Getting changes live.
WHAT THE SEQUENCE USUALLY IS
Back up Deploy the artefact Install dependencies Apply migrations Rebuild caches Restart processes Verify
WHAT TO AUTOMATE
All of it.
WHY
Manual deployment is inconsistent, and inconsistency is where incidents originate.
WHAT TO NEVER DO
Edit files directly on the server.
WHY
The change is lost at the next deployment and exists nowhere else.
WHAT TO KEEP
The previous version, ready to restore.
WHAT TO PLAN BEFORE DEPLOYING
How to revert, and how quickly.
WHAT TO RESTART
Application processes, and queue workers.
WHY WORKERS SPECIFICALLY
They hold code in memory, and continue running the old version otherwise.
WHAT TO VERIFY AFTERWARDS
Key paths work The log is clean Scheduled work still runs Queue workers are processing
WHAT TO MONITOR
Error rates, for the following hour.
WHY
Problems appear under real traffic, not during verification.
WHAT TO SCHEDULE
Deployment when someone is available afterwards.
