[feature request] support scale subresource for mode: daemonset

[nazarewk]

Looks like Vertical Pod Autoscaler works only with deployment and statufulset modes, because the operator implements only those two:

opentelemetry-operator/pkg/collector/reconcile/opentelemetry.go

Lines 58 to 70 in 2d18b7b

	func updateScaleSubResourceStatus(ctx context.Context, cli client.Client, changed *v1alpha1.OpenTelemetryCollector) error {
	mode := changed.Spec.Mode
	if mode != v1alpha1.ModeDeployment && mode != v1alpha1.ModeStatefulSet {
	changed.Status.Scale.Replicas = 0
	changed.Status.Scale.Selector = ""
	return nil
	}
	name := naming.Collector(*changed)
	// Set the scale selector
	labels := collector.Labels(*changed)

It would be useful to have the same working for daemonset mode.

Is there any particular reason why it was omitted from #785 ?

field similar to replicas is present in DaemonSet.status https://kubernetes.io/docs/reference/kubernetes-api/workload-resources/daemon-set-v1/#DaemonSetStatus

resources:

• original implementation Add scale subresource status to the OpenTelemetryCollector CRD status #785
• slack thread on #sig-autoscaling

[yuriolisa]

I have no objections to implementing it. @jaronoff97, @TylerHelmuth, @pavolloffay WDYT?

[jaronoff97]

yeah this would be great to have, not sure why we didn't have it already.

[TylerHelmuth]

Let's do it

[pavolloffay]

No objections, how does the scaling work with daemonset? Will it schedule additional pods on the missing nodes?

[KitSutliff]

Seems like a good idea. Would be glad to have a go at it.

Just to clarify, we want to set a selector on the daemon set so that the vertical pod autoscaler can scale the collector, correct?

We could probably handle sidecar with similar logic. Should we set selectors for that too while we're at it?

[jaronoff97]

closed by #2665