[
https://issues.apache.org/jira/browse/QPID-6154?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Keith Wall updated QPID-6154:
-----------------------------
Summary: Handle rollback of node when use of weak durability has allowed
nodes to diverge (was: Handle rollback of node when use of weak durability has
allow nodes to diverge)
> Handle rollback of node when use of weak durability has allowed nodes to
> diverge
> --------------------------------------------------------------------------------
>
> Key: QPID-6154
> URL: https://issues.apache.org/jira/browse/QPID-6154
> Project: Qpid
> Issue Type: Improvement
> Components: Java Broker
> Reporter: Keith Wall
> Labels: bdbstore, ha
>
> JE throws an asynchronous exception (RolllbackException) to indicate that,
> owing to the use of weak durability options a node that is trying to rejoin
> the group has transactions that do not match those of the current master, and
> it must rollback one or more transactions in order to join.
> This condition can't occur in use of the group with default quorum options,
> but if a user were to use node priority zero, electable group override or
> designate primary (options that all weaken durability), creates the situation
> were nodes can diverge (split brain).
> In the case where JE detects this, it delivers a RollbackException
> asynchronously shortly after environment creation and requires that the
> environment restarts so the diverged transactions can be dropped.
> Java Broker should catch this exception, log it and organise for the
> environment to restart.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
