[Sig-scale] Defining SLIs/SLOs

[Ryan Hallisey US]

Hey folks,

For several years, sig-scale has been monitoring scale and performance metrics across kubevirt releases. We've reached a point where we would like to define several SLIs and SLOs for the project, similar to what Kubernetes does. This would be a big step for kubevirt and the community to support. So sig-scale is writing this thread to share our recommendations on how the community can support this.

Here's what we need your feedback on:

- What SLIs should kubevirt support?

SIG-Scale recommendations:

1. VMI create-to-running time
2. VM create-to-running time
3. VMI delete-to-cleanup time
4. VM delete-to-cleanup time

Future SLIs we can evaluate:

1. KubeVirt Subresources
2. Hotplug
3. Webhooks
4. Kubevirt controlplane mem/cpu usage

- Should the Performance metrics used to measure SLIs be release blocking?

SIG-Scale recommendation: Yes

- Should the Performance test used to measure SLIs be expanded to account for ~95% of Kubevirt deployments - 100 Nodes and 1000 VMIs?

SIG-Scale recommendation: Yes

-Ryan