open-cluster-management/development.md

Development Guide

This document provides comprehensive guidelines for new contributors developing the Open Cluster Management (OCM) project. It also serves as a reference for AI tools performing code reviews and analysis.

Table of Contents

• Project Overview
• Development Environment Setup
• Code Organization
• Development Workflow
• Code Standards and Style
• Testing Guidelines
• API Development
• Build and Release
• Troubleshooting

Project Overview

Open Cluster Management (OCM) is a CNCF sandbox project that provides multicluster and multicloud management capabilities for Kubernetes. The project consists of five core components:

• registration: Manages cluster registration and certificate lifecycle
• placement: Handles content placement decisions across clusters
• work: Distributes and applies manifests to managed clusters
• registration-operator: Deploys and manages OCM components
• addon-manager: Manages OCM add-ons lifecycle

Key Technologies

• Language: Go 1.24.0
• Framework: Kubernetes operators with controller-runtime
• Build System: Make with OpenShift build machinery
• Testing: Ginkgo/Gomega for BDD-style tests
• Packaging: Helm charts and OLM bundles

Development Environment Setup

Prerequisites

• Go 1.24.0 (installation guide)
• Docker or Podman (container engine)
• Kind (local Kubernetes clusters)
• kubectl (Kubernetes CLI)
• clusteradm (OCM CLI)

Quick Start

1. Create development cluster:

   kind create cluster --name ocm-dev
   

2. Initialize OCM control plane:

   # Replace <TAG> with a concrete release (e.g., vX.Y.Z)
   curl -fsSL https://raw.githubusercontent.com/open-cluster-management-io/clusteradm/<TAG>/hack/install.sh | bash
   clusteradm init --wait --bundle-version="<BUNDLE_VERSION>"
   

3. Join cluster as managed cluster:

   clusteradm join --hub-token <token> --hub-apiserver <url> --wait --cluster-name cluster1 --force-internal-endpoint-lookup --bundle-version="<BUNDLE_VERSION>"
   clusteradm accept --clusters cluster1
   

4. Verify installation:

   kubectl get clustermanager,managedcluster
   kubectl get pods -A | grep open-cluster-management
   

Code Organization

ocm/
├── cmd/                    # Main applications entry points
│   ├── addon/             # Addon manager
│   ├── placement/         # Placement controller
│   ├── registration/      # Registration controller
│   ├── registration-operator/ # Operator
│   ├── server/            # GRPC server
│   └── work/              # Work controller
├── pkg/                   # Core business logic
│   ├── addon/             # Addon framework and controllers
│   ├── operator/          # Operator controllers
│   ├── placement/         # Placement logic and controllers
│   ├── registration/      # Registration logic
│   └── work/              # Work distribution logic
├── manifests/             # Deployment manifests
├── deploy/                # Helm charts and OLM bundles
├── test/                  # Test suites
│   ├── e2e/              # End-to-end tests
│   └── integration/      # Integration tests
└── solutions/             # Usage examples and solutions

Key Packages

• pkg/common/: Shared utilities and helpers
• pkg/registration/hub/: Hub-side registration controllers
• pkg/registration/spoke/: Managed cluster registration logic
• pkg/work/hub/: Hub-side work distribution
• pkg/work/spoke/: Managed cluster work application
• pkg/placement/controllers/: Placement scheduling and decision logic
• pkg/addon/controllers/: Addon lifecycle management controllers

Development Workflow

1. Making Code Changes

# Fork and clone the repository
git clone https://github.com/<your-username>/ocm.git
cd ocm

# Create feature branch
git checkout -b feature/your-feature-name

# Make changes and commit with DCO sign-off
git commit -s -m "Add new feature: description"

2. Building Images

Build all images:

IMAGE_REGISTRY=quay.io/your-org IMAGE_TAG=dev make images

Build specific component:

IMAGE_REGISTRY=quay.io/your-org IMAGE_TAG=dev make image-registration

Load images to kind cluster:

kind load docker-image quay.io/your-org/registration:dev --name ocm-dev

3. Testing Changes

Update operator images:

kubectl set image -n open-cluster-management deployment/cluster-manager registration-operator=quay.io/your-org/registration-operator:dev

Update component images via CustomResource:

kubectl edit clustermanager cluster-manager  # For hub components
kubectl edit klusterlet klusterlet           # For spoke components

Code Standards and Style

Go Code Style

• Follow standard Go conventions: Use gofmt, go vet, and golangci-lint
• Variable naming: Use camelCase for local variables, PascalCase for exported items
• Comments: Write clear comments for exported functions and complex logic
• Interface design: Keep interfaces small and focused (interface segregation principle)

Import Organization

Imports must be organized in the following order:

1. Standard library packages
2. Third-party packages
3. OCM packages

import (
    // Standard library
    "context"
    "fmt"
    "time"

    // Third-party
    "github.com/spf13/cobra"
    "k8s.io/apimachinery/pkg/api/errors"
    "sigs.k8s.io/controller-runtime/pkg/client"

