fix capi-kubeadmconfig rule for hybrid providers #1494
Conversation
| expr: capi_kubeadmconfig_status_condition{type="Ready", status="False"} > 0 | ||
| expr: |- | ||
| ( | ||
| app_operator_app_info{status="not-installed", catalog=~"giantswarm|cluster|default", team!~"^$|noteam"} |
There was a problem hiding this comment.
What does that have to do with the app operator? I think the original metrics should bé capi_kubeadmconfig_status_condition{type="Ready right?
There was a problem hiding this comment.
welp, what did I do!?? I screwed the query with a broken copy/paste, and there were no unit tests to catch it! 😱
Thanks for catching it!
There was a problem hiding this comment.
I really wonder how I did it, because my Grafana Explore panel was actually using the right query 🤔
Anyway, hopefully I've now pushed the proper one 🤞
| expr: |- | ||
| ( | ||
| capi_kubeadmconfig_status_condition{type="Ready", status="False"} | ||
| * on(cluster_id) group_left(provider) | ||
| sum( | ||
| label_replace( | ||
| capi_cluster_info, "provider", "vsphere", "infrastructure_reference_kind", "VSphereCluster" | ||
| ) | ||
| ) by (cluster_id, provider) | ||
| ) > 0 |
There was a problem hiding this comment.
I'm not an expert on PromQL so I may be miss understanding the query, but aren't we basically disabling the alert for non vsphere clusters? We want this alert for all providers.
There was a problem hiding this comment.
Not at all :)
What we are doing here is replace the provider label to vsphere if the cluster type is a VSphereCluster
There was a problem hiding this comment.
That makes more sense 😄 Thanks! I guess the same would be needed for cloud-director?
There was a problem hiding this comment.
We do not need it for cloud-director because we do not have cloud director WCs running on cloud MCs yet.
We might need this for promox and so on i guess
There was a problem hiding this comment.
But I would prefer that we try to find a better solution in the long run :)
There was a problem hiding this comment.
That sounds good. Thank y'all for the explanations!
Towards https://github.com/giantswarm/giantswarm/issues/32587
This PR fixes multi-provider routing for
KubeadmConfigNotReadyalert.Checklist
oncall-kaas-cloudGitHub group).