a server-side thing.
> Recovery after ZK session expiration happens in a single thread for all cores
> in a node
> ---------------------------------------------------------------------------------------
>
> Key: SOLR-7503
> URL: https://issues.apache.org/jira/browse/SOLR-7503
> Project: Solr
> Issue Type: Bug
> Components: SolrCloud
> Affects Versions: 5.1
> Reporter: Shalin Shekhar Mangar
> Assignee: Timothy Potter
> Labels: impact-high
> Fix For: Trunk, 5.2
>
> Attachments: SOLR-7503.patch
>
>
> Currently cores are registered in parallel in an executor. However, when
> there's a ZK expiration, the recovery, which also happens in the register
> call, happens in a single thread:
> https://github.com/apache/lucene-solr/blob/trunk/solr/core/src/java/org/apache/solr/cloud/ZkController.java#L300
> We should make these happen in parallel as well so that recovery after ZK
> expiration doesn't take forever.
> Thanks to [~mewmewball] for catching this.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
