Rundeck job chain stopped partway

When a job chain stops at a failed job, rerunning that one job or skipping it is the quick way out, but a lone rerun leaves every job after it unrun, and forcing damaged data through spreads the damage downstream. 2501 finds the job that failed and why, fixes the underlying cause and reruns the whole Rundeck chain, within limits you set.

The job's error message says what broke, not why. Behind it may be a data problem: an input file that arrived cut short, or a record count that disagrees with the control file the upstream system sent.

What 2501 does

The Batch Scheduling specialty reads the failed run in Rundeck to find the job that stopped the chain, then looks past the error message to the underlying cause and fixes it. The chain is then rerun from its start, and the incident stays open until it completes. The report says which job failed, why, what was fixed and how each job did on the rerun, with the runs it looked at, so an operator can retrace the work in Rundeck.

You stay in control

• You choose the level: investigate only, act after a person confirms, or act on its own within your rules. The level is set per specialty, not per ticket.
• Business controls are never bent: a control file or check is never changed to make a job pass, and the upstream system's values stay the reference.
• No job is switched off to clear a failure: it never disables, unschedules or skips a job, never edits a chain to step around one, and never stops the scheduler itself.
• Your rules apply: which projects and chains it may rerun unattended, and which wait for a person.

Frequently asked

Why rerun the whole chain rather than the failed job? Because the jobs after the failure never ran. A single rerun can confirm a fix; only the whole chain finishes the work.

Can it just tell us why the chain stopped? Yes. In investigate-only mode it reports the failed job, the cause and the fix it would make, and reruns nothing.

Related

• Batch Scheduling
• ServiceNow incidents and change requests, worked end to end
• Failed CI run: flaky or real?
• Read-only AI investigation