Course contentsShow
← Cloud and Infrastructure
Lesson 1167 of 2,98923. Kubernetes WorkloadsFree lesson

Rolling Update Strategy

Configure rolling update behavior with maxSurge and maxUnavailable to control update speed and availability.

Rolling Update Strategy

What you'll learn: You'll learn how to fine-tune rolling updates using maxSurge and maxUnavailable to balance update speed with service availability.

The Core Idea

When you update a Deployment, Kubernetes doesn't immediately shut down all old Pods and start new ones—that would cause downtime. Instead, it gradually replaces them using a rolling update strategy. Two key parameters control this behavior:

  • maxSurge: How many extra Pods above the desired replica count can exist temporarily during the update
  • maxUnavailable: How many Pods can be unavailable during the update

Think of it like renovating a restaurant: maxSurge is how many extra tables you can temporarily squeeze in while replacing old ones, and maxUnavailable is how many tables can be out of service at once. You balance customer capacity with renovation speed.

How They Work Together

Both settings can be absolute numbers (e.g., 2) or percentages (e.g., 25%). Here's what happens:

  • Higher maxSurge: More new Pods spin up before old ones terminate → faster updates, more resources used temporarily
  • Lower maxUnavailable: Fewer Pods go down at once → better availability, slower updates
  • Aggressive update: maxSurge: 50%, maxUnavailable: 0 (never lose capacity)
  • Conservative update: maxSurge: 1, maxUnavailable: 0 (one-by-one replacement)
  • Balanced: maxSurge: 25%, maxUnavailable: 25% (Kubernetes default)

Configuration Example

spec:
  replicas: 4
  strategy:
    type: RollingUpdate
    rollingUpdate:
      maxSurge: 1
      maxUnavailable: 1

This allows up to 5 Pods total (4 + maxSurge) and tolerates 3 available (4 - maxUnavailable) during updates.

Key Takeaway: maxSurge and maxUnavailable give you precise control over how fast your Deployment updates roll out and how much capacity fluctuation you'll tolerate—tune them based on your application's availability requirements and resource constraints.