GKE Autopilot Overview

Autopilot is a managed mode of operation in Google Kubernetes Engine (GKE) where Google manages your infrastructure configuration, including nodes, scaling, security, and other preconfigured settings.

What is Autopilot?

Autopilot mode is optimized to run most production workloads, and provisions compute resources based on your Kubernetes manifests. You can run an entire cluster in Autopilot mode, or run specific workloads in Autopilot mode within GKE Standard clusters.

Benefits

Key Benefits Summary

Benefit	Description
Focus on Apps	Google manages infrastructure; you build and deploy applications
Security	Default hardened configuration with automatic security patches
Pricing	Simplified billing forecasts and attribution
Node Management	No need to create nodes or configure upgrades
Scaling	Automatic node provisioning with HPA support
Resource Management	Pre-configured defaults if not specified
Networking	Network security features enabled by default
Release Management	Automatic enrollment in release channels

Container-Optimized Compute Platform

In GKE version 1.32.3-gke.1927002 and later, Autopilot includes a specialized container-optimized compute platform:

Platform Benefits

• Compute capacity matches workloads: Dynamically adjusts based on Pod count and resource consumption
• Fast scaling times: Dynamic resizing means Pods don't need to wait for new nodes

Pricing Model

Workload Type	Billing Model	Example
General-purpose Pods	Pod-based	Container-optimized platform, Balanced/Scale-Out ComputeClasses
Specific hardware Pods	Node-based	GPU workloads, specific machine series

Planning Your Clusters

Networking Options

Use Case	Solution
Expose apps to internet	Use Services
Internal communication	Configure Network Policies
Multi-cluster connectivity	Use Multi-cluster Services (MCS)
External HTTP(S) traffic	Use Ingress for Application Load Balancers
Internal traffic	Use Internal Load Balancers
Network security	GKE Dataplane V2 with network policies

Scaling Configuration

Key Points:

• Autopilot automatically scales nodes based on Pod count
• Clusters can scale down to zero nodes with no running workloads
• Use Horizontal Pod Autoscaling for automatic Pod scaling

Security Features

Autopilot clusters enable security best practices by default:

Deploying Workloads

Workload Deployment Options

Use Case	Solution
Control node properties	Create custom ComputeClass
Run Autopilot in Standard cluster	Use Autopilot ComputeClass
Run Arm workloads	Request Arm CPU machine series
AI/ML workloads	Request GPUs in manifest
Batch jobs at lower cost	Use Spot Pods or autopilot-spot ComputeClass
Minimal disruptions	Use safe-to-evict=false annotation
Pod bursting	Set limits higher than requests

Workload Separation

Pod Scheduling Controls

Feature	Description
Zonal Topology	Place Pods in specific Google Cloud zones
Pod Affinity	Colocate Pods on same nodes
Pod Anti-affinity	Spread Pods across nodes
Topology Spread	Control Pod distribution across topology domains

Managing and Monitoring

Version Upgrades

All Autopilot clusters are enrolled in a GKE release channel:

• Rapid: Latest features
• Regular: Balanced stability (default)
• Stable: Most tested versions

Monitoring

Autopilot clusters have built-in observability:

Type	Features
Logs	System logs, Workload logs, Audit logs
Metrics	System metrics, Workload metrics (via Prometheus)
Dashboards	GKE dashboard, Observability dashboard
Security	Security posture dashboard

Quick Start

# Create an Autopilot cluster
gcloud container clusters create-auto my-autopilot-cluster \
    --region=us-central1

# Get credentials
gcloud container clusters get-credentials my-autopilot-cluster \
    --region=us-central1

# Deploy a workload
kubectl create deployment hello-server \
    --image=us-docker.pkg.dev/google-samples/containers/gke/hello-app:1.0

# Expose the deployment
kubectl expose deployment hello-server \
    --type=LoadBalancer \
    --port=80 \
    --target-port=8080

---

Last updated: December 2025