[jboss-jira] [JBoss JIRA] (JGRP-1656) RELAY2: Requester ignores remote sites members responses

[Bela Ban (JIRA)]

    [ https://issues.jboss.org/browse/JGRP-1656?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12789956#comment-12789956 ] 

Bela Ban commented on JGRP-1656:
--------------------------------

This is due to a misunderstanding of how RELAY2 works: LON and NYC are completely autonomous clusters and don't know each other. When you invoke a cluster wide RPC in LON, the null destination means that we'll block util we've received responses from A and B. 

While RELAY2 does forward the multicast to NYC where it will be invoked on C and D, which in turn send responses to the caller in LON, the latter won't wait for the responses from C and D as it only blocks on A and B.
The reason is that 'null' in LON means the current membership, which is A and B. The LON cluster doesn't know anything about the NYC cluster or their members.

So you're using the wrong tool; *RELAY* instead of RELAY2 would work the way you want, but RELAY is not actively maintained.

                
> RELAY2: Requester ignores remote sites members responses
> --------------------------------------------------------
>
>                 Key: JGRP-1656
>                 URL: https://issues.jboss.org/browse/JGRP-1656
>             Project: JGroups
>          Issue Type: Bug
>    Affects Versions: 3.3.3
>            Reporter: Karim AMMOUS
>            Assignee: Bela Ban
>             Fix For: 3.4
>
>
> Let's:
> - A and B two members from site LON
> - C and D two members from site NYC
> - B and C are site masters
> A sends an RPC request (RpcDispacther.callRemoteMethods) to all members (dest=null) and blocks (RspMode.GET_ALL) until receiving all responses. We expect 4 responses, but only 2 responses are recovered into RspList.
> According to tests with low level traces, remote site responses have been sent back, relayed by site masters and reached the requester, but response collector at requester side (GroupRequest) ignored them.

--
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