`ezpz extend` / `ezpz update-job`⚓︎

Extend or retarget a queued job — bump its walltime, move it to another queue/partition, or recharge it to a different account — without resubmitting. Wraps IRI PUT /compute/job/{resource_id}/{job_id} (preferred) and qalter / scontrol update over SSH as fallback.

ezpz extend 8879044 --remote aurora --time 02:00:00
ezpz extend 8879044 --remote aurora --time 02:00:00 --queue prod --account AuroraGPT

update-job is an alias:

ezpz update-job 8879044 --remote aurora --time 02:00:00

Queued jobs only

The scheduler and the IRI API both reject updates to running jobs. Check ezpz status --remote <host> <job-id> first; if the state is active/running the update will fail.

Options⚓︎

Flag	Description
`JOB_ID` (positional, required)	Job to update
`-r`, `--remote HOST`	Required. Target host (`aurora`/`polaris`/`crux` or any SSH alias)
`-t`, `--time WALLTIME`	New walltime — `HH:MM:SS`, `MM:SS`, or seconds (e.g. `7200`)
`-q`, `--queue NAME`	Move to this queue / partition
`-A`, `--account NAME`	Charge this account / project
`--backend {auto,iri,ssh}`	Backend (default: `auto` — IRI-first on IRI hosts, else SSH)
`--dry-run`	Print what would be sent without sending

You must supply at least one of --time / --queue / --account; otherwise the command exits with a usage error.

Backend selection⚓︎

Same rule as ezpz submit --remote: auto tries IRI first on Aurora/Polaris/Crux and falls back to SSH only on transport/auth failures. Explicit failures (400/422, Unknown Job Id, not served by IRI) are never retried on the other backend — they mean "fix the request".

Examples⚓︎

Extend walltime⚓︎

ezpz extend 8879044 --remote aurora --time 04:00:00          # HH:MM:SS
ezpz extend 8879044 --remote aurora --time 7200              # seconds
ezpz extend 8879044 --remote aurora --time 30:00             # MM:SS

Retarget queue and account together⚓︎

ezpz extend 8879044 --remote aurora --queue prod --account AuroraGPT

Force the SSH path (qalter)⚓︎

ezpz extend 8879044 --remote aurora --queue prod --backend ssh

Preview⚓︎

ezpz extend 8879044 --remote aurora --time 02:00:00 --dry-run
# would PUT /api/v1/compute/job/{rid}/8879044 {duration:7200, queue:None, account:None} backend=auto

Output⚓︎

On success the updated job is printed:

8879044.aurora-pbs-0001 queued queue=prod (via iri)

On failure the API/scheduler message is printed on stderr and the process exits non-zero.

ezpz extend / ezpz update-job⚓︎