[
https://issues.apache.org/jira/browse/SOLR-9106?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15381851#comment-15381851
]
Alan Woodward commented on SOLR-9106:
-------------------------------------
bq. Not watching clusterprops was done deliberately. it is read rarely and
watching is expensive.
This seems exactly backwards to me? Watches are pretty lightweight. Do you
have some numbers here?
If it's true, then it needs to be explicitly documented on the
.getClusterProperties() method that it will always read data from ZK - maybe it
should be called .readClusterProperties() instead. And it ends up being used
in some expected places, for example during admin requests from a
CloudSolrClient.
> Cache cluster properties in ZkStateReader
> -----------------------------------------
>
> Key: SOLR-9106
> URL: https://issues.apache.org/jira/browse/SOLR-9106
> Project: Solr
> Issue Type: Improvement
> Affects Versions: master (7.0)
> Reporter: Alan Woodward
> Assignee: Alan Woodward
> Fix For: 6.1
>
> Attachments: SOLR-9106.patch, SOLR-9106.patch
>
>
> ZkStateReader currently makes calls into ZK every time getClusterProps() is
> called. Instead we should keep the data locally and use a Watcher to update
> it.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
