Job States
Every Slurm job moves through a series of states from submission to completion. The current state is shown in the State column on the Jobs page.
Common states
| State | Meaning |
|---|---|
| PENDING | Waiting in the queue. The job has not started yet. See below for common reasons. |
| RUNNING | Executing on a compute node. |
| COMPLETING | The job has finished but Slurm is still cleaning up (collecting exit codes, releasing nodes). Transient — usually lasts a few seconds. |
| COMPLETED | Finished successfully (exit code 0). |
| FAILED | Finished with a non-zero exit code. Check the output and error files for details. |
| CANCELLED | Cancelled by the user or an administrator before or during execution. |
| TIMEOUT | Reached the time limit and was killed by Slurm. |
| OUT_OF_MEMORY | Exceeded the memory limit and was killed. Resubmit with a higher memory allocation. |
| NODE_FAIL | The allocated node failed during the job. Resubmit; consider using Exclude nodes to avoid the failed node. |
Why is my job PENDING?
A pending job is waiting for one of the following reasons:
- Resources — not enough free CPUs, memory, or nodes in the partition right now. The job will start automatically when resources become available.
- Priority — other jobs have higher fair-share priority. Your priority increases the longer you wait.
- QOS or account limits — you have reached a per-user or per-account job or CPU limit set by the QOS. Wait for a running job to finish.
- Time limit — your requested time limit exceeds the partition maximum. Lower the time limit or switch partitions.
Terminal states
COMPLETED, FAILED, CANCELLED, TIMEOUT, and OUT_OF_MEMORY are terminal — the job will not change state again. The detail panel stops polling once a terminal state is detected.