[jboss-jira] [JBoss JIRA] (JGRP-1654) Logical names: a view change in one cluster marks for removal the entries of all the other clusters

[Bela Ban (JIRA)]

    [ https://issues.jboss.org/browse/JGRP-1654?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12789869#comment-12789869 ] 

Bela Ban commented on JGRP-1654:
--------------------------------

It seems a very simple change can fix this: instead of removing all members not in the current view, we simply remove the members which left. E.g. if we have view A,B,C,D and then get A,B,C, we could simply remove D from the UUID cache. 
This would allow us to continue using UUID as a singleton object.
                
> Logical names: a view change in one cluster marks for removal the entries of all the other clusters
> ---------------------------------------------------------------------------------------------------
>
>                 Key: JGRP-1654
>                 URL: https://issues.jboss.org/browse/JGRP-1654
>             Project: JGroups
>          Issue Type: Bug
>    Affects Versions: 3.3.3
>            Reporter: Dan Berindei
>            Assignee: Bela Ban
>            Priority: Minor
>             Fix For: 3.4
>
>
> This is related to JGRP-918.
> The UUID class keeps a static cache of logical names, which is shared by all the clusters running in the same JVM. When TP receives a VIEW_CHANGED event, it calls {{UUID.retainAll(members)}}, which marks all the addresses on other clusters for removal. If there are enough members to exceed the logical names cache capacity (500 by default), some of those members will be actually removed from the logical names cache and their addresses will be printed as UUIDs.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira