cachemachine — JupyterLab image prepuller#

The Docker images used for lab pods run by the Notebook Aspect are quite large, since they contain the full Rubin Observatory software stack. If the image is not already cached on a Kubernetes node, starting a lab pod can take as long as five minutes and may exceed the timeout allowed by JupyterHub.

Cachemachine is an image prepulling service designed to avoid this problem by ensuring every node in the Science Platform Kubernetes cluster has the most frequently used lab images cached. It is also responsible for reporting the available images to Nublado, used to generate the menu of images when the user creates a new lab pod.

View on GitHub

/applications/cachemachine Application template

Source

lsst-sqre/cachemachine

Type

Helm

Namespace

cachemachine

Environments

base

values

Argo CD

ccin2p3

values

Argo CD

idfint

values

Argo CD

roe

values

Argo CD

summit

values

Argo CD

tucson-teststand

values

Argo CD

usdfdev

values

Argo CD

usdfprod

values

Argo CD

Guides#