[Automated/Action] Install Kepler dashboard

[rossf7]

Description

Grafana is installed by kube-prometheus but it would be good to be able to add our own dashboards. Likely starting with the Kepler dashboard. https://sustainable-computing.io/installation/kepler-operator/

Issue

Parent: #5

[nikimanoledaki]

Thank you @rossf7, agreed! Here is a list of suggested steps.

Steps / Requirements

• Setup the TF Grafana provider
• Setup basic auth to sign in localhost:3000 using admin:admin
• Import the default Kepler dashboard from Kepler operator repository
• Make the dashboard public using the TF Grafana provider by setting share = "public"

[dipankardas011]

Currect me if I am wrong
so this issue tries to solve the import of kepler grafana dashboard to the already up and running k3s cluster provided that kepler is already successfully installed

specifically these steps for automations?
#17 (comment)

[dipankardas011]

Thank you @rossf7, agreed! Here is a list of suggested steps.

Steps / Requirements

• Setup the TF Grafana provider
• Setup basic auth to sign in localhost:3000 using admin:admin
• Import the default Kepler dashboard from Kepler operator repository
• Make the dashboard public using the TF Grafana provider by setting share = "public"

Also the planned PR of installing kube-prometheus already installs the grafana. so may be we might need to stop the grafana as a component install?
never see multiple grafana instances in a single k8s clusters thats why asking

[AntonioDiTuri]

Yes @dipankardas011 you are right when installing prometheus we also get grafana.
However two points need to be still achieved:

• Import the default Kepler dashboard from Kepler operator repository.
• Make the dashboard public using the TF Grafana provider by setting share = "public"

In order to achieve the seconds we can use gateway api with cilium

[dipankardas011]

so for the importing dashbaord. is it manual one time operation or we are planning for automatation of this?

[rossf7]

@dipankardas011 It needs investigation but the kube-prometheus-stack helm chart includes the grafana helm chart
https://github.com/prometheus-community/helm-charts/tree/main/charts/kube-prometheus-stack

The grafana helm chart can deploy a sidecar that looks for configmaps with a label.
https://github.com/grafana/helm-charts/tree/main/charts/grafana#sidecar-for-dashboards

So we could import the Kepler dashboard as a configmap managed by Flux. Later if we need our own dashboards they could also be managed by Flux. WDYT?

@nikimanoledaki You likely know this better than I do! :) Do you think this approach makes sense?

I haven't used the Grafana TF provider. Are there benefits to using it instead of the Helm chart? With Helm we will need to use Gateway API (or Ingress) to make the Grafana instance accessible from the cluster.

[dipankardas011]

now I am understanding thanks @rossf7

[dipankardas011]

yes we can install the grafana dashboard without much effort

namespace doesnt matter as the sidecar looks for all the configmaps in all ns with specific label

just

apiVersion: v1
kind: ConfigMap
metadata:
  name: sample-grafana-dashboard
  labels:
     grafana_dashboard: "1"
data:
  kepler-dashboard.json: |-
    <raw...>

it works when the sidecar for the grafana in added during installing
I tested on kube-prometheus-stack
but was unable to make the kepler helm install when using kube-prometheus-stack up but was able to when using kube-prometheus

kube-prometheus-stack artifact link
https://artifacthub.io/packages/helm/prometheus-community/kube-prometheus-stack

[dipankardas011]

may be kepler service monitor work for kube-prometheus
docs uses that 😕

[rossf7]

@dipankardas011 Nice, thanks for investigating. I think we can store it in the monitoring namespace since we're going to move kepler there. We can use kepler-dashboard for both the CM name and filename.

It would also be good to have something in docs/ for syncing the dashboard with upstream. I don't think it changes often so no need to automate that.

may be kepler service monitor work for kube-prometheus docs uses that 😕

Hmm let's wait till prometheus is deployed with kube-prometheus-stack and kepler is moved to monitoring. As having everything in the monitoring ns may help with the service monitor.

cc @AntonioDiTuri

[dipankardas011]

was trying but it failed. so messaged in the kepler-project slack channel for there help
will be commenting in the slack

[dipankardas011]

other than that we can easily install the kepler grafana dashboard irrespective of the prometheus stack we are using

[dipankardas011]

Also this is the grafana Dashboard which is working

https://raw.githubusercontent.com/sustainable-computing-io/kepler/main/grafana-dashboards/Kepler-Exporter.json