Skip to content

Quickstart: Launch Interactive Build Workloads with Connected Ports


This Quickstart is an extension of the Quickstart document: Start and Use Interactive Build Workloads

When starting a container with the Run:AI Command-Line Interface (CLI), it is possible to expose internal ports to the container user.

Exposing a Container Port

There are a number of ways to expose ports in Kubernetes:

  • NodePort - Exposes the Service on each Node’s IP at a static port (the NodePort). You’ll be able to contact the NodePort service, from outside the cluster, by requesting <NodeIP>:<NodePort> regardless of which node the container actually resides in.
  • LoadBalancer - Useful for cloud environments. Exposes the Service externally using a cloud provider’s load balancer.
  • Ingress - Allows access to Kubernetes services from outside the Kubernetes cluster. You configure access by creating a collection of rules that define which inbound connections reach which services.
  • Port Forwarding - Simple port forwarding allows access to the container via localhost:<port>

The document below provides examples for Port Forwarding and Ingress. Contact your Administrator to see which methods are available in your cluster


The step below use a Jupyter Notebook as an example for how to expose Ports. There is also a special shortcut for starting a Jupyter Notebook detailed here.

Port Forwarding, Step by Step Walkthrough


  • Login to the Projects area of the Run:AI Administration user interface at
  • Add a Project named "team-a"
  • Allocate 2 GPUs to the Project

Run Workload

  • At the command-line run:
runai config project team-a
runai submit jupyter1 -i jupyter/base-notebook -g 1 --interactive \ --service-type=portforward 
  --port 8888:8888  --command -- --NotebookApp.base_url=jupyter1
  • The Job is based on a generic Jupyter notebook docker image jupyter/base-notebook
  • We named the Job jupyter1. Note that in this Jupyter implementation, the name of the Job should also be copied to the Notebook base URL.
  • Note the interactive flag which means the Job will not have a start or end. It is the Researcher's responsibility to close the Job.
  • The Job is assigned to team-a with an allocation of a single GPU.
  • In this example, we have chosen the simplest scheme to expose ports which is port forwarding. We temporarily expose port 8888 to localhost as long as the runai submit command is not stopped

Open the Jupyter notebook

Open the following in the browser


You should see a Jupyter notebook. To get the full URL with the notebook token, run the following in another shell:

runai logs jupyter1 -p team-a

Ingress, Step by Step Walkthrough

Note: Ingress must be set up by your Administrator prior to usage. For more information see: Exposing Ports from Researcher Containers Using Ingress.


  • Perform the setup steps for port forwarding above.

Run Workload

  • At the command-line run:
runai config project team-a
runai submit test-ingress -i jupyter/base-notebook -g 1  --interactive \ 
  --service-type=ingress --port 8888  --command -- --NotebookApp.base_url=team-a-test-ingress
  • An ingress service URL will be created, run:
    runai list jobs

You will see the service URL with which to access the Jupyter notebook



With ingress, Run:AI creates an access URL whose domain is uniform (and is IP which serves as the access point to the cluster). The rest of the path is unique and is build as: <project-name>-<job-name>. Thus, with the example above, we must set the Jupyter notebook base URL to respond to the service at team-a-test-ingress

See Also

Last update: January 3, 2021