Are you running DSM Services? It would be potential IP/port conflicts issues you have to overcome. Otherwise, you run a logic Websphere cluster with multiple JVMs within the same physical server (only thing you need to make sure is that you ensure unique set of ports for each node in the Websphere cluster). We run this configuration extensively for our internal environments.

For the logs, for each JVM we've added
-Dpega.logdir=/myApp/was/var/log

Since it's at server_X level, I guess you could define different value for each server_X running on same physical

You mention that only the first server starts, what error is shown in log for the second?

 

Thanks for the responses - Unfortunately they do not answer what I am looking for - the cluster configuration means that I only have 1 config file - for all instances - vertical or horizontally . This means that I need to specify these variables as dynamic for each instance - Obviously if I have only a fixed path or port - then the 1st instance that loads will work, but any others will fail as port is already assigned, runtimetemp is already being written to, logs are already allocated to the first instance. I understand that this may be more WebSphere related question, was hoping that someone else had done this in websphere who could give some pointers on the config file.  - Thanks

We've got websphere apps running with vertical & horizontal scaling.
This is done by our websphere team and this is common so it shouldn't be that unusual.
Each process has its own port on one server, its own log & config dir.

Then, for each PEGA JVM, you can define in its parameters, its own Log/Config/pegatemp directory.
for Hazelcast port you can also define in each prconfig.xml the range the instance can use.

so all combined, you should have a valid vertical design.