Cloud Deploy Complete Guide: GitOps for GKE (2026)

Cloud Deploy is GCP's continuous delivery (CD) service that manages progressive releases (Canary / Blue-Green) to GKE, Cloud Run, Anthos, and more. Combined with Cloud Build (CI), it gives you a complete end-to-end CI/CD pipeline.

Core Concepts

Delivery Pipeline: Defines the Dev → Staging → Prod sequence
Target: The destination (GKE cluster, Cloud Run service, etc.)
Release: A release candidate for a specific version
Rollout: An execution unit that deploys a Release to a Target
Promotion: The action of advancing to the next Target
Phase: A canary stage (10% → 50% → 100%)

Sample Delivery Pipeline (clouddeploy.yaml)

apiVersion: deploy.cloud.google.com/v1
kind: DeliveryPipeline
metadata:
  name: my-app-pipeline
serialPipeline:
  stages:
  - targetId: dev
    profiles: [dev]
  - targetId: staging
    profiles: [staging]
    strategy:
      canary:
        runtimeConfig:
          kubernetes:
            serviceNetworking:
              service: my-app
        canaryDeployment:
          percentages: [25, 50, 75]
          verify: true
  - targetId: prod
    profiles: [prod]
    strategy:
      canary:
        canaryDeployment:
          percentages: [10, 50]
          verify: true
---
apiVersion: deploy.cloud.google.com/v1
kind: Target
metadata:
  name: prod
requireApproval: true
gke:
  cluster: projects/PROJECT/locations/asia-northeast1/clusters/prod-cluster

Deployment Strategies

Strategy	Description	Supported Targets
Standard	Single 100% cutover	All
Canary	Staged rollout by percentage	GKE / Cloud Run / Custom
Blue-Green	Switch the Service target	GKE
Custom	Your own shell script	Custom Target

Approval Gate

Set requireApproval: true on the Target
On promotion, send a Pub/Sub notification → Slack/email
Approve or reject from the GUI or CLI
Audit logs record who approved each promotion

Verify Phase

Run a test job via the Skaffold verify hook
Automate smoke tests at each canary stage
Automatic rollback on failure

Pricing

Item	Price
Delivery Pipeline	$50 / month
Or per-Release pricing	$0.50 per Release
Custom Target Job	Billed by Cloud Build time
Free tier	2 Delivery Pipelines / month

Comparison with Other CD Tools

Item	Cloud Deploy	ArgoCD	AWS CodeDeploy	Spinnaker
GitOps	—	◎	—	○
Native Canary	◎	Argo Rollouts (separate)	○	◎
Managed	◎	—	◎ (AWS)	—
GUI	◎	◎	○	◎
Pricing	Per Pipeline	Free	Free (within AWS)	Free / OSS

Typical Workflow

Push to GitHub
Cloud Build builds, tests, and pushes the image
Cloud Build runs gcloud deploy releases create to create a Release
Automatic deployment to the Dev Target (Verify)
Manual promotion to Staging (Approval), with Canary stages at 25/50/75
After clearing the Prod Approval Gate, Canary stages at 10/50
SLO monitoring via Cloud Monitoring, with automatic rollback on issues

What is the difference between Cloud Deploy and Cloud Build?

Cloud Build handles build and test (CI), while Cloud Deploy manages deployments (CD). Cloud Deploy takes the image built by Cloud Build and progressively releases it across multiple environments.

Which deployment targets are supported?

GKE, Cloud Run, GKE Enterprise (Anthos), and Custom Target (which lets you deploy anywhere via a shell script).

Does it support Canary and Blue-Green deployments?

Canary is supported natively (staged rollouts such as 10% → 50% → 100%). Blue-Green is implemented on GKE by switching Services.

How do Approval Gates work?

You can require approval before promoting to each target, with Slack/email notifications. Essential for production release governance.

How is the Verify phase implemented?

Use the Skaffold verify hook to run a test job. You can configure automatic rollback on failure.

What is the pricing model?

$50 per month per Delivery Pipeline, or $0.50 per Release. Target compute (Cloud Run, GKE, etc.) is billed separately.

How does it compare to AWS CodeDeploy and ArgoCD?

CodeDeploy is tightly integrated with AWS, ArgoCD is a flexible OSS GitOps tool, and Cloud Deploy offers deep GCP integration, native canary support, and an easy-to-use GUI.

What is the relationship with Skaffold?

Cloud Deploy uses Skaffold internally for rendering and deployment. You can fine-tune behavior through skaffold.yaml.

Check what you learned with practice questions

Practice with certification-focused question sets

Explore GCP exam prep

Author

NicheeLab Editorial Team

NicheeLab editorial team focused on data engineering and cloud certification learning. Content is structured around practical study needs and official exam domains.

Cloud Deploy Complete Guide: Canary, Blue-Green, and Progressive Deployment on GKE/Cloud Run