escorciav / kaust-cluster-status Goto Github PK

The current way to parse requested GPUs from the queue doesn't take into account jobs requesting GPUs across multiple nodes.

• Implement a patch for cluster.queue_status

Create a single page where we can go to each of the cluster (Ibex or Skynet)

• Remove price stuff from template

• Link website running on different servers.

Blocked by #4

Suggested by @ModarTensai

currently, it displays nodes with zero GPUs which does not add any value.

It's highly possible that the current version of gdragon.py doesn't work in Ibex under certain circumstances. I have a functional (buggy though) version in the Ibex server.

• Push local changes from Ibex.
• Deploy ibex backend.
• Make sure Skynet is not broken 😅

it's desirable to have a column with the gpu type of each host.

Get and report stats about time to get job dispatched e.g. min and std of waiting time of all jobs in the queue.

Get and report info about memory and cores available per node such that ppl can tweak sbatch-script.

Document what's going on under the hood