Image Builder feature is released, administrators can only add images from valid registries on Image Management of PrimeHub.
Image Builder feature, administrators can build custom images for users according to requirements of environment such as pre-installed python packages, environmental variables configurations, and so on. Those prerequisites which were required to set up manually every time can be set up once in images. Users are able to re-use images to proceed to projects without spending time on environment setup after images are instantiated as containers.
Basically, we can use
Image Builder to build any images we want; in terms of images for PrimeHub, we use images for two conditions, one is for Job Submission, the other is for JupyterHub.
For Job Submission
Since requirements for running jobs are varied with tasks. As long as images are valid to be instantiated as containers, we can select those images as environments for running jobs.
Only images which are installed with JupyterHub environment are able to be used for launching JupyterHub in PrimeHub. PrimeHub expects installed JupyterHub inside the image; after starting the container, PrimeHub tries to launch Jupyter environment.
For the sake of building images with JupyterHub, we must use base images which are JupyterHub-compatible; we can have them from official registries.
There are two official registries where we can have JupyterHub-compatible base images, one is from Jupyter, the other is from InfuseAI.
|InfuseAI||A read_registry token is required to access the registry, please contact InfuseAI to acquire the token.|
Images provided by InfuseAI are:
|minimal-notebook||R and Julia|
Build Images via Image Builder
Build Images via Dockerfile
E.g. a Dockerfile using a existing Jupyter image.
FROM jupyter/minimal-notebook:177037d09156 # Get the latest image tag at: # https://hub.docker.com/r/jupyter/minimal-notebook/tags/ # Inspect the Dockerfile at: # https://github.com/jupyter/docker-stacks/tree/master/minimal-notebook/Dockerfile # install additional package for customization RUN pip install --no-cache-dir astropy
E.g. a Dockerfile using a existing image provided by InfuseAI.