Continuous Delivery and Release Management
What to know
Continuous delivery keeps software in a always-releasable state; continuous deployment goes further and releases automatically. Candidates must know deployment strategies: rolling, blue-green (two environments, instant switch and rollback), and canary (small traffic slice, progressive expansion). Understand feature flags as decoupling deployment from release, dark launches, A/B testing (experimentation, not risk reduction), and progressive delivery with automated analysis. Know rollback vs. roll-forward, database migration compatibility (expand-contract pattern), and release gates based on health metrics. Understand semantic versioning and changelog/audit expectations.
Traps
Blue-green is often misdescribed as requiring no database changes-schema changes must still be backward compatible during the switch. Canary analysis needs a baseline comparison and automated abort criteria; a canary without metrics is just a small rollout. Another trap conflates feature flags with deployment: flags let you release code dark and enable later, but stale flags become technical debt and must be cleaned up. Also, rollback of a database migration is often unsafe-prefer roll-forward with expand-contract.
Rule of thumb
Deployment is technical; release is a business decision. Separate them with flags, and make every change reversible or forward-fixable.
Worked example
A schema change renames a column. Sequence: (1) add the new column, dual-write; (2) backfill; (3) switch reads to the new column; (4) stop writing the old one; (5) drop it later. This expand-contract sequence keeps old and new app versions working simultaneously during a rolling deploy-a single-step rename breaks in-flight instances.