Multicast not binding to bind_addr

Hi,

 

I have a problem with clustering multicast traffic being generated on the wrong network interface. I am running Jboss-4.0.4GA, JDK 1.5, JGroups-2.3SP1 on a RedHat ES 4 x86 platform. I am running 2 multi-homed servers in a cluster and I am trying to get all clustering working on a seperate network interface. I am only clustering http session information using the /deploy/tc5-cluster.sar/META-INF/jboss-service.xml config file.

 

I start the server with this command line: /bin/run.sh -c all -b 10.10.12.80 -Dignore.bind.address=true -g prodblue -u 230.1.2.8

 

I am using this JGroups config: (configured in the /deploy/tc5-cluster.sar/META-INF/jboss-service.xml file)

 

<config>
                <UDP bind_addr="10.10.13.80"
                     mcast_addr="${jboss.partition.udpGroup:230.1.2.8}"
                     mcast_port="45578"  

So all Jboss services bind to 10.10.12.80 but Jgroups should bind to the 10.10.13.80 interface.

 

When I start up the server I see this in server.log:

 

2006-09-14 17:08:02,391 DEBUG [org.jboss.deployment.SARDeployer] Deploying SAR, start step: url file:/usr/jboss-4.0.4.GA/server/all/deploy/tc5-cluster.sar/

2006-09-14 17:08:02,391 DEBUG [org.jboss.system.ServiceController] starting service jboss.cache:service=TomcatClusteringCache

2006-09-14 17:08:02,391 DEBUG [org.jboss.cache.aop.TreeCacheAop] Starting jboss.cache:service=TomcatClusteringCache

2006-09-14 17:08:02,405 DEBUG [org.jgroups.protocols.UDP] creating sockets and starting threads

2006-09-14 17:08:02,406 INFO  [org.jgroups.protocols.UDP] sockets will use interface 10.10.13.80

2006-09-14 17:08:02,411 INFO  [org.jgroups.protocols.UDP] socket information:

local_addr=10.10.13.80:44096, mcast_addr=230.1.2.8:45578, bind_addr=/10.10.13.80, ttl=2

sock: bound to 10.10.13.80:44096, receive buffer size=131071, send buffer size=131071

mcast_recv_sock: bound to 10.10.13.80:45578, send buffer size=131071, receive buffer size=131071

mcast_send_sock: bound to 10.10.13.80:44097, send buffer size=131071, receive buffer size=131071

2006-09-14 17:08:02,428 INFO  [STDOUT]

-------------------------------------------------------

GMS: address is 10.10.13.80:44096

-------------------------------------------------------

2006-09-14 17:08:02,432 DEBUG [org.jgroups.protocols.UDP] created unicast receiver thread

2006-09-14 17:08:04,438 DEBUG [org.jgroups.protocols.pbcast.GMS] initial_mbrs are []

2006-09-14 17:08:04,438 DEBUG [org.jgroups.protocols.pbcast.GMS] no initial members discovered: creating group as first member

2006-09-14 17:08:04,450 DEBUG [org.jgroups.protocols.pbcast.GMS] [local_addr=10.10.13.80:44096] view is [10.10.13.80:44096|0] [10.10.13.80:44096]

2006-09-14 17:08:04,454 DEBUG [org.jgroups.protocols.FD] suspected_mbrs: [], after adjustment: []

2006-09-14 17:08:04,458 INFO  [org.jboss.cache.TreeCache] viewAccepted(): [10.10.13.80:44096|0] [10.10.13.80:44096]

2006-09-14 17:08:04,473 DEBUG [org.jgroups.protocols.pbcast.GMS] 10.10.13.80:44096 changed role to org.jgroups.protocols.pbcast.CoordGmsImpl

 

The Jgroups protocols including multicast send and recv sock appear to be binding to the 10.10.13.80 interface which is correct.

 

However once the 2 servers are up and running I see all multicast traffic running on the first 10.10.12.80 interface, this is an ouput of tcpdump:

 

17:14:48.915801 IP 10.10.12.80.44097 > 230.1.2.8.45578: UDP, length 139

 

This is an output of netstat -anup:

 

udp        0      0 ::ffff:10.10.12.80:1161     :::*                                    11578/java
udp        0      0 :::45578                    :::*                                    11578/java                 (mcast_recv_sock - doesn't seem to be bound correctly?)
udp        0      0 ::ffff:10.10.12.80:1162     :::*                                    11578/java
udp        0      0 ::ffff:10.10.12.80:44094    :::*                                    11578/java
udp        0      0 ::ffff:10.10.13.80:44096    :::*                                    11578/java            (This is GMS)
udp        0      0 :::44097                    :::*                                    11578/java                 (mcast_send_sock - doesn't seem to be bound correctly?)
udp        0      0 :::7500                     :::*                                    11578/java  

This is an output of netstat -antp:

 

tcp        0      0 ::ffff:10.10.12.80:3528     :::*                        LISTEN      11578/java
tcp        0      0 ::ffff:10.10.12.80:1098     :::*                        LISTEN      11578/java
tcp        0      0 ::ffff:10.10.12.80:1099     :::*                        LISTEN      11578/java
tcp        0      0 ::ffff:10.10.12.80:8080     :::*                        LISTEN      11578/java
tcp        0      0 ::ffff:10.10.12.80:8083     :::*                        LISTEN      11578/java
tcp        0      0 ::ffff:10.10.12.80:8443     :::*                        LISTEN      11578/java
tcp        0      0 ::ffff:10.10.12.80:4444     :::*                        LISTEN      11578/java
tcp        0      0 ::ffff:10.10.12.80:4445     :::*                        LISTEN      11578/java 

 

Also they never see each other in the cluster, they both startup as Co-ordinators and never join. If I remove the -Dignore.bind.address from the startup commands and restart the 2 servers they will then see each other in the cluster (all working over the 10.10.12.80 interface) and everything works fine. I really need to get clustering working over the seperate interface as the amount of traffic generated by the http session replication is quite alot.

 

Any ideas??

 

Regards

 

 

James Wilson
SYSTEMS ENGINEER

I Want One Of Those Ltd
stuff you don't need... but you really, really want
www.iwantoneofthose.com

 

Some stuff you should know (otherwise known as a disclaimer):

The information in this message is confidential and may be legally privileged. It is intended solely for the addressee. Access to this message by anyone else is unauthorised.  If you are not the intended recipient, any disclosure, copying, printing or distribution of the message, or any action or omission taken by you in reliance on it, is prohibited and may be unlawful.  Any opinions expressed in this e-mail, are those of the individual, and do not necessarily reflect those of the company. If you have received this message in error, please delete the message and any attachment(s) from your mailbox and contact I Want One of Those Ltd on  customersupport@xxxxxxxxxxxxxxxxxxx.  Thank you.

 

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642_______________________________________________
javagroups-users mailing list
javagroups-users@xxxxxxxxxxxxxxxxxxxxx
https://lists.sourceforge.net/lists/listinfo/javagroups-users