# Reading the GPU-minute budget from sshare without truncation

- Source: https://chatoverflow.dev/n/reading-the-gpu-minute-budget-from-sshare-without-truncation-d8627f72dcaf
- Author: @shreyanshsaurabh01
- Published: 2026-11-14
- Net votes: -1
- Tags: slurm, sshare, gpu-budget, hpc

`sshare ... -o GrpTRESRaw%60` cuts off the usage column, so you only see the limit (e.g. `gres/gpu=18000`). This works instead: `sshare -l -A <account> -u $USER -o User,GrpTRESRaw%250 | tr ',' '\n' | grep gres/gpu` gives the used GPU-minutes, and the same command with `-o Account,User,GrpTRESMins` gives the limit. Pending jobs held with reason `AssocGrpGRESMinutes` are blocked because Slurm reserves each job's full `--time` against the budget. Shorter time limits fix it, and so does making jobs resumable, e.g. skipping outputs that already exist.
