Kubespray: Monitoring based on prometheus

[arslanbekov]

FEATURE REQUEST

It would be very convenient if monitoring was automatically raised monitoring cluster based on prometheus-operator.

As an example: https://itnext.io/kubernetes-monitoring-with-prometheus-in-15-minutes-8e54d1de2e13

[bobahspb]

Hi @owners!

I'm ready to prepare PR with 2 roles:

• prometheus operator
• cluster metrics collector + alermanager using Prometheus + Grafana

We use this configuration in our company.
Part of k8s-cluster.yml :

# Prometheus Operator. Needs for k8s metrics. Set true or install it by another way (helm?)
prometheus_operator_enabled: false

# K8s cluster metrics. Installed Prometheus Operator is required.
k8s_metrics_enabled: false

# Separate namespace for monitoring/metrics
monitoring_namespace: "monitoring"

I'm happy to help this project now and use metrics out of the box later, without local patches.

[arslanbekov]

@bobahspb nice! big thanks!
you role correct work with rbac?

[mattymo]

@bobahspb If you can use helm, I would strongly prefer it. We want to keep Kubespray code as simple as possible and not duplicate efforts other teams are using. Now that ansible has a native helm module, it should be quite easy to get prometheus-operator deployed. I had some good luck on my own actually.

[bobahspb]

@Arslanbekov , yes, it works with RBAC. I can send you our current version directly.

@mattymo , thanks for fast answer. Metrics collector is as popular subsystem as logs. EFK is included. What is different between metrics and logs?
Any way I can try to reconfigure it using HELM.

[woopstar]

@bobahspb Any news on this?

[arslanbekov]

if anyone needs this role (prometheus), we can put it in Ansible Galaxy repo

[woopstar]

I could try it out :) We have a big dev cluster testing on. And we're currently looking into how to best monitor with ease.

So please :)

[arslanbekov]

I'll try to do this week

[bobahspb]

@woopstar , PR has been sent.
https://github.com/kubernetes-incubator/kubespray/pull/2499

[AnatolyRugalev]

Looks like it was fixed by #2547

[desaintmartin]

Should it declares some ingress so that it is automatically exposed through host-based route?

[haoyehaoye]

2547 has been removed from master at https://github.com/kubernetes-incubator/kubespray/pull/2658.

Is there any pr is going on for installing Monitoring based on prometheus?

[haoyehaoye]

@desaintmartin I think it is better to have such a function, or implement it by using Node port or externalIP.