operational-hygiene

The Only Force That Defeats Cloud Entropy Is Enforced Discipline

Cloud infrastructure degrades through entropy. Every manual change, every forgotten resource, every untagged instance, every disabled alarm that was never re-enabled -- these are small acts of disorder that compound into large, expensive, ungovernable messes. Nobody wakes up one morning with a $40,000 surprise bill. They get there through twelve months of "we'll clean that up later."

Operational hygiene is not a project. It is a daily practice. It is the cloud infrastructure equivalent of washing dishes after every meal instead of letting them pile up for a week. The five pillars -- clean as you go, cost attribution, monitoring, drift detection, and lifecycle policies -- form a system where each reinforces the others.

Pillar 1: Clean As You Go

The most expensive cloud resources are the ones nobody remembers creating. After every migration, every experiment, every proof-of-concept, clean up immediately. Not next sprint. Not after the launch. Now.

The Rule

If a resource has served its purpose, delete it in the same week. Temporary resources that survive longer than one sprint become permanent. Permanent resources that nobody owns become liabilities.

Good vs. Bad Patterns

Bad: "We'll clean up after the migration"

Week 1:  Migrate service-A from old cluster to new cluster
Week 4:  Migrate service-B
Week 8:  Migrate service-C
Week 12: "We should clean up the old cluster"
Week 20: Old cluster is still running, costing $2,400/month
Week 52: Nobody remembers what the old cluster does. Too risky to delete.

Good: Clean up is part of the migration ticket

Ticket: Migrate service-A to new cluster
  Subtask 1: Deploy service-A on new cluster
  Subtask 2: Reroute traffic to new cluster
  Subtask 3: Verify new deployment (48h monitoring)
  Subtask 4: Delete old service-A resources    <-- same ticket
  Subtask 5: Verify old resources are gone     <-- same ticket

Common Cleanup Targets