|
Before a Kubernetes upgrade or a routine node drain, cluster operators still check each application by hand to see whether it will survive. The Kubernetes project wants to replace that ritual with a shared, testable standard, and on Tuesday it announced a Workload Conformance Working Group to write one.
Certified Kubernetes tests the cluster, and Kubernetes AI Conformance tests whether a cluster can run AI and ML workloads. Nothing comparable exists for the workloads themselves, wrote Nabarun Pal of Broadcom on behalf of the group. The failures operators find are familiar ones: an app that can't tolerate disruption shows up during a drain, an upgrade breaks something that still depends on a removed API version, and some workloads request privileges they don't need or reserve resources they never use.
What the group plans to produce
The charter lists three deliverables: a specification of workload behavior for each Kubernetes minor version, a method for testing each requirement against a running workload, and upstream conformance tooling that runs those tests. Scope covers security, resilience during disruptions, resource use, networking, storage, API usage and observability, but the group hasn't agreed on specific requirements or how to test them.
One rule is set. Every requirement needs an objective test that a test binary can run, "so nothing subjective goes into the specification," the post says. The spec covers how a workload runs and leaves application logic out. The binary won't retest anything Certified Kubernetes already covers, and it won't require a particular packaging format or a vendor tool such as a policy engine or security scanner. |
