infinispan question

[lmcdasi]

Hi,

  In order to have the jgroups activated in as7 I add in the inifinispan subsystem setup the following entry: start="EAGER" and that for each cache-container.

  The behaviour is that I would have 4 clusters started:

GMS: address=traffic_instance_PL-3-18819, cluster=cluster

GMS: address=traffic_instance_PL-3-16418, cluster=web

GMS: address=traffic_instance_PL-3-29553, cluster=sfsb

GMS: address=traffic_instance_PL-3-22689, cluster=hibernate

  So far so good. The question is that I do see the some warnings. Here is an example :

08:49:28,274 WARNING [org.jgroups.protocols.UDP] (OOB-16,web,traffic_instance_PL-3-16418) discarded message from different cluster "sfsb" (our cluster is "web"). Sender was traffic_instance_PL-3-29553

  Since the jgroups can define only one interface where all cluster messages are going through then I wonder if this is 'ok' since I will have messages that go from a cluster to a different cluster for 'nothing' just to get discarded.

  Should each cluster have a distinct port in order to avoid the above ? If yes, how this can be achieved ?

BR,

Dan S.

[pferraro]

It looks like there's something screwy with the shared transport setup, which should be the default behavior (meaning, all stacks should share the same UDP layer by default).  I'll look into this today.

In the meantime, can you try the following workaround?  In the jgroups subsystem, add the following to the UDP transport:

<transport type="UDP" ... >

   <property name="singleton_name">udp</property>

</transport>

[lmcdasi]

Hi Paul,

   Sorry - being busy with the other thread related to the services issue that I have. I can wait for the full fix on this one, unless the services issue gets addressed/solved faster or I there is a break on it ....

   Also, any link that would help me understand inifinispan clustering on AS7 ? I am just seeing the 'tip of the iceberg' for now

  Thanks!

BR,

Dan S.

[pferraro]

I should have an initial draft of the "High-Availability Guide" finished today.  Until then here are a couple resources:

(the section on clustering changes)

https://docs.jboss.org/author/display/AS7/How+do+I+migrate+my+application+from+AS5+or+AS6+to+AS7

http://infinispan.blogspot.com/2011/07/infinispan-in-jboss-as7.html

[pferraro]

BTW - those log messages are annoying, but harmless.

[lmcdasi]

OK

[pferraro]

OK - I found the issue.  There's a bug in the jgroups subsystem that causes the 1st channel created for a given stack to not be configured using a shared transport, even though subsequent channels are.  I'm submitting a pull request for this now.

https://issues.jboss.org/browse/AS7-1768

[lmcdasi]

Excellent ! Thank you for your help !

BR,

Dan S