[
https://issues.apache.org/jira/browse/SOLR-17381?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17904224#comment-17904224
]
ASF subversion and git services commented on SOLR-17381:
--------------------------------------------------------
Commit a9e7506c618e392d357dd5cb636a0c8772e79fc5 in solr's branch
refs/heads/branch_9x from aparnasuresh85
[ https://gitbox.apache.org/repos/asf?p=solr.git;h=a9e7506c618 ]
SOLR-17381: Fix CloudSolrClient.getClusterState when HTTP CSP (not ZK) (#2853)
When using the HTTP ClusterStateProvider (not ZK), getClusterState() wasn't
working correctly; a regression from the first PR for this JIRA issue (not
released).
Also,
* Optimization: fix O(N^2) algorithm to be O(N) for the number of collections
when calling getClusterState
* liveNodes is now immutable, and probably a non-sorted ordering
* removed "health" and some other keys from DocCollection that aren't present
when using ZK CSP
Minor details:
* clearly differentiate internal ClusterState processing from getting one
DocCollection
* Use GenericSolrRequest not QueryRequest
* test the both CSPs better
---------
Co-authored-by: David Smiley <[email protected]>
(cherry picked from commit d2045f679e55e15cf4e974947b9417c4465fccd2)
> Make CLUSTERSTATUS request configurable
> ---------------------------------------
>
> Key: SOLR-17381
> URL: https://issues.apache.org/jira/browse/SOLR-17381
> Project: Solr
> Issue Type: Improvement
> Reporter: Aparna Suresh
> Priority: Minor
> Labels: pull-request-available
> Fix For: 9.8
>
> Time Spent: 9h 40m
> Remaining Estimate: 0h
>
> Fetching {{CLUSTERSTATUS}} remotely is resource-intensive and should be done
> with caution. Currently, if no parameters are specified, the call returns all
> information, including collections, shards, replicas, aliases, cluster
> properties, roles, and more. This can have significant performance
> implications for clients using a Solr cluster with thousands of collections.
> Several performance [issues|https://issues.apache.org/jira/browse/SOLR-14985]
> have been identified when switching {{CloudSolrClient}} to use HTTP-based
> CSP, particularly in two instances where the entire cluster state is fetched
> unnecessarily.
> *Proposal:* Modify the requests to retrieve only the necessary information,
> such as the cluster status for a specific collection, live nodes, or cluster
> properties. Ensure these changes maintain backward compatibility.
> Additionally, update the HTTP CSP to reflect these optimizations.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
