Permalink
51 lines (35 sloc) 2.75 KB

Run Heapster in a Kubernetes cluster with an InfluxDB backend and a Grafana UI

Setup a Kubernetes cluster

Bring up a Kubernetes cluster, if you haven't already. Ensure that you are able to interact with the cluster via kubectl (this may be kubectl.sh if using the local-up-cluster in the Kubernetes repository).

Start all of the pods and services

In order to deploy Heapster and InfluxDB, you will need to create the Kubernetes resources described by the contents of deploy/kube-config/influxdb.

If you're running a different architecture than amd64, you should correct the image architecture for the grafana-deployment.yaml and the influxdb-deployment.yaml.

Ensure that you have a valid checkout of Heapster and are in the root directory of the Heapster repository, and then run

$ kubectl create -f deploy/kube-config/influxdb/

Grafana service by default requests for a LoadBalancer. If that is not available in your cluster, consider changing that to NodePort. Use the external IP assigned to the Grafana service, to access Grafana. The default user name and password is 'admin'. Once you login to Grafana, add a datasource that is InfluxDB. The URL for InfluxDB will be http://localhost:8086. Database name is 'k8s'. Default user name and password is 'root'. Grafana documentation for InfluxDB here.

Take a look at the storage schema to understand how metrics are stored in InfluxDB.

Grafana is set up to auto-populate nodes and pods using templates.

The Grafana web interface can also be accessed via the api-server proxy. The URL should be visible in kubectl cluster-info once the above resources are created.

Troubleshooting guide

See also the debugging documentation.

  1. If the Grafana service is not accessible, it might not be running. Use kubectl to verify that the heapster and influxdb & grafana pods are alive.

    $ kubectl get pods --namespace=kube-system
    ...
    monitoring-grafana-927606581-0tmdx        1/1       Running   0          6d
    monitoring-influxdb-3276295126-joqo2      1/1       Running   0          15d
    ...
    
    $ kubectl get services --namespace=kube-system monitoring-grafana monitoring-influxdb
    
  2. If you find InfluxDB to be using up a lot of CPU or memory, consider placing resource restrictions on the InfluxDB & Grafana pod. You can add cpu: <millicores> and memory: <bytes> in the Controller Spec and relaunch the controllers by running kubectl apply -f deploy/kube-config/influxdb/influxdb-grafana-controller.yaml and deleting and old influxdb pods.