Virtual Machine CPU Usage Analysis - exclude time?

[Jarv]

Hi,

just wondering if anyone can think of a clever (or even obvious!) way round this

It would appear the backup system we are using is putting real stress on VM's when its backing them up at say 1AM in a morning. So for say half an hour the CPU spikes considerably and then drops right down. Because we use an auto ticketing system from SCOM set to 20 minutes a lot of this is causing tickets and of course alerts that just close off. While I am trying to persue the obvious route of fix the route problem...

Now unless I am wrong I can only see a couple of ways round this. First of all create a group for these servers and do an override that takes x amount of samples rather than 1 to cover that amount of time. So lets say half an hour. However my worry is we would then miss important stuff throughout the day if there was a genuine spike.
My only other thought is schedule some sort of maintenance mode to always cover this time period but again messy.

Any others thoughts how I could do this??

[sergey.g]

Hi Steve,

The solution may depend on the backup system you are using. For instance for Veeam Backup and Replication we have a special Management Pack with set of overrides which hide unwanted alerts and objects, this MP also collects events into SCOM which then could be used for further automation.

If your Backup solution produces events which can be collected by SCOM, then you should be able to create a monitor which can change its state when backup is running (backup started - unhealthy event, backup finished - healthy event), this monitor can have a recovery action which runs a PowerShell script which either disables CPU usage monitor for pre-configured group of VMs or enables it.

Of course such a script could be scheduled, but this is not as flexible as monitor tracking backup start-finish events.

We'll investigate if we can create some kind of example for you.

Also I'm not sure how ticketing system is configured in your environment, but maybe it is possible to block forwarding of certain alerts for a specific time, for instance you can specify a schedule for notifications in SCOM.

Thanks.

[Jarv]

Hi Sergey,

I have found a little bit more out about this. We are using Tivoli backup and that seems to cause the spike. I would add a couple of things on this. Checking the event logs the backup causes a common event each time it starts. Therefore a monitor could be written that could then do some sort of action. Second from our point of view I have been advised that we are changing to another system next month so this may not matter.
That said if you do have time I would be interested in the method you would use to suspend the monitoring for potential usage in similar scenarious.
Thanks!