Use new kube-state-metrics Helm chart

[tokyowizard]

As reported in #151, the KSM helm chart moved. But the KSM helm chart has moved yet again!

The current requirements.yaml is pointing to the version, 2.13.2, of the KSM helm chart that specifies KSM version 1.9.8. The new location of the KSM helm chart is here.

We'd like to be on KSM version 2+ because:

• We're on AWS EKS 1.20 and KSM 2+ has full support for 1.20. See KSM compatibility matrix
• We'd like to enable Datadog's Kubernetes State Metrics Core

[clamoriniere]

Hello,

Thanks for opening this issue.

We are aware of this new move :) We have plan to use the new repo and the version 2.13.2. But it is not a real urgence since no new ksm v1.9.x as been released yet.

But, we will not update this dependency on any major version (3.x.x) for 2 reasons:

• the kubernetes_state check is only compatible with kube-state-metrics v1.x.x, so keeping the kube-state-metrics chart version to v2.13.2. allow use to use the image version 1.9.8. This is to keep the metrics backward compatibility for user.
• the new kubernetes_state_core check is base and it expose the new KSM 2+ metrics, but doesn't require to deploy kube-state-metrics chart (more info in this blog post).
  So if you plan to use the new check, you should not worry anymore about deploying and updating the kube-state-metrics chart with the datadog chart.

the proper way to do so is to add the following configuration in your values.yaml file:

datadog:
  # disable kube-state-metrics nested chart
  kubeStateMetricsEnabled: false
  # enable the new `kubernetes_state_core` check
  kubeStateMetricsCore:
    enabled: true

It will:

• uninstall (during an upgrade) or not install the kube-state-metrics chart.
• enable the kubernetes_state_core check.
• disable the autodiscovery of the kubernetes_state check. So if you would like to deploy on your own kube-state-metrics on the same cluster, it will not trigger the auto-discovery and the execution of the kubernetes_state check.

Please let us know if I answered to your question, and if the solution that we propose works for you.

Cheers
Cedric

[clamoriniere]

Please reopen the issue, it you need more information