Normally, during a patching cycle, we currently HOLD an ESP class and then RELEASE the class when the patching cycle completes.
We ran into an issue last month where there was a problem with a handful of servers. We did want to release the class because not all servers were fully functional. As a result, the batch processing was late and there was customer impact.
Are there other ways to release some of the workload but keep others from running?
Release : 11.4 & 12.0
Component : CA ESP WORKLOAD AUTOMATION
Applicable scenarios for possible solutions:
- Quiesce the agent:
It's the easiest if you know that a particular server was problematic. Only negative impact is, the COMMQ file usage can increase since the requests will be queued up in it.
- Create one resource for every agent, and make the resource as not available when it needs the maintenance:
It will not impact the COMMQ file usage since the agent wobs will stay in RESWAIT status. Note: you can use AGENT_MON with AMNOTIFY to automatically set the agent resource as not available when it becomes INACTIVE.
- Use more EVENT CLASSes or HOLD events:
This can be done when the agent wobs for the particular agent are in separate events. If they are among many events while wobs for other agents are also defined, then neither of the two methods can help.