AutoSys Exit Code 15: Why Your Job Was Terminated (SIGTERM, term_run_time, KILLJOB)

You open the job history and see TERMINATED with exit code 15. The script's log just stops mid-line with no error. That's the signature of SIGTERM: the job didn't fail, it was stopped. This guide shows how to find what stopped it in a couple of commands, and how to fix each cause.

What Exit Code 15 Means

On Unix, signal 15 is SIGTERM, the polite "please stop" signal. When an AutoSys job's process is ended by a signal, the job shows that signal number as its exit code and usually goes TERMINATED rather than FAILURE. So exit code 15 tells you the job was stopped from outside, and that the error message you're looking for won't be in the job's own log.

# Last runs, with status and exit code
autorep -J FIN_EXTRACT -d

# The job's definition: look for term_run_time
autorep -J FIN_EXTRACT -q | grep -E 'term_run_time|max_run_alarm'

Cause 1: term_run_time Expired

term_run_time is the number of minutes AutoSys lets a job run before killing it. It's the most common cause of exit code 15, and its signature is easy to spot: the job is killed at the same elapsed time on every bad run. Jobs that were fine for months start hitting it as data volume grows.

update_job: FIN_EXTRACT
max_run_alarm: 75     /* alert at 75 min */
term_run_time: 120    /* kill only at 120 min */

Cause 2: Someone Sent KILLJOB

sendevent -E KILLJOB stops a running job, and the process typically sees SIGTERM. If the kill time doesn't match any limit, check the event history for a KILLJOB on that job, which also shows when it was sent. Operators often kill a job that looks hung without telling the owning team.

# Recent events for the job (look for KILLJOB)
autorep -J FIN_EXTRACT -d

# What a manual kill looks like
sendevent -E KILLJOB -J FIN_EXTRACT

Cause 3: The Agent or Host Went Down

When a server is rebooted or the AutoSys agent is restarted, running processes get SIGTERM. The clue is scale: several jobs on the same machine all end with exit code 15 within a minute or two. Check the agent log and the host's uptime, and agree maintenance windows with the batch schedule.

# On the agent machine
uptime
# Jobs that ran on this machine around the kill time
autorep -M finbatch01

Make the Script Survive a SIGTERM

Whatever the cause, a killed job can leave half-written files that the next run picks up. Trap SIGTERM in the script to clean up and exit with a clear message, so the log shows the job was stopped from outside.

#!/bin/bash
cleanup() {
  echo "$(date) received SIGTERM, removing partial output" >&2
  rm -f /data/out/extract.tmp
  exit 15
}
trap cleanup TERM

run_extract > /data/out/extract.tmp && mv /data/out/extract.tmp /data/out/extract.csv