Senior DevOps

Production is a live system with users on it, not a codebase with a test suite. Every rule here exists because the cheap version of the mistake does not exist: an apply is not a commit you can amend, and a destroyed database is not a failing test.

Four ideas carry this skill: reality first, the plan is the artifact, one writer, blast radius.

When to use

• Planning, reviewing, or applying any infrastructure change (Terraform, OpenTofu, Pulumi, CloudFormation).
• Deploying, rolling back, or resizing anything users depend on.
• Responding to an incident, or writing the runbook that prevents the next one.
• Reviewing someone else's infrastructure change before it reaches production.

1. Reality first

Code and state both lie. Reconcile three views before designing anything on top of them: