Monitor a scheduled GitHub Actions workflow.

One step at the end of the job. Four minutes, start to finish.

Before you start

You need a heartbeat in Upglory with its period and grace set to match the workflow's schedule. Its page shows the ping address; keep it in a repository secret called UPGLORY_PING_URL.

Settings that match this schedule

• Period24 hoursThe cron in the workflow's own schedule block, not the run time.
• Grace35 minutesThe longest run you have seen, plus the late start GitHub sometimes gives a scheduled workflow.
• Channelsemail, TelegramInherited from the workspace default. Change it on this heartbeat if it matters less than the others.

The line

Add one step at the end of the job. if: success() is the whole trick: the step is skipped when anything before it failed, so a failed run looks the same to us as a run that never started.

# the last step of the job
- name: Tell Upglory the job finished
  if: success()
  run: >
    curl -fsS -m 10 --retry 3
    "${{ secrets.UPGLORY_PING_URL }}"

The URL is unique to this heartbeat and is not a secret in the security sense: anyone holding it can only say "the job ran". Keep it in repository secrets anyway, so it does not appear in a public log. A scheduled workflow on GitHub can start late under load, so give the grace window room.

Measure the run time

Make it the first step of the job. The step at the end then records how long the run took, and a slow run is not mistaken for a missed one.

- name: Tell Upglory the job started
  run: curl -fsS -m 10 "${{ secrets.UPGLORY_PING_URL }}/start"

Report a failure

if: failure() runs this step only when an earlier one failed, so you hear at once instead of when the grace window runs out.

- name: Tell Upglory the job failed
  if: failure()
  run: curl -fsS -m 10 "${{ secrets.UPGLORY_PING_URL }}/fail"

Test it

Use Send a test ping on the heartbeat's page, or add workflow_dispatch to the workflow's triggers and run it by hand from the Actions tab. Nothing alerts until the first ping arrives, so a job you have not deployed yet never raises an alarm.

What you will see if it stops

Thirty-five minutes after the workflow should have finished, the heartbeat turns down and you get the alert with the last ping time in it. Check the Actions tab first: a workflow disabled after sixty days of no repository activity is the common cause, and it looks exactly like a job that stopped.

The other guides