More JMS Server Problems with Vista 4.2 no less!

When we upgraded to Vista 4.2 on August 22nd, and in subsequent testing to that event, we did several simulated JMS failovers. They were very successful. Relieved, we reconfigured our little application cluster (2 nodes only) so that NodeA was preferred and NodeB would be available.

But now…

We’re experiencing some kind of network latency -our best guess- between the Admin (which the forces that be required us to build as a virtual slice) and our Nodes …

The past two weeks now we’ve had 4 incidents with the Administrative VMSlice. Three of those events, although I’m not certain how to prove it without historical network bandwidth charts… seem to be that the ClusterAdmin was not receiving multicast back from NodeA to confirm its Aliveness, and began the migration process of JMS servers from NodeA to NodeB. NodeB then becomes unresponsive, causing the ClusterAdmin to change its mind? Is this a SWAG or truth? When I got to the Nodes the application wasn’t running on either one of them…

So… I configured Weblogic 9.2 to keep the JMS servers (Integration and Chat) on NodeA no matter what. At least that way, we keep the application servers up…

Here are the screen caps. Can anybody see anything wrong with this method? (Could worse go wrong that what already has?)

 

So I moved only NodeA over to “Chosen” for the effect, I trust, of pinning the JMS servers to that Node alone.

Advertisements

Comments are closed.