POC: how to use jupyterhub api to determine active/inactive users

[asmacdo]

Just a POC so we dont lose this work

[yarikoptic]

good to know, but as @satra pointed out this would just work for the duration of the instance and we would not have information persist across restarts of the hub. If it is correct, we can close unless we add some persistence to it - e.g. dumping those into some file to be reloaded upon start etc. Is that easily doable?

[asmacdo]

This info is stored in the JupyterHub database, which is persistent across restarts—but not if the hub is fully torn down and the database is lost. Since the DB lives on EBS, we could create backups or snapshots and restore from those if needed. That seems feasible, though I’m not familiar with the full restore workflow.

FWIW, I think it’s reasonable to leave DB backup/restore out for now. In practice, we shouldn’t need to tear down the entire hub often. And even in the rare cases that we do, the only effect would be potentially carrying a some extra data for 30–45 days afterward.

That said, having a DB backup/restore process would be nice to have—especially before upgrades—but probably not a high priority since this data is only relevant for cleanup and isn’t critical to core functionality.