Not familiar with CSM, but with any load balancer, you need to set up session-based load balancing (first connection from client A goes to random server, subsequent connections from client A go back to the same server).
My question is more of LB usage on earlier statements made about LB persistent sessions:
(subsequent connections from client A go back to the same server).
Does this also mean that if I have 2 machines as clients to IS cluster of 4 machines over LB each of my clients will establish this persistent session to selected IS and any requests come from this server allways will end up on same IS server because they are directed to Loadbalancer like Big IP or CSM? And cinse I only have 2 clients they allways use same IS servers adn other 2 servers will not be used at all by Load balancer?
If I understand you right, “We’ve setup two 6.1 IS instances into a (wM) cluster and we just configured CSM with an alias to round robin between the addresses.”
Q 1) What is the benefit of having a CISCO CSM when you are performing a software cluster using webMethods? Are you using the Cisco CSM to create a common alias for HTTP/S calls? Have you set up load balancing on HTTP ports?
Q 2) Are you using RR as opposed to RR using Client Re-direct?
Another idea would be to have a third IS redirect calls to the two IS’s. RR using Server redirect.
You can set the persistant “time” on the LB. So as long as your two clients don’t “resend” during that “time”, the LB will/should send to a different IS instance.
The “time” is configurable… well should be configurable
Q1) The CSM is doing the balancing on the “internal” net and yes, as well to provide a common alias for the calls again internally. The RI will connect to the LB port on the IS instances for the “external” connections.
Q2) Round Robin just made sense for the internal net… and
Q3) This architecture was a choice, no technical reason for one method or the other. We decided if we were going to run an IS, we wanted it to process documents, not manage data flow.
