Hi jonathan,
it seems that you are running an activemq node for every servicemix
instance. It's entirely possible that the activemq brokers might not
be networked together (I don't have your activemq configuration so
can't tell).
What I would recommend in general is to have two standalone broker
instances running, that are networked together - see: http://
www.activemq.org/Networks+of+Brokers - to ensure that they
communicate - otherwise you can get orphaned servicemix containers -
who do not know about other instances running.
Then the jmsFlow should be pointed at the activemq brokers that are
clustered together.
This will ensure that the activemq cluster is stable and is always
running, and allow you scale up the number of clustered servicemix
containers.
cheers,
Rob
On 3 Dec 2005, at 00:20, [EMAIL PROTECTED] wrote:
For the most part, once a cluster has been set up using the JmsFlow,
everything is now flowing between components in the cluster as
expected.
But...there's always a but, especially on Friday...the trick now
seems to be
getting the cluster up in the first place.
In this scenario I have three nodes, on the same host. Each node
fires up
its own ActiveMQ instance, and connects to it via the following
mechanism:
flowName="jms?jmsURL=reliable:(tcp://<activeMqInstance1>,tcp://
<activeMqInst
ance2>,tcp://<activeMqInstance3>), where activeMqInstance 2 and 3
are the
ports for the other nodes in the cluster.
So I startup node1, and wait for it to come up cleanly. Then I
start node2,
and wait for it to startup, and then exchange information with
node1 via
ActiveMq. And I wait...and wait some more, and it never connects.
Or, it
connects right away. If it doesn't connect I have to bounce one or
both of
the nodes. Adding the third node further complicates the procedure
- but
ultimately I have to bounce each node before they are all fully
crossed
referenced. I assume that this should not be the case? This is
using 2.1,
version 992.
regards,
/jonathan
