prometheus: mimir optional integration

[NachoxMacho]

I've currently got prometheus remote writing to mimir, and would like to use the prometheus plugin to pull the metrics, but can't provider the mimir endpoint as the prometheus api is on the path /prometheus. I could probably run another query frontend just for headlamp, but that seems like overkill to just enable this one plugin.

Ideally I think we should be able to append a path to the service definition, but barring that able to pass a full url as an alternative in the case the prometheus instance isn't directly available inside the cluster.

[jerry-pond]

please also support basic auth for prometheus service

[fatsolko]

same problem with Victoria Metrics

[DevOpsPop]

please also support basic auth for prometheus service

PR is on draft

[wilson0x4d]

i just wanted to point out that at least in the current version i am able to point at a mimir back-end:

port 1362 is a kind: Service in k8s that proxies to port 9009 (http) of the mimir containers.

Headlamp is v1.30.0 (7c48c2bd72b9bf5c44d21d7338cc7bea77d0ad2a)
Prometheus Plugin is v0.5.0

This allows me to see cpu/memory/network/etc stats on resource panels/pages.

Works great. THANKS!

[wilson0x4d]

as an aside, it's very strange we can't just provide the URL, which in my case would be:

svc-prod-ops-mimir.prod-ops-grafana:1362/prometheus

or expanded:

svc-prod-ops-mimir.prod-ops-grafana.svc.cluster.local:1362/prometheus

i've never seen anything that expects a "cluster/service:port" format, including k8s services. but maybe it's familiar to users of some k8s wrapper i've never used. seems to me k8s's own convention for DNS (service.namespace.resource.cluster:port) would make more sense. just an opinion.

[illume]

This was fixed in:

• prometheus: Add support for configuring subpath #187

and is part of Headlamp v0.30.0

[illume]

@yolossn @wilson0x4d makes a good point... why can't the whole address be used? Why is there a subpath?