[
https://issues.apache.org/jira/browse/SOLR-918?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12657885#action_12657885
]
Yonik Seeley commented on SOLR-918:
-----------------------------------
bq. Is it ok to let it be a ConcurrentHashMap or should we make it a
synchronized LinkedHashMap
Either should be fine... new components are added so infrequently and checked
so infrequently that there won't be a performance difference.
If it's changed to synchronized LinkedHashMap, it should be noted in the
javadoc for SolrCore.getInfoRegistry() (which is new in 1.3) that it's a
synchronized Map and thus users need to synchronize during iteration. This
synchronization is currently being done by all users of this API since it
previously was a synchronized map (before the synchronization was mistakenly
removed).
If we stay with ConcurrentHashMap (and hence don't care about order), then the
current synchronization while iterating should probably be removed too.
> Thread safety issue in SolrCore.infoRegistry
> --------------------------------------------
>
> Key: SOLR-918
> URL: https://issues.apache.org/jira/browse/SOLR-918
> Project: Solr
> Issue Type: Bug
> Reporter: Shalin Shekhar Mangar
> Assignee: Shalin Shekhar Mangar
> Priority: Minor
> Fix For: 1.4
>
> Attachments: SOLR-918.patch
>
>
> SolrCore.infoRegistry is used for keeping a reference to the SolrInfoMBean
> objects. When JMX is disabled, it is initialized to a LinkedHashMap which is
> not synchronized.
> We can change this to a ConcurrentHashMap instead of a LinkedHashMap.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
