Prefer reversible changes.
Operational confidence grows when a change has a known path backward. Backups are useful; tested rollback procedures are better.
FIVE OPERATING PRINCIPLES
Complexity has to
earn its place.
Keep the public surface small.
Every unnecessary listener, dependency and exposed interface becomes another assumption the system has to defend.
Measure from outside.
Internal health is necessary but incomplete. A service should also be verified from the point where users actually meet it.
Make failure legible.
Systems will fail. The operational advantage comes from knowing where the failure is, what changed, and which action is safe next.
Boring is a feature.
Stable components and understandable configurations often outperform clever architecture over the lifetime of a service.