feat(opentelemetry-collector): add VPA and in-place resize support

[boqu]

Summary

• Add VerticalPodAutoscaler (VPA) support to the opentelemetry-collector chart, following the pattern already established in the opentelemetry-operator chart
• Add container resizePolicy support for Kubernetes in-place pod resource resize (>= 1.27)

VPA

• New templates/vpa.yaml template gated on autoscaling.k8s.io/v1 API capability and verticalPodAutoscaler.enabled
• Supports all three modes: daemonset, deployment, statefulset
• Configurable: recommenders, controlledResources, maxAllowed, minAllowed, updatePolicy, and full containerPolicies override
• New vpaKind helper in _helpers.tpl
• Values schema validation in values.schema.json
• CI test file ci/vpa-deployment-values.yaml

In-place resize

• New resizePolicy field on the collector container in _pod.tpl
• Supports NotRequired and RestartContainer restart policies per resource

[boqu]

@TylerHelmuth Thank you for your review. I pushed all the changes according to your review. Please let me know.

[boqu]

The CI failed due missing VPA CRD. See https://github.com/open-telemetry/opentelemetry-helm-charts/actions/runs/22117983560/job/63930896483?pr=2076.

VPA CRDs are not part of core Kubernetes. Is it ok to add VPA CRDs in the CI test cluster?

[dmitryax]

I would be strongly against bringing non-standard crds to this chart. Is there currently a way to install VPA separately?

[TylerHelmuth]

@boqu does .Values.extraManifest work here?

[boqu]

@boqu does .Values.extraManifest work here?

I added a minimal stub VPA CRDs here. But it's failing with below error. If my understanding is correct, helm fails when resolving all resource types against the cluster API servers before applying anything. So it looks like we still need to install the CRDs before the CI test.

Error: INSTALLATION FAILED: unable to build kubernetes objects from release manifest: resource mapping not found for name: "opentelemetry-collector-o6n8pl90u0" namespace: "opentelemetry-collector-o6n8pl90u0" from "": no matches for kind "VerticalPodAutoscaler" in version "autoscaling.k8s.io/v1"

[boqu]

I would be strongly against bringing non-standard crds to this chart. Is there currently a way to install VPA separately?

I think we only need to install the CRDs in the github action to run the CI test. If this is not acceptable, I can simply remove CI test.

[boqu]

I install the VPA CRD in the github setup action. The CI test can pass now.

[TylerHelmuth]

I have misunderstood how this feature is implemented in kubernetes. Is this feature not available in a standard +1.27 k8s install? What are the additional CRDs it needs?

[boqu]

No, the VPA doesn't come with kubernetes in default. It need to be installed separately. See https://kubernetes.io/docs/concepts/workloads/autoscaling/#scaling-workloads-vertically and https://github.com/kubernetes/autoscaler/tree/master/vertical-pod-autoscaler. The CRD can be found here.

[TylerHelmuth]

We've typically been against adding direct support for external CRDs, but seeing as how this is a kubernetes-owned CRD I'm ok adding this feature. @open-telemetry/helm-approvers what do you think?

[boqu]

Can I know if we can move this PR forward or anything outstanding concern?

[boqu]

I'd appreciate if we can move this PR forward.

[github-actions]

This PR was marked stale due to lack of activity. It will be closed in 14 days.

[boqu]

Any chance we can move this PR forward?