Arvados on Kubernetes

Arvados on Kubernetes is implemented as a Helm Chart.

Note:

This Helm Chart does not retain any state after it is deleted. An Arvados cluster created with this Helm Chart is entirely ephemeral, and all data stored on the cluster will be deleted when it is shut down. This will be fixed in a future version.

Overview

This Helm Chart provides a basic, small Arvados cluster.

Current limitations, to be addressed in the future:

  • An Arvados cluster created with this Helm Chart is entirely ephemeral, and all data stored on the cluster will be deleted when it is shut down.
  • No dynamic scaling of compute nodes (but you can adjust values.yaml and reload the Helm Chart
  • All compute nodes are the same size
  • Compute nodes have no cpu/memory/disk constraints yet
  • No git server

Requirements

  • Kubernetes 1.10+ cluster with at least 3 nodes, 2 or more cores per node
  • kubectl and helm installed locally, and able to connect to your Kubernetes cluster

If you do not have a Kubernetes cluster already set up, you can use Google Kubernetes Engine for multi-node development and testing or another Kubernetes solution. Minikube is not supported yet.

Initialize helm on the Kubernetes cluster

If you already have helm running on the Kubernetes cluster, proceed directly to Start the Arvados cluster below.

$ helm init
$ kubectl create serviceaccount --namespace kube-system tiller
$ kubectl create clusterrolebinding tiller-cluster-rule --clusterrole=cluster-admin --serviceaccount=kube-system:tiller
$ kubectl patch deploy --namespace kube-system tiller-deploy -p '{"spec":{"template":{"spec":{"serviceAccount":"tiller"}}}}'

Test helm by running

$ helm ls

There should be no errors. The command will return nothing.

Clone the repository

Clone the repository and nagivate to the arvados-kubernetes/charts/arvados directory:

$ git clone https://github.com/curoverse/arvados-kubernetes.git
$ cd arvados-kubernetes/charts/arvados

Start the Arvados cluster

Next, determine the IP address that the Arvados cluster will use to expose its API, Workbench, etc. If you want this Arvados cluster to be reachable from places other than the local machine, the IP address will need to be routable as appropriate.

$ ./cert-gen.sh <IP ADDRESS>

The values.yaml file contains a number of variables that can be modified. At a minimum, review and/or modify the values for

  adminUserEmail
  adminUserPassword
  superUserSecret
  anonymousUserSecret

Now start the Arvados cluster:

$ helm install --name arvados . --set externalIP=<IP ADDRESS>

At this point, you can use kubectl to see the Arvados cluster boot:

$ kubectl get pods
$ kubectl get svc

After a few minutes, you can access Arvados Workbench at the IP address specified

  • https://<IP ADDRESS>

with the username and password specified in the values.yaml file.

Alternatively, use the Arvados cli tools or SDKs:

Set the environment variables:

$ export ARVADOS_API_TOKEN=<superUserSecret from values.yaml>
$ export ARVADOS_API_HOST=<STATIC IP>:444
$ export ARVADOS_API_HOST_INSECURE=true

Test access with:

$ arv user current

Reload

If you make changes to the Helm Chart (e.g. to values.yaml), you can reload Arvados with

$ helm upgrade arvados .

Shut down

Note:

This Helm Chart does not retain any state after it is deleted. An Arvados cluster created with this Helm Chart is entirely ephemeral, and all data stored on the Arvados cluster will be deleted when it is shut down. This will be fixed in a future version.

$ helm del arvados --purge

Previous: Arvados-in-a-box Next: Prerequisites

The content of this documentation is licensed under the Creative Commons Attribution-Share Alike 3.0 United States licence.
Code samples in this documentation are licensed under the Apache License, Version 2.0.