Hello Arjen,

One vital piece of information is missing in your post. Which version of MMBase are yoy using. The multicast configuration changed during the 1.7.* releases.

In 1.7.1 the settings were in the config/modules/mmbaseroot.xml file. Based on the info of your post I ttink you found the properties in this file.

1.7.2 and up moved the properties to config/utils/multicast.xml and the mmbaseroot.xml has a property sharedstorage. In 1.8 this is renamed to clustering.

in 1.7.1 and previous release the logging of multicast was all on debug level. It didn't even mention that it was starting. 1.7.2 and up echos the settings from the config in the logfile on startup.

When you use 1.7.1 add a logger for class org.mmbase.module.core.MMBaseMultiCast with debug in the log4j.xml and you should see some messages passing around.

Another thing nice to know is: are the machines on the same wire or are there any routers/switches between them. The implementation uses a ttl of 1 which means that the packets won't leave the wire.

Nico Klasens


Arjen Roodselaar wrote:
Hi all,

I'm trying to setup MMbase to run in a cluster environment. So far I failed however and I'm running out of options to try. Maybe someone reading this will be able to point me in the right direction.

The relevant part of my MMbase configuration:

node 1:
<property name="machinename">ou.work2</property>
<property name="multicasthost">226.0.0.2</property>
<property name="multicastport">9119</property>
<property name="host">work2.esp</property>

node 2:
<property name="machinename">ou.work3</property>
<property name="multicasthost">226.0.0.2</property>
<property name="multicastport">9119</property>
<property name="host">work3.esp</property>

To avoid confusion, work2.esp and work3.esp do resolve to a valid IP address (we're using a private tld's inside our cluster)

The weird thing is that I don't see any MMbase multicast traffic going over the wire when sniffing, using "tcpdump ip multicast. I do see muticast traffic coming from both Tomcat processes, but it's addressed to port 8012 which seem to be some default value. Multicast is active on the OS layer as ifconfig is telling me my NIC is running in multicast mode and I'm able to ping 224.0.0.1 and receive a response from all machines. I've a feeling I'm missing some sort of master switch, which I need to turn from false to true to enable MMbase clustering.

Something I noticed in the log when editing nodes (which should trigger multicast traffic):

2006-03-23 11:34:14,382 WARN mmbase.module.core.ClusterBuilder addRelationDirections.1656 - No relation defined between educations and images using RelationStep(tablename:pos2rel, alias:pos2rel, nodes:[], dir:destination, role:498) with direction(s) BOTH. Searching in 'destination' direction now, but perhaps the query should be fixed, because this should always result nothing.2006-03-23 11:20:04,826 FATAL STDERR println.258 - java.lang.NullPointerException 2006-03-23 11:20:04,827 WARN STDERR println.250 - at org.mmbase.module.builders.MMEvents.probeCall(MMEvents.java:119) 2006-03-23 11:20:04,827 WARN STDERR println.250 - at org.mmbase.module.builders.MMEventsProbe.run(MMEventsProbe.java:56) 2006-03-23 11:20:04,828 WARN STDERR println.250 - at java.lang.Thread.run(Thread.java:534)

I'm told however this is should not be directly related to my multicast problem.

Somebody who has an idea of what I'm missing?
Thanks in advance for any input!

Arjen Roodselaar


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

_______________________________________________
Developers mailing list
[email protected]
http://lists.mmbase.org/mailman/listinfo/developers


_______________________________________________
Developers mailing list
[email protected]
http://lists.mmbase.org/mailman/listinfo/developers

Reply via email to