[
https://issues.apache.org/jira/browse/SOLR-5749?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13931919#comment-13931919
]
Shalin Shekhar Mangar commented on SOLR-5749:
---------------------------------------------
Thanks Mark.
[~tim.potter] - I didn't use the metrics APIs (that's a big issue!) but you'll
find that all of your demands are met by this patch.
I think we should rename "stats" to "operations" and have the timing done per
minute instead of per-second since Overseer operations are not that frequent. I
am working on capturing the past N operations and past N failures (exceptions)
per operation to the stats. Right now the stats are in-memory which means that
we lose them if the overseer dies. I think that we should periodically, say
every 15 minutes, save the stats to ZK and initialize the stats from ZK when a
new Overseer starts.
> Implement an Overseer status API
> --------------------------------
>
> Key: SOLR-5749
> URL: https://issues.apache.org/jira/browse/SOLR-5749
> Project: Solr
> Issue Type: Sub-task
> Components: SolrCloud
> Reporter: Shalin Shekhar Mangar
> Fix For: 5.0
>
> Attachments: SOLR-5749.patch
>
>
> Right now there is little to no information exposed about the overseer from
> SolrCloud.
> I propose that we have an API for overseer status which can return:
> # Past N commands executed (grouped by command type)
> # Status (queue-size, current overseer leader node)
> # Overseer log
--
This message was sent by Atlassian JIRA
(v6.2#6252)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
