Per-cluster Kosmoo vs multiple-clusters-per-OpenStack

[MPV]

Hi, In what way (cardinality) do you deploy and run Kosmoo across clusters and OpenStack environments/tenants?

I'm asking since we've initially set it up to run "one per Kubernetes cluster", but since we have multiple Kubernetes clusters running in the same OpenStack region and tenant, it causes us to get the same metrics (and thus alerts) from multiple clusters (for example broken load balancers/servers when erroring in provisioning state).

[tobiasgiese]

Hi @MPV.

We are also using Kosmoo to run as "one per Kubernetes cluster". But the difference in our provisioning/operation is, that we're using one Openstack project (i.e., tenant) per cluster. So we have only metrics for this particular cluster.

Edit: but maybe it's enough for your use-case to use the name label in the kosmoo metrics, which describes the OpenStack resource name. Or we could add a feature to filter for resource names that matches a specific value.

[MPV]

@tobiasgiese Could you take a look at #121? 🙏

[tobiasgiese]

@MPV sorry, I haven't reviewed the PR because it's still a draft. But let me start with the review :)