Added
- New 'create an S3 bucket' in Portal for users
- Default S3 volume mounted for site (internal to site only)
- Daskhub shows machine costs on Server Options page
- (2) Jupyter GPU Tutorials
- Daskhub additions: jupyterlab-plotly, jupyter-ai, juptyer-archive extensions
- New GPU & Dask tutorials, AI tutorials
- Better UX (warnings on slow startups, improved Tutorials, etc)
- Bleeding Edge:
- Daskhub has jupyter-ai
- /scratch_space visible to daskhub workers if in same zone
- s3fs-fuse in as a hook for future improvements on S3 access
Changed
- Improved Portal page, shows more instances, shows S3 - costs, better VM costing
- Improved login (OAuth) and logos now addable.
- Container updates (many) including SunPy & PySPEDAS updates, emacs-nox, lftp
- Costing improved, including:
- better updates and summaries
- summary costs on Portal page
- S3 costs shown on Portal page
- Under the hood:
- Cluster ingress reconfigure improved portal/daskhub login
- Daskhub reverse proxy reduces load balancer costs ($768/yr)
- Updated AMI choices
- Additional Daskhub burst instances added for more reliable start-up
- Deprecated RHEL, focusing on AWSLinux & Ubuntu
- Updated kubernetes to 1.33 (good through July 29 2027)
- Registration page moved into the cluster
- Auto-makes scratch user directories
- Daskhub ipydatagrid, bqplot removed
Fixed
- lots of bug fixes, install fixes, teardown fixes
