[
https://issues.apache.org/jira/browse/SOLR-9784?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16332729#comment-16332729
]
David Smiley commented on SOLR-9784:
------------------------------------
I'm writing a test wherein I call
{{solrClient.getClusterStateProvider().getState(collectionName)}} and in
certain cases it can throw an NPE due to no zkStateReader. Apparently it
wasn't connect()'ed first. I think this is a design flaw – no connect() method
ought to be needed IMO. It needs to connect, sure, but can't that detail be
handled completely internally to the provider imol?
ZkClientClusterStateProvider.getZkStateReader could become what connect() is
right now, and then all code in ZkClientClusterStateProvider that needs the
zkStateReader should call that method and not access the field directly.
> Refactor CloudSolrClient to eliminate direct dependency on ZK
> -------------------------------------------------------------
>
> Key: SOLR-9784
> URL: https://issues.apache.org/jira/browse/SOLR-9784
> Project: Solr
> Issue Type: Sub-task
> Components: SolrJ
> Reporter: Noble Paul
> Assignee: Noble Paul
> Priority: Major
> Fix For: 6.4, 7.0
>
> Attachments: SOLR-9584.patch
>
>
> CloudSolrClient should decouple itself from the ZK reading/write. This will
> help us provide alternate implementations w/o direct ZK dependency
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
