Blue-Green Deployment
Subject: Computer and Technology noun /ˌbluːˈɡriːn dɪˈplɔɪ.mənt/ 1 definition
- software-deployment
- continuous-delivery
- high-availability
- cloud-computing
-
Definition (noun)
Entry #110 · UpdatedBlue-Green Deployment is a software release strategy that maintains two identical, isolated production environments—conventionally labeled "blue" and "green"—where only one serves live user traffic at any given time. While the active environment handles production requests, engineering teams deploy the new application release to the idle environment and run automated smoke and integration tests under real production conditions. Once verified, a load balancer, reverse proxy, or DNS change atomically switches incoming traffic to the updated environment, leaving the previous environment intact on standby for rapid rollback if defects surface. Unlike rolling updates or canary releases that run mixed application versions concurrently across shared nodes, blue-green deployment cleanly isolates full release stacks at the cost of temporarily provisioning duplicate compute capacity and requiring backward-compatible database migrations.
Common Forms
Load-Balancer Routing: Updates target groups or upstream pools on a Layer 4 or Layer 7 load balancer (such as AWS ALB or NGINX) to shift all production traffic instantaneously without waiting on client DNS caches.
DNS-Based Cutover: Reassigns domain records (such as weighted CNAME or alias entries) from the active environment to the standby stack, simplifying cross-region switching while introducing transition delays governed by record Time-to-Live (TTL) caching.
Container-Orchestrator Selector Switching: Maintains parallel blue and green workload sets (such as Kubernetes
Deploymentobjects) and flips aServicelabel selector or ingress backend rule to redirect traffic across pod sets within seconds.Database-Decoupled (Expand-and-Contract) Deployment: Pairs dual stateless application stacks with a shared database using backward-compatible schema migrations so both outgoing and incoming application versions can read and write safely during cutover and rollback.
Example
A payments engineering team serves
v2.4.0of its checkout API from a livebluecluster behind an Application Load Balancer while deployingv2.5.0to an identicalgreencluster and running synthetic payment tests against an internal preview endpoint. After verification passes, the CI/CD pipeline switches the production listener rule to route 100% of traffic togreen, and when an unexpected latency spike triggers an alert three minutes later, the on-call engineer reverts the listener rule back tobluein under ten seconds.