Grafana Helm Chart

This repo contains helm charts to deploy Prometheus and Grafana to your Kubernetes cluster. The Prometheus helm chart deploys the node_exporter and kube-state-metrics to expose cluster metrics. The Grafana helm chart comes with useful dashboards preconfigured. It also contains Pod Security Policies and is intended to work with no additional configuration for version of Pivotal Container Services 1.4+. It will also work with clusters from all major cloud providers.

Prerequisites

Configure Helm

The following commands may need to be executed for helm to function correctly.

kubectl create serviceaccount --namespace kube-system tiller 
kubectl create clusterrolebinding tiller-cluster-rule \
--clusterrole=cluster-admin --serviceaccount=kube-system:tiller
helm init --service-account tiller

Install

cd ~/workspace/charts-grafana
helm install . --name cluster-health --namespace observability

Port Forwarding

Grafana

1. Retrieve the grafana dashboard password by running the following

   # Assuming you are on a Mac OSX
   kubectl get secret cluster-health-grafana --namespace observability --output json | jq -r '.data."admin-password"' | base64 --decode

2. Create the port forward to the Grafana dashboard

   kubectl port-forward deployment/cluster-health-grafana 3000:3000 --namespace observability

3. Open your browser window and go to http://localhost:3000
4. Enter the username admin and the previously retrieved password.

Prometheus

1. Create the port forward to the Prometheus dashboard

   kubectl port-forward deployment/cluster-health-prometheus-server 9090:9090 --namespace observability
   

2. Open your browser window and go to http://localhost:9090

Installation Caveats

Cluster Size

Prometheus includes a few deployments, such as the prometheus server, alertmanager, node exporter daemonset, and kube-state-metrics. Grafana includes a deployment as well. You should consider this when deploying this chart and sizing your cluster.

Cluster with Admission Controllers

PodSecurityPolicy

The chart has PodSecurityPolicy enabled by default. See here for more information.

SecurityContextDeny

The grafana deployment uses PodSecurityContext to configure the grafana container to run with user id 472. This means that the grafana deployment will fail on any cluster with SecurityContextDeny enabled.