Part 1 · 1 chapters · ~8 min

Rolling, Recreate and Blue-Green

Recreate for when versions cannot coexist, rolling updates with surge and unavailability settings and readiness gates, blue-green with a full second environment and an instant switch, and the shared database that complicates every strategy.

Three replacement strategies

code

# Kubernetes rolling update: never below desired capacity, one extra pod at a time
spec:
  replicas: 10
  strategy: { type: RollingUpdate, rollingUpdate: { maxSurge: 1, maxUnavailable: 0 } }
  minReadySeconds: 20                      # a pod must stay ready 20 s before it counts
  template:
    spec:
      containers:
      - name: ledger
        readinessProbe: { httpGet: { path: /readyz, port: 8080 }, periodSeconds: 5 }
      terminationGracePeriodSeconds: 45    # longer than the app's drain time (Node course P7)

strategy	downtime	mixed versions	extra capacity	rollback speed
recreate	yes	no	none	another deploy
rolling	no	yes, during rollout	maxSurge	another rolling deploy
blue-green	no	no (at the LB)	100% during switch	seconds

ROLLING, RECREATE AND BLUE-GREEN

three ways to replace version 1 with version 2

swipe the figure sideways, or tap expand for full screen

1/5

recreate

Stop all v1, start all v2. Simple and guarantees no mixed versions, at the cost of downtime. Acceptable for batch workers or when two versions cannot coexist.

stop everything, start everythingdowntime, but never mixed versions