after our last maintenance window–when nothing scom-related was changed, just windows updates– most if not all of my agents have piled up hundreds or thousands of conhost and cscript processes, consuming all of the memory. manually flushing the agent cache (the ops console task says it ‘succeeds’ but generally does nothing) seems to fix it. however, on some boxes, I can’t delete the Health Service State folder even when the Microsoft Monitoring Agent service is stopped, because that folder is ‘in use by another process.’ so I have to reboot to clear the file lock, then manually flush the cache. which is obviously not ideal for production servers.
another interesting thing is that on all boxes, even with the Agent service stopped. I can’t kill any of the cscript processes, either with task manager or with powershell stop-process. stop-process doesn’t throw any errors, it just doesn’t actually stop the process. i’m an admin running powershell as administrator, and stop-process has no problem stopping all the conhost processes when the Agent service is stopped. it only has problems with the cscript process.
I’m having a similar problem in my environment. But it is on 2008R2 servers. It only started recently. I have some hanging cscript processes that I am unable to kill. I have to reboot the server to end them. Have you found any helpful information regarding this phenomenon?