Post

The old approach relied on crawling files to estimate directory sizes. That’s expensive on shared storage, so metrics had to run slowly to avoid impacting users. Now jupyterhub-home-nfs exports per-user usage / limits as Prometheus metrics, using filesystem quota accounting so it's much faster!

github.comGitHub - 2i2c-org/jupyterhub-home-nfs: In-cluster NFS server for JupyterHub home directories with usage quota enforcementIn-cluster NFS server for JupyterHub home directories with usage quota enforcement - 2i2c-org/jupyterhub-home-nfs