Skip to content

Frequently asked questions

What Kubernetes version do I need?

Kubernetes 1.33 or newer. CruiseKube relies on in-place pod resource updates for continuous optimization without recurring rollouts for request changes. Older clusters may install but will not get the full runtime story.

Does CruiseKube replace HPA or Karpenter?

No. It addresses vertical sizing (requests/limits policy per pod). HPA changes replica count; Karpenter / Cluster Autoscaler change node supply. CruiseKube works alongside them—see Use cases.

What happens if I use CPU- or memory-based HPA?

Workloads using CPU or memory metrics in HPA are skipped by CruiseKube so the two controllers do not fight. Other HPA modes may still interact indirectly; treat combined behavior as something to validate in staging. More detail in Tradeoffs.

Does it work with Java / JVM workloads?

Limited. The JVM heap is bounded by process and container limits; aggressive memory changes at runtime do not always map to a safe live resize. Size Java workloads conservatively and validate behavior. We expect this area to evolve.

Do you account for memory pressure?

Not yet in the way operators often mean (e.g. full memory pressure integration). Memory handling today leans on usage statistics, limits with headroom, and OOM-driven feedback. Treat this as a roadmap-sensitive area; see Algorithm — OOM handling.

Why does CruiseKube avoid CPU limits?

CPU limits throttle workloads and can hide contention behind artificial ceilings. CruiseKube prefers CPU requests for fairness and scheduling, uses PSI (where available) for contention signal, and documents the tradeoffs in Why CPU limits aren't needed.

Can CruiseKube help if I do not use node auto-provisioning?

Much less. The system assumes the cluster can grow nodes when honest demand requires it. Without that, you may still get recommendations, but you cannot rely on the same safety margins when the node pool is fixed and tiny.

Can I run it next to Cluster Autoscaler or Karpenter?

Yes. That is the expected deployment model. CruiseKube improves request honesty; autoscalers improve capacity supply.

Where are container images hosted?

Official charts reference images from TrueFoundry’s registry (see Helm chart reference). Security policy also mentions ghcr.io/truefoundry/cruisekube* for reporting scope—confirm your installed chart/image for your environment.

How do I report a security vulnerability?

Do not open a public GitHub issue. Email security@truefoundry.com with reproduction, impact, and versions. Full policy: Security (mirrors the repository SECURITY.md).

License?

CruiseKube is distributed under BUSL-1.1 (see the license badge in the README). Review terms before production use.

Where can I get help?