feat: propagate trace via annnotation for limitador cr & auth configs

[KevFan]

Description

Closes: #1690
Requires: Kuadrant/limitador-operator#222 Kuadrant/authorino#574

• Propagates trace to limitador cr via annotation and links the limitador operator trace to the kuadrant opertor trace
• Propagates trace to authorino auth configs cr via annotation and links the authorino control plane trace to the kuadrant opertor trace

Limitador Operator -> Kuadrant Operator

Authorino -> Kuadrant Operator

[alexsnaps]

👍

[KevFan]

Only setting annotation here. Did not also set the annotation at limitador reconciler at:

kuadrant-operator/internal/controller/limitador_reconciler.go

Lines 62 to 84 in a4526a7

	limitador := &limitadorv1alpha1.Limitador{
	TypeMeta: metav1.TypeMeta{
	Kind: "Limitador",
	APIVersion: "limitador.kuadrant.io/v1alpha1",
	},
	ObjectMeta: metav1.ObjectMeta{
	Name: kuadrant.LimitadorName,
	Namespace: kobj.Namespace,
	OwnerReferences: []metav1.OwnerReference{
	{
	APIVersion: kobj.GroupVersionKind().GroupVersion().String(),
	Kind: kobj.GroupVersionKind().Kind,
	Name: kobj.Name,
	UID: kobj.UID,
	BlockOwnerDeletion: ptr.To(true),
	Controller: ptr.To(true),
	},
	},
	},
	Spec: limitadorv1alpha1.LimitadorSpec{
	MetricLabelsDefault: ptr.To("descriptors[1]"),
	},
	}

since if we apply the annotation at both these places, one will override the other and cause a reconcile loop. Maybe this is a non-issue since I'm guessing we only care about tracing the reconcile of the limitador limits portion.

But I wonder if we should/want to merge the reconcilers for limitador creation and the limitador limits

[adam-cattermole]

Yeah that's a good point, I do tend to agree we probably only care about limit reconciles but it does open up the question of why we have both

[KevFan]

Updated PR to also inject the trace annotation into auth configs to link traces (Kuadrant/authorino#574 is the related authorino PR)

[adam-cattermole]

Managed to verify this with the other changes and it's working as it's intended.

I don't think it's necessarily tied to this PR; but we might need to revisit this solution - once we set the annotation all reconciles whether they were triggered by the kuadrant-operator, other changes in the cluster, or users changing the CR will be linked to the last time we set the traceparent. I don't think this is necessarily a bad thing as it's still providing the link to the last time the kuadrant-operator updated it; but the source of that specific reconcile or change may not be at all related.

[KevFan]

Managed to verify this with the other changes and it's working as it's intended.

I don't think it's necessarily tied to this PR; but we might need to revisit this solution - once we set the annotation all reconciles whether they were triggered by the kuadrant-operator, other changes in the cluster, or users changing the CR will be linked to the last time we set the traceparent. I don't think this is necessarily a bad thing as it's still providing the link to the last time the kuadrant-operator updated it; but the source of that specific reconcile or change may not be at all related.

Yeah, that's the only thing with this. The traceparent annotation should only be viewed as the last time that kuadrant operator reconciled any "managed-fields" of these CRs 🤔