This project is deprecated. Please see ✨ stencila/sparkla which provides compute sessions that are faster to start, more secure, and have finer grained control, and real-time user notifications, for resource usage.

In Stencila, execution Contexts are provided by a Host (if you are familiar with Jupyter, then Stencila Contexts are like Jupyter kernels and a Host is like a kernel gateway). In addition to providing Contexts, Hosts can also provide execution Environments (which have one or more Contexts with in them e.g. PythonContext, RContext) plus language specific packages (e.g. pandas, ggplot2). There is a Host HTTP API (currently in draft) available here.

This package, stencila/cloud, implements the Stencila Host API for running alternative execution Environments as Docker containers within a Kubernetes cluster. It is intended as a way of providing users of Stencila with an easy way to render Stencila documents withing alternative execution environments, without having to install packages themselves, or install and run Docker.

An example of a document using a RContext, hosted within the stencila/core execution environment, provided by stencila/cloud.

See the Dockerfile for building a container and deploy.yaml and minikube.yaml for example Kubernetes deployments.

Quickstart:

git clone https://github.com/stencila/cloud.git
cd cloud
npm install

Most development tasks can be run directly from npm or via make recipes.

You can run the sever locally but get it to create new session pods within an Minikube cluster.

Install minikube and kubectl. Then start the Minikube cluster,

minikube start
make run-with-minikube

The server will be available on http://localhost:2000 but will create new pods on Minikube. e.g. Create a new session using HTTPie

http PUT :2000/execute environment:='{"id":"alpine"}'

Use minikube dashboard or kubectl get pods to confirm that the session pods are getting created.

You can run the server inside Minikube and create new session pods there too.

Deploy stencila/cloud to the Minikube cluster,

minikube start
make run-inside-minikube

Check the Deployment is ready (the dashboard can be useful for this too: minikube dashboard),

kubectl get deployments

NAME                        DESIRED   CURRENT   UP-TO-DATE   AVAILABLE   AGE
stencila-cloud-deployment   1         1         1            0           1d

You can then get the URL of the host:

minikube service stencila-cloud-server --url

And check that it responds:

curl $(minikube service stencila-cloud-server --url)

If you're developing the Docker images in the stencila/images repo you can save time (and bandwidth) by not pushing/pulling images to/from the Docker Hub registry and the Minikube cluster. To do that, configure your local Docker client to use the Docker engine running inside the Minikube cluster:

eval $(minikube docker-env)