[jboss-jira] [JBoss JIRA] (JGRP-1519) RELAY2: relaying of multicasts between multiple clusters leads to cycles
Bela Ban (JIRA)
jira-events at lists.jboss.org
Mon Sep 24 06:43:34 EDT 2012
[ https://issues.jboss.org/browse/JGRP-1519?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12720930#comment-12720930 ]
Bela Ban commented on JGRP-1519:
--------------------------------
OK, the issue is the following:
- Sites LON={A,B,C}, SFO={M,N,O} and NYC={X,Y,Z}. Site masters are A, M and X
- A receives a multicast message originating from its local cluster
- A forwards the message to M and X (as unicasts)
- M and X broadcast the message locally, *but also forwards it to all other sites except itself and the sender*, so:
- M forwards the message to X (NYC), and
- X forwards the message to M (SFO)
In the next step,
- X forwards the message to A, and
- M forwards the message to A, too
and so on, this cycle never ends until the number of sites goes down to 2 !
The root cause is that both the site master of the originating node and all remote site masters forward a multicast message to *all* routes (except self and sender).
SOLUTION:
- Only the site master of the originating site forwards the message to all remote site masters; the remote site master now only broadcast the message locally, but don't forward it.
The drawback here is that this only works for sites that are all connected to the same global bridge, also this won't work for hierarchical routing.
We have to revisit this issue when tackling hierarchical routing.
> RELAY2: relaying of multicasts between multiple clusters leads to cycles
> ------------------------------------------------------------------------
>
> Key: JGRP-1519
> URL: https://issues.jboss.org/browse/JGRP-1519
> Project: JGroups
> Issue Type: Bug
> Reporter: Bela Ban
> Assignee: Bela Ban
> Fix For: 3.2
>
> Attachments: global.xml, LON.xml, NYC.xml, relay2.xml, SFO.xml
>
>
> - Run 2 Draw instances for LON.xml, SFO.xml and NYC.xml (RELAY2.relay_multicasts needs to be enabled)
> - Draw in any instance: this leads to the multicast being relayed forever
> - Note that these cycles don't occur with 2 clusters
--
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
More information about the jboss-jira
mailing list