[jboss-jira] [JBoss JIRA] Closed: (JBAS-5630) Web session persistent store cleanup with total repl and no marshalling

[Brian Stansberry (JIRA)]

     [ https://jira.jboss.org/jira/browse/JBAS-5630?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Brian Stansberry closed JBAS-5630.
----------------------------------

    Resolution: Done


I decided to add purgeStateOnStartup=true to the JBC cache loader configs for session caches.  With this, a complete restart of a cluster will not allow any state to survive.

It is valid for users who wish state to survive to remove this config/set it to false. However, if this is done, care should be taken such that the last cluster node to shutdown is the first to restart; otherwise the first node to restart will load stale data from its persistent store and propagate it to the other nodes when they join (overwriting whatever they had stored).  The potential for error in this is the reason I did not make this the default behavior.

> Web session persistent store cleanup with total repl and no marshalling
> -----------------------------------------------------------------------
>
>                 Key: JBAS-5630
>                 URL: https://jira.jboss.org/jira/browse/JBAS-5630
>             Project: JBoss Application Server
>          Issue Type: Sub-task
>      Security Level: Public(Everyone can see) 
>          Components: Clustering, Web (Tomcat) service
>            Reporter: Brian Stansberry
>            Assignee: Brian Stansberry
>             Fix For: JBossAS-5.0.0.CR2
>
>
> See parent issue for basic problem description. This subtask is for when total replication is used and region based marshalling is not configured. 
> Need to think about this one a bit more.  Here the cache might remain active after webapp undeploy, and will remain consistent with the cluster.  When the webapp starts, it doesn't do a state transfer (cache already did a full transfer when *it* started), so it can't remove local state and then transfer.
> Need to determine if we want to support a concept of letting a cache that's the first one to start (e.g. after a full cluster restart) go ahead and read persistent state.  Or, do we simply set the cache's cache loader's "purgeOnStartup" to true and flush persistent state.  With the latter, the persistent store is purely used for passivation overflow and not for HA; HA is provided by the cluster.  But no state survives a complete cluster restart.

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: https://jira.jboss.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira