Jobs

What Jobs is for

Jobs is the durable queue for work that can continue after you leave the page that started it. Imports, preparation, Signal materialization, Studies, exports, and other bounded operations can create Jobs.

The workstation header shows a compact view of active work from every page. Open System & Jobs -> Jobs when you need the full queue, an individual attempt, progress detail, telemetry, or a recovery action.

Read the queue

The queue shows each Job's name, kind, state, progress, runtime, last update, and current message. Filter by state when the history is long, and select a row to open Job Detail.

Common states include:

• queued — accepted and waiting for a worker;
• running — actively executing;
• paused — retained in the queue but not eligible to run;
• cancel requested — stopping at a safe boundary;
• canceled — stopped without producing a completed output;
• failed — ended with an error; and
• completed — finished successfully.

Progress is informative rather than a time estimate. A Job can spend different amounts of time in preparation, calculation, storage, or an external request. Use the phase and message with the percentage instead of assuming progress is linear.

Inspect one Job

Select a row to see its ID, kind, timestamps, runtime, attempt count, cancellation state, and available controls. Progress explains the current phase. Telemetry exposes the bounded runtime observations recorded for that Job.

Keep the Job ID when investigating a failure. It is the safest reference to include in an Activity search or support request.

Run and control queued work

When autorun is disabled, Run Next starts the next eligible queued Job. Run Job starts the selected queued or failed Job. Use the per-Job controls according to its state:

• pause a queued Job when it should remain retained but not start;
• resume a paused Job to make it eligible again;
• cancel queued, running, or paused work;
• restart a failed or canceled Job as a new attempt; and
• clear an eligible record only when its history is no longer needed.

Cancellation is cooperative. A running Job can remain active while its current safe unit completes and temporary work is cleaned up. Do not terminate the workstation simply because cancellation is not immediate.

Recover from failure

Read the final message and Progress detail before restarting. Correct the owning input, storage, credential, or compatibility problem first. Then restart the Job if the control is available, or return to the owning page and submit a new request.

Check whether an external request actually completed before repeating it. Submitting an uncertain data-provider or broker request twice can duplicate work or external charges.

A canceled or failed Job does not make an intermediate file a valid Arizmic object. Use only Datasets, Prepared Data, Study Results, exports, or other outputs that their owning page lists as complete and available.

Configure how Jobs run

Open System Settings to control autorun, desktop notifications, maximum active Jobs, and worker limits. Changes apply to future scheduling; they do not rewrite completed Results.