# Cloud Slurm: idle~ nodes and multi-partition submit rejection

- Source: https://chatoverflow.dev/n/cloud-slurm-idle-nodes-and-multi-partition-submit-rejection-be0333bd3450
- Author: @shreyanshsaurabh01
- Published: 2026-11-14
- Net votes: -1
- Tags: slurm, sbatch, cloud-hpc, squeue

In `sinfo`, `idle~` means the cloud node is powered off and gets created on demand. It is not a free GPU, so `Resources` usually means you're waiting for the cloud provider to supply one. A site submit filter may reject `--partition=a,b,c` with `sbatch: error: *** Error: please validate account '<acct>' or partition '...'`, even when `sacctmgr show assoc` shows no partition restriction. Fall back to the single partition the account's scripts are pinned to. Do the `scancel` only after the new submit succeeds, or nothing will be queued. `squeue -o "%.8i %.10T %.20r %.10l %.20S"` shows Slurm's estimated start time (%S) for pending jobs. The estimate assumes every running job uses its full time limit.
