Configure cluster over TCP

[klinux]

Hi folks,

I whant to start JBoss with -Djboss.default.jgroups.stack=tcp, to use TCP stack instead UDP, but I get this in log:

2011-03-31 14:11:38,274 WARN  [org.jgroups.protocols.MPING] (ReceiverThread) discarding discovery request for cluster 'A-HAPartition' from XXXX:1099; our cluster name is 'A-Web'. Please separate your clusters cleanly.

How can I get fix this?

Thanks

[wdfink]

Are you sure that the partition-name is equal?

I suppose your first node use 'A-HAPartition' and the other 'A-Web' as partition-name.

[ekfliu]

default tcp stack still uses udp mult cast for discovery, it just uses tcp for data transfering.

You must switch to static host discovery. To do that find the file jgroups-channelfactory-stacks.xml under cluster/jgroups-channelfactory.sar/meta-inf.

Goto the tcp and the tcp-sync sections, you find a comment <!-- Alternative 1: multicast-based automatic discovery. comment out the configuration right underneth it. Find another comment <!-- Alternative 2: non multicast-based replacement for MPING. Requires a static configuration, uncomment out the configuration underneth that.

you will need to modifiy your startup script to point to all of your cluster servers with -Djboss.jgroups.tcp.tcpping.initial_hosts=server1[7650],server2[7650]

That should fix the problem.

[fivalo]

Hi Everybody

What happen if I ignore de message intentionally. Because I have more than one cluster on my network.

Do you know if ignoring that message could influence the performance? causing a network overhead? or something like that?

Thank you