Prometheus configuration to scrape Kubernetes outside the cluster

prometheus.yml

# Prometheus configuration to scrape Kubernetes outside the cluster
# Change master_ip and api_password to match your master server address and admin password
global:
  scrape_interval:     15s
  evaluation_interval: 15s


scrape_configs:
# metrics for the prometheus server
- job_name: 'prometheus'
  static_configs:
    - targets: ['localhost:9090']

# metrics for default/kubernetes api's from the kubernetes master
- job_name: 'kubernetes-apiservers'
  kubernetes_sd_configs:
  - role: endpoints
    api_server: https://master_ip
    tls_config:
      insecure_skip_verify: true
    basic_auth:
      username: admin
      password: api_password
  scheme: https
  tls_config:
    insecure_skip_verify: true
  basic_auth:
    username: admin
    password: api_password
  relabel_configs:
  - source_labels: [__meta_kubernetes_namespace, __meta_kubernetes_service_name, __meta_kubernetes_endpoint_port_name]
    action: keep
    regex: default;kubernetes;https

# metrics for the kubernetes node kubelet service (collection proxied through master)
- job_name: 'kubernetes-nodes'
  kubernetes_sd_configs:
  - role: node
    api_server: https://master_ip
    tls_config:
      insecure_skip_verify: true
    basic_auth:
      username: admin
      password: api_password
  scheme: https
  tls_config:
    insecure_skip_verify: true
  basic_auth:
    username: admin
    password: api_password
  relabel_configs:
  - action: labelmap
    regex: __meta_kubernetes_node_label_(.+)
  - target_label: __address__
    replacement: master_ip:443
  - source_labels: [__meta_kubernetes_node_name]
    regex: (.+)
    target_label: __metrics_path__
    replacement: /api/v1/nodes/${1}/proxy/metrics

# metrics from service endpoints on /metrics over https via the master proxy
# set annotation (prometheus.io/scrape: true) to enable
# Example: kubectl annotate svc myservice prometheus.io/scrape=true
- job_name: 'kubernetes-service-endpoints'
  kubernetes_sd_configs:
  - role: endpoints
    api_server: https://master_ip
    tls_config:
      insecure_skip_verify: true
    basic_auth:
      username: admin
      password: api_password
  scheme: https
  tls_config:
    insecure_skip_verify: true
  basic_auth:
    username: admin
    password: api_password
  relabel_configs:
  - source_labels: [__meta_kubernetes_service_annotation_prometheus_io_scrape]
    action: keep
    regex: true
  - source_labels: [__meta_kubernetes_pod_annotation_prometheus_io_port]
    action: replace
    regex: (\d+)
    target_label: __meta_kubernetes_pod_container_port_number
  - source_labels: [__meta_kubernetes_service_annotation_prometheus_io_path]
    action: replace
    regex: ()
    target_label: __meta_kubernetes_service_annotation_prometheus_io_path
    replacement: /metrics
  - source_labels: [__meta_kubernetes_namespace, __meta_kubernetes_service_name, __meta_kubernetes_pod_container_port_number, __meta_kubernetes_service_annotation_prometheus_io_path]
    target_label: __metrics_path__
    regex: (.+);(.+);(.+);(.+)
    replacement: /api/v1/namespaces/$1/services/$2:$3/proxy$4
  - target_label: __address__
    replacement: master_ip:443
  - action: labelmap
    regex: __meta_kubernetes_service_label_(.+)
  - source_labels: [__meta_kubernetes_namespace]
    action: replace
    target_label: kubernetes_namespace
  - source_labels: [__meta_kubernetes_service_name]
    action: replace
    target_label: kubernetes_name
  - source_labels: [__meta_kubernetes_pod_node_name]
    action: replace
    target_label: instance

