Experiment №3
ARC Pulse, a heartbeat a day
A morning roll call for ARC's GPU nodes: who showed up, who's busy, and who called in sick.
Status history
Today
| Partition | Nodes | Idle | Mixed | Allocated | Down or drained |
|---|
Node states as defined in the sinfo manual: idle, the node is not allocated to any jobs and is available for use; mixed, the node has some of its CPUs allocated while others are idle; allocated, the node has been allocated to one or more jobs.
Out of service
| Partition | Node | State | Since | Reason given by ARC |
|---|
Collected each day with sinfo (node states and the reasons ARC gives) and squeue (jobs waiting and running), the standard SLURM commands any ARC user can run.
Why ARC Pulse
We were preparing an AI-assisted HUB workshop that runs on ARC's GPUs. Two days before it, more than half of ARC's GPU nodes were unavailable, and the timeline for bringing them back was uncertain. We wanted a simple view of who had shown up each morning, so we built one: a small daily page inspired by Ganglia, using only what any ARC user can already see.