[
https://issues.apache.org/jira/browse/SOLR-6261?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14089078#comment-14089078
]
Ramkumar Aiyengar edited comment on SOLR-6261 at 8/7/14 9:43 AM:
-----------------------------------------------------------------
This is possibly exposing a design issue with OCP. There is a watch being
created with each {{workQueue.peekTopN}}, even if there are work items present.
It could then fail exclusivity check, and then try again, find items and create
one more watch..
{code}
~/builds/lucene-solr/solr/core| less out.log | grep 'Exclusivity check failed'
| wc -l
6960
{code}
Obviously with the new code you run into issues because that creates 1000s of
threads, but it's a problem to begin with that the code is creating 1000s of
watches..
was (Author: andyetitmoves):
This is possibly exposing a design issue with OCP. There is a watch being
created with each {{workQueue.peekTopN}}, even if there are work items present.
It could then fail exclusivity check, and then try again, find items and create
one more watch..
{code}
~/builds/lucene-solr/solr/core| less out.log | grep 'Exclusivity check failed'
| wc -l
6960
{code}
> Run ZK watch event callbacks in parallel to the event thread
> ------------------------------------------------------------
>
> Key: SOLR-6261
> URL: https://issues.apache.org/jira/browse/SOLR-6261
> Project: Solr
> Issue Type: Improvement
> Components: SolrCloud
> Affects Versions: 4.9
> Reporter: Ramkumar Aiyengar
> Assignee: Mark Miller
> Priority: Minor
> Fix For: 5.0, 4.10
>
>
> Currently checking for leadership (due to the leader's ephemeral node going
> away) happens in ZK's event thread. If there are many cores and all of them
> are due leadership, then they would have to serially go through the two-way
> sync and leadership takeover.
> For tens of cores, this could mean 30-40s without leadership before the last
> in the list even gets to start the leadership process. If the leadership
> process happens in a separate thread, then the cores could all take over in
> parallel.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
