Closed consideRatio closed 1 year ago
We have done this now. Here is a snippet from the JupyterHub helm chart configuration related to this.
singleuser:
# /dev/shm is mounted as a filesystem path, where writing to it means to
# write to memory.
#
# How to: https://stackoverflow.com/questions/46085748/define-size-for-dev-shm-on-container-engine/46434614#46434614
# Request for this by Ellie: https://fperezgroup.slack.com/archives/C020XCEFPEH/p1658168872788389
#
storage:
extraVolumes:
- name: dev-shm
emptyDir:
medium: Memory
extraVolumeMounts:
- name: dev-shm
mountPath: /dev/shm
/dev/shm
is a trick one can apply to mount RAM memory to a disk location. Can we make a k8s pod take some of its allocated memory and mount for use within the pod?This is recommended when using Ray, but can we do that in a k8s environment where we have Pods etc, and where we both have requested memory that we are guaranteed, and a memory limit, where we are not guarnateed but could potentially use?