akirill0v / prometheus-kubernetes

Most common Prometheus deployment example with alerts for Kubernetes cluster

Prerequisites

Kubectl

kubectl should be configured.

Namespace

This example uses monitoring namespace.

Upload etcd TLS keypair

In case when you use TLS keypair and TLS auth for your etcd cluster, please put corresponding TLS keypair into the etcd-tls-client-certs secrets:

kubectl --namespace=monitoring create secret generic --from-file=ca.pem=/path/to/ca.pem --from-file=client.pem=/path/to/client.pem --from-file=client-key.pem=/path/to/client-key.pem etcd-tls-client-certs

Upload Ingress controller server TLS keypairs

In order to provide secure endpoint available trough the Internet you have to set example-tls secret inside the monitoring Kubernetes namespace.

kubectl create --namespace=monitoring secret tls example-tls --cert=cert.crt --key=key.key

Detailed information is available here. Ingress manifest example.

Create Ingress basic auth entry

With the internal-services-auth name. More info is here. Ingress manifest example.

Set proper external URLs to have correct links in notifications

Run EXTERNAL_URL=https://my-external-prometheus.example.com ./deploy.sh to deploy Prometheus monitoring configured to use https://my-external-prometheus.example.com base URL. Otherwise it will use default value: https://prometheus.example.com.

Assumptions

Disk mount points

This repo assumes that your Kubernetes worker nodes contain two observable mount points:

• root mount point / which is mounted as readonly /root-disk inside the node-exporter pod
• data mount point /localdata which is mounted as readonly /data-disk inside the node-exporter pod

If you wish to change these values, you have to modify node-exporter-ds.yaml, prometheus-rules/low-disk-space.rules, grafana-import-dashboards-configmap and then rebuild configmap manifests before you run ./deploy.sh script.

Data storage

This repo uses emptyDir data storage which means that every pod restart will cause data loss. In case when you wish to use persistant storage please modify the following manifests correspondingly:

• prometheus-deployment.yaml
• alertmanager-deployment.yaml
• grafana-deployment.yaml

Grafana dashboards

Initial Grafana dashboards were taken from this repo and adjusted.

Rebuild configmaps manifests

You have to rebuild manifests in case when you modify one of these directories:

• alertmanager-templates
• grafana-import-dashboards-configmap
• prometheus-rules

Command to rebuild the configmaps manifests:

./rebuild-configmaps.sh

Ingress controller

Example of an ingress controller to get an access from outside:

apiVersion: extensions/v1beta1
kind: Ingress
metadata:
  annotations:
    ingress.kubernetes.io/auth-realm: Authentication Required
    ingress.kubernetes.io/auth-secret: internal-services-auth
    ingress.kubernetes.io/auth-type: basic
    kubernetes.io/ingress.allow-http: "false"
  name: ingress-monitoring
  namespace: monitoring
spec:
  tls:
  - hosts:
    - prometheus.example.com
    - grafana.example.com
    secretName: example-tls
  rules:
  - host: prometheus.example.com
    http:
      paths:
      - path: /
        backend:
          serviceName: prometheus-svc
          servicePort: 9090
      - path: /alertmanager
        backend:
          serviceName: alertmanager
          servicePort: 9093
  - host: grafana.example.com
    http:
      paths:
      - path: /
        backend:
          serviceName: grafana
          servicePort: 3000

Alerting

Included alert rules

Prometheus alert rules which are already included in this repo:

• NodeCPUUsage > 50%
• NodeLowRootDisk > 80% (relates to /root-disk mount point inside node-exporter pod)
• NodeLowDataDisk > 80% (relates to /data-disk mount point inside node-exporter pod)
• NodeSwapUsage > 10%
• NodeMemoryUsage > 75%

Notifications

alertmanager-configmap.yaml contains smtp_* and slack_* inside the global sections. Adjust them to meet your needs.

Updating configuration

Prometheus configuration

Update command line parameters

Modify prometheus-deployment.yaml and apply a manifest:

kubectl --namespace=monitoring apply -f prometheus-deployment.yaml

If deployment manifest was changed, all Prometheus pods will be restarted with data loss.

Update configfile

Update prometheus-configmap.yaml or prometheus-rules directory contents (in this case don't forget to run ./rebuild-configmaps.sh) and apply them:

./update_prometheus_config.sh
# or
./update_prometheus_rules.sh

These scripts will update configmaps, wait until changes will be delivered into the pod volume (if the configmap was not changed, this script will work forever) and reload the configs. You can also reload configs manually using the commands below:

curl -XPOST --user "%username%:%password%" https://prometheus.example.com/-/reload
# or
kubectl --namespace=monitoring exec $(kubectl --namespace=monitoring get pods -l app=prometheus -o jsonpath={.items..metadata.name}) -- killall -HUP prometheus

Alertmanager configuration

Update command line parameters

Modify alertmanager-deployment.yaml and apply a manifest:

kubectl --namespace=monitoring apply -f alertmanager-deployment.yaml

If deployment manifest was changed, all Alertmanager pods will be restarted with data loss.

Update configfile

Update alertmanager-configmap.yaml or alertmanager-templates directory contents (in this case don't forget to run ./rebuild-configmaps.sh) and apply them:

./update_alertmanager_config.sh
# or
./update_alertmanager_templates.sh

These scripts will update configmaps, wait until changes will be delivered into the pod volume (if the configmap was not changed, this script will work forever) and reload the configs. You can also reload configs manually using the commands below:

curl -XPOST --user "%username%:%password%" https://prometheus.example.com/alertmanager/-/reload
# or
kubectl --namespace=monitoring exec $(kubectl --namespace=monitoring get pods -l app=alertmanager -o jsonpath={.items..metadata.name}) -- killall -HUP alertmanager

Pictures