We have PEGA application running in Websphere application server v7 cluster environment. When ever there is an issue with login, we are following a PEGA instruction https://pdn.pega.com/how-clear-websphere-tempwork-directory to clear the cache and delete markup file. I do not understand why we have shutdown all the JVMs in the cluster same time, which make application not available to users. Why the cache is maintained between the jvms instead every JVM has to have its own cache.
You should not have to do that unless there are some rare unrecoverable situations. Deleting caches (including pega assembly cache, pega temp dir (not to be confused with websphere temp dir), WAS temp dir etc) should NOT be the normal practice, it only masks the real issue and the chance of finding out potential defects. Shutdown all JVMs is slightly different scenario. It may be a scheduled shutdown (e.g., applying hotfixes), but they should not have to be scheduled at the same time. For example, you can take out several nodes out of the rotation from the load balancer and users can still work on other nodes (although no new code changes).