For those of you who have encountered the services running on your management server consuming a crazy amount of memory and CPU, my team found one of the root causes.
We came into work Monday and Service Manager was not responding and we restarted the services.
Once we logged in, we found that SCSM wasn’t importing tickets via the Exchange connector. Checking the connector in the Administration section, we saw that it was started, but not finishing.
We started seeing alerts for memory consumption and found that Microsoft.MOM.sdk.ServiceHost.exe quickly ramped up its memory consumption to nearly 100%, starving the OS, and crashing the system.
Since this service and the MonitoringHost.exe run under our SCSM system service account, we felt that the stalling Exchange connector was probably the issue.
We opened a PSA (Premier Support Agreement) ticket with Microsoft and started rolling out possible solutions, landing on the SPN update. This solved the problem. Here’s some more details:
First, check your Management Server’s event log for Event Id 26371 – The System Center Data Access service failed to register an SPN (I know, thank you “Captain Obvious”).
We then followed Kevin Holman’s blog post to fix it. While this post is an Ops Manager issue, the solution applies to any server with this particular event ID.
Jonathan Almquist also walks you through a solution for this.
Once you make the changes, restart your services (or reboot the server) and wait while the registration propagates. Depending on your domain, this could take an hour, as it did in our case.
Once fully registered, our Exchange connector finished successfully and the system returned to full operational capabilities. Here’s a screenshot of the services running as expected. Note that I took the shot as our Exchange connector was running showing it consuming CPU, which is expected. It will do this and then quickly return to normal.