The problem here is that we don't necessarily want to copy data every time, but if we don't copy data than databases can get screwed up.
In a teaching scenario, we could have ~50 people online at the same time. Our HDD is 3.8 TB. If our data directory is 10 GB (it is smaller), than we are still under 0.5 GB, so this is fine. But we are within a factor of 10 of the total disk space.
We know how to use nbgitpuller, read-only directories for common data, and are building scripts to distribute project directores, so i think this can be marked closed. We will use copies for most things, for now.
The problem here is that we don't necessarily want to copy data every time, but if we don't copy data than databases can get screwed up.
In a teaching scenario, we could have ~50 people online at the same time. Our HDD is 3.8 TB. If our data directory is 10 GB (it is smaller), than we are still under 0.5 GB, so this is fine. But we are within a factor of 10 of the total disk space.
As I type this, I think that we can start by copying data directories on user creation, but this needs to be scripted. Or maybe not? What about https://tljh.jupyter.org/en/latest/howto/content/nbgitpuller.html#howto-content-nbgitpuller?
@tngTUDOR Please share your opinion, if any.