[ 
https://issues.apache.org/jira/browse/SOLR-6621?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14173785#comment-14173785
 ] 

Renaud Delbru commented on SOLR-6621:
-------------------------------------

I have added some comments, and created a pull request at:
https://github.com/apache/lucene-solr/pull/100

> SolrZkClient does not guarantee that a watch object will only be triggered 
> once for a given notification
> --------------------------------------------------------------------------------------------------------
>
>                 Key: SOLR-6621
>                 URL: https://issues.apache.org/jira/browse/SOLR-6621
>             Project: Solr
>          Issue Type: Bug
>          Components: SolrCloud
>    Affects Versions: Trunk
>            Reporter: Renaud Delbru
>         Attachments: SOLR-6621
>
>
> The SolrZkClient provides methods such as getData or exists. The problem is 
> that the client automatically wraps the provided watcher with a new watcher 
> (see 
> [here|https://github.com/apache/lucene-solr/blob/6ead83a6fafbdd6c444e2a837b09eccf34a255ef/solr/solrj/src/java/org/apache/solr/common/cloud/SolrZkClient.java#L255])
>  which breaks the guarantee that "a watch object, or function/context pair, 
> will only be triggered once for a given notification". This creates 
> undesirable effects when we are registering the same watch is the Watcher 
> callback method.
> A possible solution would be to introduce a SolrZkWatcher class, that will 
> take care of submitting the job to the zkCallbackExecutor. Components in 
> SolrCloud will extend this class and implement their own callback method. 
> This will ensure that the watcher object that zookeeper receives remains the 
> same.
> See SOLR-6462 for background information.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org

Reply via email to