[jboss-jira] [JBoss JIRA] (JGRP-1461) SEQUENCER leaks sequence numbers
Bela Ban (JIRA)
jira-events at lists.jboss.org
Thu May 10 06:19:18 EDT 2012
[ https://issues.jboss.org/browse/JGRP-1461?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12691754#comment-12691754 ]
Bela Ban commented on JGRP-1461:
--------------------------------
Ok, at this point, I think the seqno assigned to a forward-message by C should *not* be used to order it; instead it should be used as unique timestamp. When we forward a message 5, and later receive 5, we can remove it from the forward-table. When a coordinator crashes, we can resend the messages left in the forward-table.
This does *not* require the forward-table to be sorted, so we can downgrade forward-table from a TreeSet to a simple ArrayList.
Again, total order is established by the coordinator (A) broadcasting messages on behalf of all cluster members. The messages received from A will be delivered on the same thread (as it is the same sender), so we don't have to order them again.
> SEQUENCER leaks sequence numbers
> --------------------------------
>
> Key: JGRP-1461
> URL: https://issues.jboss.org/browse/JGRP-1461
> Project: JGroups
> Issue Type: Bug
> Affects Versions: 3.0.9, 3.1
> Reporter: David Hotham
> Assignee: Bela Ban
> Fix For: 3.0.10, 3.1
>
>
> As noted in JGRP-1458, it looks as though SEQUENCER always thinks that sequence numbers must start at zero. But a member who joins a group after some messages have already been broadcast may never receive sequence number zero. So, I think, his SeqnoTable will grow indefinitely.
> Note that this one's mostly through code reading. I think I'm right; but I'm happy to be told otherwise!
> Pull request to follow shortly.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.jboss.org/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
More information about the jboss-jira
mailing list