Problem
ssync status currently prints a verbose, human-readable block for every job. That is useful for debugging a single job, but it becomes noisy when monitoring many Slurm jobs or when an agent needs to summarize active work in a constrained context window.
In a recent workflow, checking a small batch of job IDs produced repeated full blocks with user, partition, resources, submit command, output/error paths, work dir, warnings, and runtime metadata for each job. Most of that was not needed for the common question: "which jobs are still running, pending, completed, failed, or cancelled?"
The verbosity also makes downstream automation harder because callers have to parse prose-style blocks instead of a compact schema.
Desired behavior
Add a compact status output that is easier for both humans and tools to consume.
Useful fields for the compact view:
job_idstatename or short job labelhostpartitionruntimereason for pending or failed jobs- optionally
submitted_at
- optionally output/stderr paths behind a verbose flag
Example output:
JOB_ID STATE HOST PARTITION RUNTIME REASON
4993854 R adastra mi300-shared 7:07 -
4993855 R adastra mi300-shared 7:07 -
4993856 R adastra mi300-shared 7:07 -
4993857 PD adastra mi300-shared 0:00 QOSGrpNodeLimit
4993881 PD adastra mi250-shared 0:00 Resources
A JSON mode would also be valuable:
ssync status --host adastra --job-id 4993854,4993855 --format json
with a stable list of job objects rather than formatted text.
Possible interface
Any of these would work:
ssync status --compactssync status --summaryssync status --format table|json|verbose- keep current output as
verbose for backwards compatibility
Why this matters
- Reduces unnecessary terminal/context output when polling many jobs.
- Makes agent and script workflows less brittle.
- Keeps the current detailed output available for debugging, while providing a better default or explicit mode for operational monitoring.
Problem
ssync statuscurrently prints a verbose, human-readable block for every job. That is useful for debugging a single job, but it becomes noisy when monitoring many Slurm jobs or when an agent needs to summarize active work in a constrained context window.In a recent workflow, checking a small batch of job IDs produced repeated full blocks with user, partition, resources, submit command, output/error paths, work dir, warnings, and runtime metadata for each job. Most of that was not needed for the common question: "which jobs are still running, pending, completed, failed, or cancelled?"
The verbosity also makes downstream automation harder because callers have to parse prose-style blocks instead of a compact schema.
Desired behavior
Add a compact status output that is easier for both humans and tools to consume.
Useful fields for the compact view:
job_idstatenameor short job labelhostpartitionruntimereasonfor pending or failed jobssubmitted_atExample output:
A JSON mode would also be valuable:
with a stable list of job objects rather than formatted text.
Possible interface
Any of these would work:
ssync status --compactssync status --summaryssync status --format table|json|verboseverbosefor backwards compatibilityWhy this matters