Kustomize Base Overlays Declarative Configuration Gke

What the Base and Overlay Architecture Is

In Kustomize, the base and overlay pattern is the standard declarative configuration pattern for managing Kubernetes manifests across multiple targets. The base/ directory contains standard, environment-agnostic Kubernetes manifests (such as Deployments, Services, and ServiceAccounts) declared inside a kustomization.yaml file. The overlays/ directory contains subdirectories for each deployment target (e.g., dev, staging, prod), each containing its own kustomization.yaml referencing the base.

How It Meets the Requirements

• Single Centralized Base: Common workload configurations are defined once in the base/ directory. Updates to shared architecture (such as adding a container port or updating a health check) occur in one place without duplication.
• Environment-Specific Customization: Overlays modify base manifests using declarative strategic merge patches, JSON 6902 patches, and built-in transformers (namespace, namePrefix, commonLabels, replicas, and resources).
• GitOps and CI/CD Compatibility: Tools like Google Cloud Deploy and Skaffold natively understand Kustomize overlays, generating fully rendered YAML artifacts per target environment during pipeline execution.

Key Advantages and Capabilities

• DRY (Don't Repeat Yourself) Principles: Eliminates redundant copy-pasted YAML across environments.
• Template-Free Customization: Preserves native, valid Kubernetes YAML specifications without template delimiters ({{ .Values }}).
• Granular Patching: Allows precise targeting of fields (such as scaling replica counts or tuning memory limits for production) without altering core specs.
• Auditable Declarative State: Ensures that the final rendered state is predictable and reproducible via kustomize build.

Why This Is Optimal

This directory structure represents the core architectural design pattern of Kustomize. It cleanly isolates shared application architecture from environment-specific configuration drift while providing continuous delivery engines like Cloud Deploy with deterministic, reviewable rendering boundaries.