We had a few presentations summarizing usage of JupyterHub, Binder and Dask on top of Kubernetes and the different challenges involved. It seems other places have or are looking into deploying similar environments, it's worth create a recipe with best practices and workarounds for common issues.
If you have information on your own deployment, please add a comment in this ticket with all the information (or a link to an external source) so we can then consolidate all the information into one recipe. Things that would be useful:
Components being used
Helm charts and values you might be using to deploy JupyterHub, Binder and other components
Highlight custom changes to support your own environment: storage system and identity integration, management of profiles and quotas, ...
Backend systems used to offload computing (HTCondor, SLURM, ...) and details on integration
This is a follow up to the CNCF Research Group meeting on February 17th 2021: https://docs.google.com/document/d/1vvXxW4Cd4P5gcmWGz-_yKbgJex2_NlSWaHtsk_56TnA/edit?ts=5d53c5ff#
We had a few presentations summarizing usage of JupyterHub, Binder and Dask on top of Kubernetes and the different challenges involved. It seems other places have or are looking into deploying similar environments, it's worth create a recipe with best practices and workarounds for common issues.
If you have information on your own deployment, please add a comment in this ticket with all the information (or a link to an external source) so we can then consolidate all the information into one recipe. Things that would be useful: