[Metricbeat] support kubernetes ressourcequotas

[f0]

it would be nice if metricbeat would fetch the kubernetes ressourcequotas.
This is very usefull to e.g display quotas per namespace vs. usage per namespace

[exekias]

It make sense to include this info, I've been checking kube-state-metrics and it doesn't retrieve these, so we would need to get them from the API server

[f0]

@exekias hm kube-state-metrics should include this https://github.com/kubernetes/kube-state-metrics/blob/master/collectors/resourcequota.go

[exekias]

That's right! I didn't see it in the docs, good to know, it should make things easier to implement then 👍

[jarrocha]

@exekias @f0
I have taken a look here:
https://github.com/elastic/beats/tree/master/metricbeat/module/kubernetes
and here:
https://github.com/kubernetes/kube-state-metrics/tree/release-1.3/Documentation
Does this involve creating a new state_resourcequotas package as in the first link?

[exekias]

yes, something similar to any of the state_* packages under https://github.com/elastic/beats/tree/master/metricbeat/module/kubernetes

[jarrocha]

I'll be working on this one.

[ruflin]

@jarrocha Great to hear. Thank you.

[jarrocha]

I have started work on this but there's some information that I am missing. Do you know where I can find more documentation to find usage for the different metric methods as it relates to Prometheus:

This is is from:
https://github.com/elastic/beats/blob/master/metricbeat/module/kubernetes/state_node/state_node.go#L42

"kube_node_info":                            p.InfoMetric(),
"kube_node_status_allocatable_pods":         p.Metric("pod.allocatable.total"),
"kube_node_spec_unschedulable":              p.BooleanMetric("status.unschedulable"),
"kube_node_status_ready":                    p.LabelMetric("status.ready", "condition"),

Also, for testing, will something similar to this work on Linux?
https://github.com/elastic/beats/blob/master/metricbeat/module/kubernetes/_meta/test/README.md

[jsoriano]

I have started work on this but there's some information that I am missing. Do you know where I can find more documentation to find usage for the different metric methods as it relates to Prometheus:

This is is from:
https://github.com/elastic/beats/blob/master/metricbeat/module/kubernetes/state_node/state_node.go#L42

"kube_node_info":                            p.InfoMetric(),
"kube_node_status_allocatable_pods":         p.Metric("pod.allocatable.total"),
"kube_node_spec_unschedulable":              p.BooleanMetric("status.unschedulable"),
"kube_node_status_ready":                    p.LabelMetric("status.ready", "condition"),

Do you have some specific question about these methods?
Automated godocs can be found here (not sure how updated they are). And you can always check the source in https://github.com/elastic/beats/blob/master/metricbeat/helper/prometheus/metric.go

For examples you can look at the state_* metricsets in the kubernetes module.

Also, for testing, will something similar to this work on Linux?
https://github.com/elastic/beats/blob/master/metricbeat/module/kubernetes/_meta/test/README.md

Yes, for Linux you can also use minikube, you will need to install kube-state-metrics service for these metrics.

If you want to use the docker daemon of the minikube VM you can do it with eval $(minikube docker-env). You may need to use it to pull or build images into the VM.