Expose kube-proxy prometheus metrics

[Cajga]

What would you like to be added:
Please expose kube-proxy metrics so, they can be scraped with (ideally, also from an external) prometheus. Currently, in EKS-A, kube-proxy metrics are only available from the node itself:

# netstat -natupl|grep 10249
tcp        0      0 127.0.0.1:10249         0.0.0.0:*               LISTEN      1428/kube-proxy
# curl -L -s http://127.0.0.1:10249/metrics |grep -v ^#|head -n 5
apiserver_audit_event_total 0
apiserver_audit_requests_rejected_total 0
go_gc_cycles_automatic_gc_cycles_total 1008
go_gc_cycles_forced_gc_cycles_total 0
go_gc_cycles_total_gc_cycles_total 1008

One simple solution would be to modify the default configmap of kube-proxy and add 0.0.0.0:10249 for metricsBindAddress:

$ kubectl describe configmaps -n kube-system kube-proxy |grep -i address
bindAddress: 0.0.0.0
bindAddressHardFail: false
healthzBindAddress: ""
metricsBindAddress: ""
nodePortAddresses: null

According to the official documentation of kube-proxy config, without a value, it defaults to 127.0.0.1:10249

Why is this needed:
To be able to scrape kube-proxy metrics with prometheus which are helpful to understand the state of the cluster (and if kube-proxy is down :) ).

[panktishah26]

@Cajga Thanks for creating a github issue. We will take a look into this and will keep you posted.

[Cajga]

@panktishah26, thanks ahead.

Just wanted to highlight that metricsBindAddress: 0.0.0.0:10249 is the default in EKS (region) clusters. See ticket about the same discussion there.