# metrics from pod endpoints on /metrics over https via the master proxy
# set annotation (prometheus.io/scrape: true) to enable
# Example: kubectl annotate pod mypod prometheus.io/scrape=true
- job_name: 'kubernetes-pods'
  kubernetes_sd_configs:
  - role: pod
    api_server: https://master_ip
    tls_config:
      insecure_skip_verify: true
    basic_auth:
      username: admin
      password: api_password
  scheme: https
  tls_config:
    insecure_skip_verify: true
  basic_auth:
    username: admin
    password: api_password
  relabel_configs:
  - source_labels: [__meta_kubernetes_pod_annotation_prometheus_io_scrape]
    action: keep
    regex: true
  - source_labels: [__meta_kubernetes_pod_annotation_prometheus_io_path]
    action: replace
    regex: ()
    target_label: __meta_kubernetes_pod_annotation_prometheus_io_path
    replacement: /metrics
  - source_labels: [__meta_kubernetes_namespace, __meta_kubernetes_pod_name, __meta_kubernetes_pod_container_port_number, __meta_kubernetes_pod_annotation_prometheus_io_path]
    target_label: __metrics_path__
    regex: (.+);(.+);(.+);(.+)
    replacement: /api/v1/namespaces/$1/pods/$2:$3/proxy$4
  - target_label: __address__
    replacement: master_ip:443
  - action: labelmap
    regex: __meta_kubernetes_pod_label_(.+)
  - source_labels: [__meta_kubernetes_namespace]
    action: replace
    target_label: kubernetes_namespace
  - source_labels: [__meta_kubernetes_pod_name]
    action: replace
    target_label: kubernetes_pod_name
  - source_labels: [__meta_kubernetes_pod_node_name]
    action: replace
    target_label: instance

Hi, thank you for the config.
Could you provide also Grafana dashboards for this metrics.

May I ask why do you use basic_auth twice for each job? one inside role, one outside?

@aghassabian The one inside role is used to authenticate with the API for discovery. The one outside is used to authenticate against the targets.

I have a K8 cluster which some applications running which exposes Prometheus end points. I want to discover all the endpoints automatically from another K8 cluster where Prometheus is installed.How can I find the API password for my k8 cluster and there is no TLS config(certificates) for HTTPS. How is it working without certificates?

I am getting below error for few endpoints;

Get https://X.X.X.X:443/api/v1/namespaces/kube-system/pods/kube-dns-6b4f4b544c-qwxxz:10053/proxy/metrics: context deadline exceeded

Get https://18.222.213.34:443/api/v1/namespaces/kube-system/pods/kube-dns-6b4f4b544c-qwxxz:53/proxy/metrics: context deadline exceeded

Get https://18.222.213.34:443/api/v1/namespaces/kube-system/pods/kube-dns-6b4f4b544c-88mrw:10053/proxy/metrics: context deadline exceeded

Get https://18.222.213.34:443/api/v1/namespaces/kube-system/pods/kube-dns-6b4f4b544c-88mrw:53/proxy/metrics: context deadline exceeded

I was getting the same, expose them over Nodeport, when services with cluster-ip are being proxied its timing out.

in the kubernetes-service-endpoints job,
if you are using Prometheus Operator and ServiceMonitor like me, you might want to use this relabelling below instead:

- source_labels: 
  - __meta_kubernetes_namespace
  - __meta_kubernetes_pod_name
  - __meta_kubernetes_pod_container_port_number
  - __meta_kubernetes_service_annotation_prometheus_io_path
  target_label: __metrics_path__
  regex: (.+);(.+);(.+);(.+)
  replacement: /api/v1/namespaces/$1/pods/$2:$3/proxy$4

because ServiceMonitor ultimately trying to scrape endpoints not service.

Hi,

I don't have metrics server running inside my k3s cluster. would i still be able to get the metrics?

Hi,

I am trying this config but getting the error.

level=error ts=2020-07-21T17:02:20.618Z caller=klog.go:94 component=k8s_client_runtime func=ErrorDepth msg="/app/discovery/kubernetes/kubernetes.go:333: Failed to list *v1.Node: Get https://master-server-ip/api/v1/nodes?limit=500&resourceVersion=0: dial tcp master-server-ip:443: connect: connection refused"

not sure if i am missing anything. any help would be greatly appreciated.

I am still seeing "server returned HTTP status 400 Bad Request" for pods and services. "/proxy/metrics".