Chapter 12B — CI/CD Release Strategies and Zero-Downtime Deployments

Typical gate sequence:

1. lint and formatting,
2. type checks,
3. unit and integration tests,
4. container build and scan,
5. migration safety checks,
6. staged deploy plus smoke tests.

Each gate should be deterministic and auditable.

Choose based on blast radius tolerance and operational maturity.

Expand-contract flow:

1. expand schema in backward-compatible way,
2. deploy app that can read old and new schema,
3. backfill data if needed,
4. switch reads/writes to new path,
5. contract old schema in later release.

Avoid single-release breaking schema flips in active systems.

Decision heuristic:

• rollback if issue is severe and previous release is known safe,
• forward-fix if rollback is unsafe due to irreversible schema or side effects,
• always document incident rationale.

Recovery speed often depends on pre-built runbooks and immutable artifacts.

Track release health signals:

• error-rate delta,
• latency delta (p95/p99),
• saturation indicators,
• failed auth spikes,
• business KPI impact.

Automatic rollback triggers should be tied to explicit thresholds.

Release hardening examples:

• pinned dependency lockfiles,
• image vulnerability scanning,
• signed artifacts,
• provenance and attestation checks,
• restricted deploy credentials.

Delivery speed without artifact trust invites high-impact incidents.