[jboss-jira] [JBoss JIRA] (WFLY-13639) Avoid duplicate expiration scheduling on member leave for non-tx invalidation cache

[Paul Ferraro (Jira)]

     [ https://issues.redhat.com/browse/WFLY-13639?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Paul Ferraro updated WFLY-13639:
--------------------------------
    Description: Following WFLY-13627, a non-tx invalidation cache will schedule session/SFSB expiration on the local member.  When that member leaves, the expiration will be reschedule on the primary owner of segment 0 (in a non-tx cache, all entries map to segment 0).  However, since that member cannot distinguish which sessions/SFSBs were owned by the leaving member, it must schedule all sessions/SFSBs.  (was: Following WFLY-13627, a non-tx invalidation cache will schedule session/SFSB expiration on the local member.  When that member leaves, the expiration will be reschedule on the primary owner of segment 0 (in a non-tx cache, all entries map to segment 0).  However, since that member cannot distinguish which sessions/SFSBs were owned by the leaving member, it must schedule all sessions/SFSBs.

By storing the address of the member that handled a given request along with a web session/SFSBs metadata, we can discriminate which entries require expiration rescheduling on topology change for non-tx invalidation caches.)


> Avoid duplicate expiration scheduling on member leave for non-tx invalidation cache
> -----------------------------------------------------------------------------------
>
>                 Key: WFLY-13639
>                 URL: https://issues.redhat.com/browse/WFLY-13639
>             Project: WildFly
>          Issue Type: Bug
>          Components: Clustering
>    Affects Versions: 20.0.0.Final
>            Reporter: Paul Ferraro
>            Assignee: Paul Ferraro
>            Priority: Major
>
> Following WFLY-13627, a non-tx invalidation cache will schedule session/SFSB expiration on the local member.  When that member leaves, the expiration will be reschedule on the primary owner of segment 0 (in a non-tx cache, all entries map to segment 0).  However, since that member cannot distinguish which sessions/SFSBs were owned by the leaving member, it must schedule all sessions/SFSBs.



--
This message was sent by Atlassian Jira
(v7.13.8#713008)