Making routine work repeatable.
WHAT TO AUTOMATE
Backups Log rotation Certificate renewal Updates, where safe Monitoring checks Cleanup of temporary files
WHAT TO NEVER AUTOMATE WITHOUT REVIEW
Deletion of anything you cannot recreate Changes to firewall rules that could lock you out Anything affecting live traffic during business hours
WHAT EVERY AUTOMATED TASK NEEDS
A log An indication of success or failure Alerting when it fails
WHY THE LAST ONE
A backup that stopped running six months ago is discovered when it is needed.
WHAT TO MONITOR
That the task ran That its output is plausible
WHAT TO BUILD IN
Idempotency, so running twice is harmless A lock, so two copies do not overlap A dry-run mode
WHAT TO TEST
The failure path, by making it fail.
WHAT TO KEEP IN VERSION CONTROL
Every script and configuration.
WHY
Servers are rebuilt, and undocumented changes are lost.
WHAT TO DOCUMENT
What runs, when, and what it touches.
WHAT TO REVIEW PERIODICALLY
Whether each automated task is still needed.
WHAT TO REMOVE
Tasks whose purpose nobody remembers.
