canonical/data-science-stack

Add `Waiting for gpu` state to `dss list`

Opened this issue · 1 comments

Why it needs to get done

In kubernetes is not possible to share a single GPU between pods. because of these when we try to create two gpu backed notebooks the latter one will go to Pending state with status of Insufficient resources .... There can be only one active Pod at a time using the single GPU in the computer.

Because of this we want to introduce a state Waiting for gpu in dss list. To recover from this state user can stop the notebook currently using the GPU. This will cause that the next notebook will get the gpu assigned.

What needs to get done

Add Waiting for gpu state to dss list which indicates that the notebook is trying to allocate gpu resource but there are currently no available.

When is the task considered done

dss list is supporting this state

Thank you for reporting us your feedback!

The internal ticket has been created: https://warthogs.atlassian.net/browse/KF-5603.

This message was autogenerated