MON-1955: apis: add thanos querier CRD

[jan--f]

This only contains a CRD proposal yet. I'd like to discuss it before adding the operator code

[sthaha]

I wonder if it is better to user *metav1.LabelSelector for selecting namespaces like the rest of the k8s api ?

So we can run label query for over the namespaces and allows for a new namespace can be added without changing the spec.

---

nit: same as above *NamespaceSelector

[fpetkovski]

I guess a *metav1.LabelSelector cannot select all namespaces, but we could embed it into a NamespaceSelector like

type NamespaceSelector struct {
        *metav1.LabelSelector
	Any bool `json:"any,omitempty"`
}

or something along those lines

[jan--f]

Sure though I wonder if we actually want that.

If we just observe labels on a namespace, doesn't that mean that an owner of a namespace can add themselves to any thanos query instance, if they know the label to set?

[fpetkovski]

That's a good point. I think that addon namespaces just have a special redhat- prefix and don't have any labels on them, so having a label selector will not be useful for us. Let's stick with the simpler approach and we can change things later

[fpetkovski]

I think this looks good as a first iteration and we can even merge it. Would you mind running make lint to get the CI to pass?

[jan--f]

Sounds good, not sure what's up with the CI jobs. locally they all succeed.

[jan--f]

The CI implicitly assumes things are run with go1.16, I have 1.17 installed which creates the lint error here.

Do we really want to lint (and track) generated files? The build target depends on the controller-gen target anyway, so the autogenerated file is created when the operator is build. I'm not sure what the impact on development is, if that file isn't present from the get go.

[fpetkovski]

We have to have the zz_generated.deepcopy.go in the repository, otherwise the code won't compile. However, we can exclude it from the diff here: https://github.com/rhobs/monitoring-stack-operator/blob/main/.github/workflows/pr-checks.yaml#L41

[fpetkovski]

On second thought, we shouldn't be ignoring that file because it is part of the codebase. Let me see how we handle this in other projects.

[fpetkovski]

I'd say let's bump go to 1.17 in the CI and we can make it a requirement for local development too.

The problem seems to be caused when controller-gen is go installed with different go versions. Eventually we'll want to update to go 1.17 so why not just do it sooner rather than later.

[jan--f]

I'd say let's bump go to 1.17 in the CI and we can make it a requirement for local development too.

That's my favorite solution.

[fpetkovski]

lgtm

[fpetkovski]

lgtm, feel free to merge it when you wish