Deployment Strategies — Deep Dive

Level: Intermediate
Pre-reading: 09 · Deployment & Infrastructure

Strategy Comparison

Strategy Downtime Rollback Resource Usage Risk
Recreate Yes Slow Normal High
Rolling No Slow Normal + buffer Low
Blue-Green No Instant 2x Low
Canary No Fast Normal + small Very Low

Rolling Update

apiVersion: apps/v1
kind: Deployment
spec:
  replicas: 4
  strategy:
    type: RollingUpdate
    rollingUpdate:
      maxSurge: 1        # Extra pods during update
      maxUnavailable: 0  # Always maintain capacity

graph LR
    subgraph Rolling Update
        S1[v1 v1 v1 v1] --> S2[v1 v1 v1 v2]
        S2 --> S3[v1 v1 v2 v2]
        S3 --> S4[v1 v2 v2 v2]
        S4 --> S5[v2 v2 v2 v2]
    end

Blue-Green

Run two identical environments; switch traffic instantly.

graph LR
    LB[Load Balancer] -->|100%| B[Blue - v1]
    LB -.->|0%| G[Green - v2]

After verification:

graph LR
    LB[Load Balancer] -.->|0%| B[Blue - v1]
    LB -->|100%| G[Green - v2]

Implementation (K8s)

# Blue deployment
apiVersion: apps/v1
kind: Deployment
metadata:
  name: order-service-blue
  labels:
    version: blue
---
# Green deployment
apiVersion: apps/v1
kind: Deployment
metadata:
  name: order-service-green
  labels:
    version: green
---
# Service points to active version
apiVersion: v1
kind: Service
metadata:
  name: order-service
spec:
  selector:
    app: order-service
    version: blue  # Switch to green to flip

Canary

Route small percentage to new version; expand gradually.

graph LR
    LB[Load Balancer] -->|90%| V1[v1]
    LB -->|10%| V2[v2 canary]

Istio Canary

apiVersion: networking.istio.io/v1beta1
kind: VirtualService
metadata:
  name: order-service
spec:
  hosts:
    - order-service
  http:
    - route:
        - destination:
            host: order-service
            subset: v1
          weight: 90
        - destination:
            host: order-service
            subset: v2
          weight: 10

Feature Flags

Toggle features without deployment.

if (featureFlags.isEnabled("new-checkout-flow", userId)) {
    return newCheckoutService.process(order);
} else {
    return legacyCheckoutService.process(order);
}
Tool Type
LaunchDarkly SaaS
Unleash OSS
Split SaaS
Flagsmith OSS + SaaS

Rollback

Strategy Rollback Method
Rolling kubectl rollout undo deployment/order-service
Blue-Green Switch Service selector back
Canary Set canary weight to 0%
Feature Flag Toggle flag off
When should you use Blue-Green vs Canary?

Blue-Green: Instant switchover; good for database migrations that need all-or-nothing. Canary: Gradual rollout; catch issues with small user impact. Use canary for most deployments; blue-green when you need atomic switches.

How do you decide canary percentage and duration?

Start small (1-5%) for initial validation. Monitor key metrics. Increase gradually (10%, 25%, 50%, 100%) over hours/days. Duration depends on traffic volume — need enough requests for statistical significance. High-traffic: faster; low-traffic: slower.

What's the difference between deployment strategy and feature flags?

Deployment strategy controls which version runs. Feature flags control which code path executes within a version. Use deployment strategy for infrastructure changes; feature flags for business logic changes. Often used together.