[Blog] Best Practices and Recommendations for Refusing workloads from running on Kubernetes

[xamebax]

Description

Running workloads on Kubernetes can give developers the freedom and velocity they need, but that freedom can also give operators a headache.

There are mechanisms like admission webhooks and operators that can enforce human-defined criteria and prevent certain workloads from being deployed to Kubernetes at all. Kyverno is an example of such an operator, but other tools also exist.

This is a vague question that might be worth exploring first before digging in to write a recommendation:

• what kind of criteria for refusing deploying Pods on Kubernetes can we potentially adopt to increase environmental sustainability? These could be criteria around for example making sure resource requests and/or limits are set, or making sure an HPA object is deployed also, or something completely different.
• what kind of tools can be used to aid refusal?

Outcome

A preliminary list of ideas that can be made concrete further, outlined in the working document.

To-Do

• dig into Kubernetes to find out which configuration of Deployments, StatefulSets, HPAs, etc. is often suboptimally set and needlessly increases a cluster's carbon footprint
• research operators and other mechanisms that can refuse objects being deployed to Kubernetes
• submit the outline for feedback

Code of Conduct

• I agree to follow this project's Code of Conduct

Comments

No response

[guidemetothemoon]

Updated the issue to be a blog post idea. We can look into transferring this as a content creation opportunity as part of the upcoming Community Group for example.