If we spawn all single user notebooks into a slice and set cpu / memory limits on that slice, we can guarantee that notebook users can never, individually or in aggregate, OOM a machine.
Ideally we would also be able to just create the slice transiently if it doesn't exist.
If we spawn all single user notebooks into a slice and set cpu / memory limits on that slice, we can guarantee that notebook users can never, individually or in aggregate, OOM a machine.
Ideally we would also be able to just create the slice transiently if it doesn't exist.