Add Image

Image, is a working environment for an instance, we could have pre-built suitable images from public/private registries. We could even build our own images with customization and push to our registry by building customer. Here is a reference, Custom Image Guideline, describing which official registries we can pull images from and how we build our own ones.

This quick-start shows how we add an pre-built image on PrimeHub for users who can choose it for launching an instance on PrimeHub. If you haven't built any custom image, here is the [quickstart] build image. Here we are going to add that custom image which is installed with fastai v1 library on PrimeHub.

Let's Add Image

1. Log in as an administrator and switch to Admin Portal.

2. Enter Images management and click + Add for adding an image spec.

3. Fill Name with gcr-fastai-v1.

4. Select Type cpu.

5. Fill Container image url with the url of our custom image. E.g. gcr.io/infuseai/fastai-v1:1d1bxxxx.

6. Since we put our image on Google Container Registry (It varies according to your real circumstance.), it requires a pull-secret to pull down the image, we check off Usage Image Pull Secret and select the right secret.

7. Enable Global to make it available to all of users.

8. Click Confirm to save the setting.

Now users can select this custom image when launching an JupyterHub instance, once the jupyterhub is launched, we can check the fastai library version in a notebook.

Alternative

We, of course, can add an image which is located on public registry without a pull-secret; using url jupyter/tensorflow-notebook at step 5 instead and leave Use Image Pull Secret unchecked.

Next

By far, we have created users, groups, instance types and added images, users are ready to launch a JupyterHub instance. Next, we can go further to try custom build images.