Continuous Deployment
Subject: Computer and Technology noun /kənˈtɪn.ju.əs dɪˈplɔɪ.mənt/ 1 definition
- software-engineering
- software-deployment
- release-automation
- continuous-delivery
-
Definition (noun)
Entry #122 · UpdatedContinuous Deployment (CD) is a software engineering practice in which every code change that passes an automated build, test, and verification pipeline is automatically released to production without human intervention. Building directly upon Continuous Integration (CI), it automates artifact promotion, infrastructure provisioning, rollout orchestration, and post-deployment health checks so verified commits reach end users within minutes of merging. Unlike Continuous Delivery—which keeps software in a deployable state while pausing at a manual approval gate before release—continuous deployment eliminates manual sign-offs entirely, relying on comprehensive test coverage, feature flags, and automated observability-driven rollbacks to maintain production stability.
Common Forms
Push-Based Pipeline Deployment: Executes deployment steps directly from a CI/CD runner once automated tests pass, pushing built artifacts, container images, or serverless packages to production targets through cloud or orchestrator APIs.
Pull-Based (GitOps) Deployment: Uses an in-environment controller (such as Argo CD or Flux) that continuously watches version-controlled manifests and automatically reconciles live production workloads whenever verified image digests or configurations update.
Progressive (Canary) Deployment: Routes a small slice of live production traffic to newly deployed instances first and automatically promotes or rolls back the rollout based on real-time latency, error-rate, and saturation telemetry.
Feature-Flagged Deployment: Decouples code release from feature activation by deploying verified commits continuously to production behind runtime configuration toggles, allowing instant exposure control without a new deployment cycle.
Example
When an engineer merges a bug fix for an e-commerce checkout service into the
mainbranch, the CD pipeline automatically builds a signed Docker image, passes 600 unit, integration, and contract tests, and updates the production Kubernetes manifest. Within four minutes, the cluster's deployment controller shifts 5% of live traffic to the new pods, verifies via Prometheus metrics that HTTP5xxerror rates remain unchanged, and completes the rollout to 100% of production instances with zero manual intervention.