ESXi Host CPU Contention in Aria Operations

ESXi Hosts have dozen of metrics in Aria Operations, one of them is CPU Contention (%).

As the hover indicates, this represents the Percent of time descendant virtual machines are unable to run because they are contending for access to the physical CPU(s). Or rather, the Percent of time the ESXi Host can't satisfy VMs demand for CPU. These are generally recognized levels of ESXi Host CPU Contention (%):

• 0–2%: Normal, no contention

• 2–5%: Low contention, monitor if sustained

• 5–10%: Moderate contention, may impact performance

• >10%: High contention, performance degradation likely

So, how is Aria Operations calculating this metric? It's defined in /usr/lib/vmware-vcops/user/plugins/inbound/vmwarevi_adapter3/conf/describe.xml as follows:

<ResourceAttribute key="host_contentionPct" nameKey="1191" unit="percent"/>
                <ComputedMetric expression="sum(${adapterkind=VMWARE, resourcekind=VMwareAdapter Instance, metric=mem|host_contention}) / (sum(${adapterkind=VMWARE, resourcekind=VMwareAdapter Instance, metric=summary|number_running_vcpus}) * 200)" key="mem|host_contentionPct"/>

The numerator is: Host System CPU Contention (ms), which I had to enable in the Policy to see.

The denominator is: (vCPU on Running VMs on that Host System) * 200

Host System CPU Contention % is: (Host System CPU Contention (ms)) / ((vCPU on Running VMs on that Host System) * 200)

Looking at an example:

CPU Contention (ms) = 20,286

vCPU on running VMs = 494

20,286 / (494 x 200) = .21, which is what we show for ESXi Host System CPU Contention (%).

Explore the describe.xml's, lots of computed metrics in them, enjoy!