[
https://issues.apache.org/jira/browse/SOLR-7636?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14578332#comment-14578332
]
Shalin Shekhar Mangar commented on SOLR-7636:
---------------------------------------------
{quote}
Before this fix, the call path was as follows
colectionshandler -> ZK Q -> OCP -> (ZK Q read + ZK Q DEL + ZK Q Write) ->
(CollectionsHandler ZK READ+ DEL)
Now it has become
CollectionsHandler -> ZK READ
So, this is definitely a win as of now.
{quote}
That is not true. Because we have to forcefully update the cluster state, we
have CollectionsHandler -> ZK getChildren (/live_nodes) + ZK read
(/clusterstate.json) + ZK getChildren + N * ZK exists (for getIndividualColls)
+ W * ZK exists + W * getData (for watched collections) + (N-W) * ZK exists +
(N-W) * ZK getData (for non-watched collections).
> CLUSTERSTATUS Api should not go to OCP to fetch data
> ----------------------------------------------------
>
> Key: SOLR-7636
> URL: https://issues.apache.org/jira/browse/SOLR-7636
> Project: Solr
> Issue Type: Improvement
> Reporter: Noble Paul
> Assignee: Noble Paul
> Priority: Minor
> Fix For: Trunk, 5.3
>
> Attachments: SOLR-7636.patch
>
>
> Currently it does multiple ZK operations which is not required. It should
> just read the status from ZK and return from the CollectionsHandler
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
