1 GB persisted space too small for custom envs
rsignell-usgs opened this issue · comments
@julienchastang, I created a custom conda env, only to find that at 3GB, it was larger than the 1GB of persisted space I had available. I think the user space should be increased to 10GB, to allow a couple of custom environments.
Out of curiosity, why is this conda environment so large?
I don't have any conda environments less than 1GB. They range from 1.8 to 4.5GB.
A lot of large packages is why they are large!
Here's an example env file that creates my IOOS environment:
https://raw.githubusercontent.com/ioos/notebooks_demos/master/environment.yml
jovyan@jupyter-julienchastang:/opt/conda/envs$ df -h
Filesystem Size Used Avail Use% Mounted on
none 59G 18G 41G 30% /
tmpfs 15G 0 15G 0% /dev
tmpfs 15G 0 15G 0% /sys/fs/cgroup
/dev/sda1 59G 18G 41G 30% /etc/hosts
/dev/sdb 976M 348M 561M 39% /home/jovyan
shm 64M 0 64M 0% /dev/shm
tmpfs 15G 0 15G 0% /sys/firmware
What about reusing /opt/conda/envs/? You have a lot more space there and it is writeable by the user I believe.
Writable, but also not persisted, right?
As I understand it, everything that is not in /home/jovyan is tied to the Docker file system so when the container gets restarted by Kubernetes, that information is lost. @zonca is this right? So, in sum, the conda environment is persisted as long as the container is alive.
@julienchastang you are right, but I think @rsignell-usgs meant persisted across restarts of the container.
I also think that 1GB is a bit too small, is 10 GB feasible in the Jetstream allocation?
Right, persisted across restarts of the container. Yes, 10GB should be feasible.
@zonca actually, how do you increase disk space?
singleuser:
capacity: 10Gi
seems to have no effect
jovyan@jupyter-throwaway:~$ df -h
Filesystem Size Used Avail Use% Mounted on
overlay 58G 5.5G 53G 10% /
tmpfs 64M 0 64M 0% /dev
tmpfs 15G 0 15G 0% /sys/fs/cgroup
/dev/sdb 976M 2.6M 958M 1% /home/jovyan
/dev/sda1 58G 5.5G 53G 10% /etc/hosts
shm 64M 0 64M 0% /dev/shm
tmpfs 15G 0 15G 0% /proc/acpi
tmpfs 15G 0 15G 0% /proc/scsi
tmpfs 15G 0 15G 0% /sys/firmware
I think it affects only newly created volumes
Right. The volume you see above was newly created. This is a "fresh" (experimental) jupyterhub.
Correct!
singleuser:
storage:
capacity: 10Gi