Deployability is an architectural quality: software should be safe, predictable, and repeatable to deliver. If a Tuesday afternoon release still depends on a fragile 40-step, 2:00 AM checklist, the system’s boundaries may not be as decoupled as they seem. section 13.12 section 13.2
Figure 13.1 The "ClickOps" Gauntlet: Manually deploying code without automated verification is a high-stakes, high-stress gamble that invites human error.

The ClickOps Gauntlet illustrates how manual deployment without automated verification invites human error. section 13.2
Traffic-routing strategies trade infrastructure cost against rollout and rollback risk:
- Rolling updates replace instances in batches, keeping costs lower but allowing old and new versions to coexist. section 13.12
- Blue-green deployments keep live and idle pools for fast cutovers and rollbacks, at a higher infrastructure cost. section 13.12
- Canary deployments route a small share of production requests, such as 2%, to limit the blast radius before wider rollout. section 13.12
Separate deployment—the act of placing code in an environment—from release, the decision to expose its behavior to users. Feature flags can keep unreleased logic hidden. Automated architectural fitness functions, such as those using NetArchTest or ArchUnit, can make CI gates reject package-boundary violations. section 13.12
Change a database schema in phases
Expand by adding new structures without changing the legacy schema; transition by writing both formats and migrating data; contract by removing deprecated structures once code and nodes are aligned. section 13.12

