A scheduled sync has one job: run at 02:00 and leave the data right. When it fails, someone has to know before the morning report goes out. A log file nobody opens doesn't count.
Two channels
| Channel | Best for |
|---|---|
| E-mail report | People: the DBA or the data owner reads what ran, how many rows, and any warning |
| Webhook | Systems: your monitoring, chat or incident tool receives a call when a job ends, with success or failure |
Set it once, override per job
- In Preferences, set the mail server (with a test e-mail button) and the default webhook URLs: one for
success, one for failure.
- In each job, tick "Notify a webhook" to use the defaults, or enter other URLs for that job only. They are
saved in the job file. A job without the option never calls a webhook.
What to alert on
- Failure, always. A failed nightly job should open a ticket or post in the team channel.
- Success, for critical jobs. "It ran" is information too: no message by 03:00 means the scheduler never
started it.
- Differences found by a status (comparison) job: route the report to the people who can explain them.
Keep it boring
- Use one webhook per environment, so staging noise doesn't wake up production on-call.
- Put the job name in your monitoring rule, not the URL; jobs change, rules should not.
More on scheduling jobs in database sync from the command line.
