Monitoring High VM CPU with Aria Operations

My lab doesn’t have a lot of resources. A couple of test VMs hit high CPU once in a while and needed a reboot. They’re only test VMs but I still wanted to catch the spikes. I’d like to automate the reboot later.

So I built an Aria Operations alert for high CPU. It works at the hypervisor level. That means no agent in the guest and no VMware Tools needed.

What I needed

• A vSphere environment with some VMs I can access
• Aria Operations that’s already monitoring that vSphere environment through the vCenter adapter
• A test VM where I can run a script to spike the CPU
• A quiet place where nobody interrupts you

1. Creating the alert definition

In Aria Operations I went to Configure → Alerts → Alert Definitions.

Then I clicked Add.

2. Naming it

I gave the alert a name and a description. I set the base object type to Virtual Machine and clicked Next.

3. Adding the condition

I went to Conditions → Metrics → CPU and dragged Demand (%) into the condition. Then I set the threshold and clicked Next.

4. Picking a recommendation

I dragged in the recommendations I’d want to see for fixing the VM, then clicked Next.

5. Choosing the policy

Policies let you pick a subset of the vCenter inventory. I could have used the vSphere Default Policy for everything. Or I could make one for a specific group of VMs.

6. Making the CPU spike

I used a Photon OS 4 VM, but any Linux or Windows test VM works. Just make sure it’s one you don’t mind spiking.

• Photon OS is free: download it here
• CPU spike script for Linux
• CPU spike script for Windows

Then I started it:

sh cpubusy.sh

7. Watching it in vCenter

In vCenter I selected the VM and watched the CPU start to climb.

8. Finding the alert

Back in Aria Operations I went to Troubleshoot → Alerts. I changed the view to Scope and picked Virtual Machine. Then I kept hitting refresh until the alert for my VM showed up.

It can take a few minutes to trigger, so give it time.

9. Stopping the spike

I went back to the VM and stopped the script.

Then I went back to Aria Operations and refreshed until the alert went away.

The alert clears itself once the condition is gone. I like that.