Prometheus or "metrics-server" ?

[kubeadm]

Hawkular has been deprecated. Bug fixes "may" be made in Kubernetes 1.11.  Setup of hawkular will be removed in K8s 1.12 and will be completely removed in K8s 1.13

ref: https://github.com/kubernetes/heapster/blob/master/docs/deprecation.md

https://github.com/kubernetes/heapster

Suggested alternatives are: Prometheus or Metrics Server

https://github.com/prometheus/prometheus

https://github.com/kubernetes-incubator/metrics-server

Which one is OCP/RedHat more inclined towards ? 

[flozano · Accepted solution]

Hi Kubeadm,

Keep in mind that new features and deprecations in the upstream might take time to happen in the supported product. Currently OpenShift supports the heapster/hawkular stack and customers will continue to receive support and patches for it according to their subscription agreement, at least until the current (v10) and release of OpenShift reaches end-of-life . The hawkular stack is also supported by CloudForms.

Red Hat usually follows the directions from the upstream community, but in a way that minimizes the impact with its customers and does not violates SLAs. Prometheus is already a tech preview feature of the current OpenShift release. Red Hat is also experimenting working with Prometheus in many fronts, for example Open Stack. The plan AFAIK is to have full support for prometheus, but no OpenShift release target is set yet.

It seems to me that prometheus replaces kawkular (metrics history and query) while the metrics server replaces heapster (current metrics for pod autoscaling). I hope I got them right.

[SimonP]

@flozanois correct. Prometheus will eventually replace Hawkular while Heapster is going to be superseded by metrics-server. This is consistent with the general trend in the upstream community.