Skip to main content

Respond to a failed job

When a job fails, your goal is to find out why, fix or account for it, and get the run moving again. The job's own output is the fastest path to the cause.

What this solves

A failed job stops the run. Without a clear path to the cause and the right action, it sits unresolved, or gets restarted before the real problem is fixed, so it just fails again.

Steps​

To respond to a failed job, complete the following steps:

  1. In Operations › Processes, open the failed job to see its detail.
  2. Read the job output and logs to find the cause. In the Summary tab's Completion section, check the Exit Code (also shown on the Output tab) and the Termination description. The exit code shows as a badge, green for 0 and red for any non-zero value. On a legacy LSAM job, also check the Agent Message row below Termination — it carries the reason the agent itself gave, which is often the actual answer when Termination holds only the exit code. See Agent Message.
  3. Check whether this is new. On the Status History tab, Job Execution History lists this job's earlier runs over a date range with each one's termination status and exit code, so you can see at once whether tonight is the first time or the fifth. A run's Notes tab carries what anyone wrote the last time — and is where to write what you find this time. See Earlier runs of the same job.
  4. Resolve the cause. Many failures are configuration or connection issues, so escalate to the Builder (job settings) or Administrator (connections, agents) as needed.
  5. Choose how to clear the job (see below).
note

The Exit Code is the code the command, script, or connector returned; it's a diagnostic detail, not the job's status. A failed job that reports no exit code shows a dash (—) rather than a 0, so "failed with code 0" reads differently from "failed, code unknown."

Check whether the job is retrying before you act on it

A job whose frequency configures failure retries does not read Failed while it still has attempts left. It waits in Wait Start Time, still showing the failed attempt's exit code and termination description, and its dependents keep waiting — so the run is not stalled and nothing downstream has reacted yet. The job's Summary tab shows Retry Count as used-of-maximum.

Let it take the attempts it has left if the cause might be transient. If it will not be — the failure is in the job's configuration, or a system it needs is down — Restart the job once that is resolved, or Cancel it, rather than waiting for the retries to run out. See Failure retries.

Your options on a failed job​

ActionUse it when
RestartThe cause is fixed and you want the job to run again. Reset Retry Count is selected by default, so the job runs with its full retry budget again; clear it to keep the count it had. For a UNIX Command, SAP R/3 Job or SAP BW Process Chain that reported its steps, pick a Restart Step to resume at the failed step rather than rerun the steps that completed — see Restart on a step. A z/OS Batch Job or z/OS Started Task offers a Start Step and an End Step, so you can rerun just a range of steps — see Restarting a z/OS job. When only some child jobs of an SAP R/3 Job failed, select them in its Child Processes table and use Restart selected — see Restarting only the failed child processes.
Mark FixedThe work was completed or resolved outside the system, and the job should count as done.
Under ReviewYou need to flag the job for follow-up without resolving it yet.
CancelThe job should not run again. A cancelled job satisfies no job dependency on it, so jobs that depend on it keep waiting until you skip, force-start or cancel them as well.
note

All four are available from a Failed job; Restart, Mark Fixed and Cancel are also available from a job Under Review. If an option is disabled in the job's action menu, or missing from its Status Update tab, the job isn't in a status that accepts it.

Before you restart​

Confirm the underlying problem is actually resolved. Otherwise the job will just fail again. If the failure is in the job's configuration, the Builder needs to change it before a restart will succeed.

When to escalate​

Escalate (with the job output, the job's status, and the workflow instance) when the failure isn't explained by the job's settings or a connection/agent issue, or when a job's status doesn't match what actually happened.

Related topics