In the metric `cluster:capacity_cpu_cores:sum` there is an attribute label `label_node_role_kubernetes_io` that has `infra` or `master`. There is no label for `worker`. If the infra nodes are missing this label, they get added into the "unlabeled" worker nodes. 

For example:
This cluster has all three types `cluster:capacity_cpu_cores:sum{_id="0702a3b1-c2d8-427f-865d-3ce7dc3a2be7"}`

But this cluster has the infra and worker merged. `cluster:capacity_cpu_cores:sum{_id="0e60ac76-d61a-4e6d-a4f3-269110b6b1f9"}`


If I count clusters that have sockets with infra but capacity_cpu without infra, I get 7,617 cluster for 2023-03-15

If I count clusters that have sockets with infra but capacity_cpu with infra, I get 2,015 cluster for 2023-03-15

That means that there are 5602 clusters that are missing the infra label. 

This metric is used to identify the vCPU/CPU count that is used in TeleSense. This is presented to the Sales teams and upper management. If there is another metric we should use, please let me know. Otherwise, this needs to be fixed. 

1.
2.
3.

refer to Slack thread: https://redhat-internal.slack.com/archives/C0VMT03S5/p1678967355450719