    // OCM packages
    "open-cluster-management.io/api/cluster/v1"
    "open-cluster-management.io/ocm/pkg/common/helpers"
)

Code Quality Requirements

All code must pass these checks:

make verify                      # Run all verification checks
make verify-gocilint            # Linting with golangci-lint
make verify-fmt-imports         # Import formatting verification
make verify-crds               # CRD consistency checks

Auto-format imports:

make fmt-imports                # Automatically format import statements

Naming Conventions

• Files: Use snake_case (e.g., cluster_controller.go)
• Packages: Use lowercase, single words when possible
• Constants: Use CamelCase (PascalCase for exported constants); reserve ALL_CAPS only when mirroring external specs
• Controllers: Suffix with Controller (e.g., PlacementController)
• Interfaces: Use descriptive names, often ending with -er (e.g., ClusterLister)

Documentation Standards

• All exported functions, types, and constants must have comments
• Comments should start with the name of the item being documented
• Use complete sentences and proper punctuation
• Include examples for complex APIs

Testing Guidelines

Unit Tests

make test-unit                    # Run all unit tests
go test ./pkg/registration/...    # Test specific package

Integration Tests

Integration tests use controller-runtime's envtest framework:

make test-integration                      # All integration tests
make test-registration-integration         # Registration only
make test-placement-integration           # Placement only
make test-work-integration               # Work only
make test-addon-integration              # Addon only

E2E Tests

End-to-end tests run against real clusters and require a kind cluster with proper setup:

# 1. Create kind cluster and set KUBECONFIG
kind get kubeconfig --name ocm-e2e > /tmp/kubeconfig-ocm-e2e
export KUBECONFIG=/tmp/kubeconfig-ocm-e2e

# 2. Run E2E tests with required variables
IMAGE_TAG=e2e KLUSTERLET_DEPLOY_MODE=Singleton make test-e2e

E2E test variables:

• IMAGE_TAG: Tag for container images (default: latest)
• KLUSTERLET_DEPLOY_MODE: Deployment mode (Default, Singleton, SingletonHosted)
• MANAGED_CLUSTER_NAME: Name of the managed cluster (default: cluster1)
• KUBECONFIG: Must be set to point to test cluster

Test Organization

• Unit tests: *_test.go files alongside source code
• Integration tests: test/integration/ directory
• E2E tests: test/e2e/ directory
• Test utilities: test/framework/ and pkg/*/testing/

Test Writing Guidelines

• Use Ginkgo/Gomega for BDD-style tests
• Write descriptive test names using Describe, Context, and It
• Use table-driven tests for multiple scenarios
• Mock external dependencies appropriately
• Ensure tests are deterministic and can run in parallel

API Development

Modifying APIs

OCM APIs are defined in the separate api repository. To modify APIs:

1. Clone both repositories:

   git clone https://github.com/<your-username>/api.git
   git clone https://github.com/<your-username>/ocm.git
   

2. Make API changes in the api repository and run:

   cd api
   make update  # Generates CRDs and DeepCopy methods
   

3. Use local API changes in ocm repository:

   cd ocm
   go mod edit -replace open-cluster-management.io/api=/path/to/local/api
   go mod tidy && go mod vendor
   make update  # Updates CRDs, Helm charts, and operator manifests
   

4. Remove replace directive before submitting PR

Update Process

The make update command handles:

• CRD copying from api repository
• Helm chart updates
• Operator manifest generation
• ClusterServiceVersion updates

Build and Release

Build System

The project uses OpenShift's build-machinery-go:

make all                        # Build all binaries
make images                     # Build all container images
make update                     # Update all generated manifests

Image Build

Images are built for multiple architectures:

• linux/amd64
• linux/arm64

Registry configuration:

export IMAGE_REGISTRY=quay.io/open-cluster-management
export IMAGE_TAG=v0.x.x

Available Make Targets

make build                      # Build binaries
make images                     # Build container images  
make test-unit                  # Run unit tests
make test-integration          # Run integration tests
make test-e2e                  # Run E2E tests
make verify                    # Run all verification checks
make update                    # Update generated files
make clean                     # Clean build artifacts

Troubleshooting

Common Issues

1. API Server Connection Issues:

   kubectl cluster-info          # Verify cluster connectivity
   kubectl get nodes            # Check cluster health
   

2. Image Pull Failures:

   kind load docker-image <image>  # Load to kind cluster
   kubectl describe pod <pod>      # Check image pull status
   

3. Certificate Issues:

   kubectl get csr               # Check certificate requests
   kubectl get secrets -A | grep tls  # Check TLS secrets
   

4. Operator Issues:

   kubectl logs -n open-cluster-management deployment/cluster-manager
   kubectl get clustermanager cluster-manager -o yaml
   

Debug Tools

• Enable debug logging: Set --v=4 or higher in component flags
• Use kubectl describe for resource status
• Check controller logs in open-cluster-management* namespaces
• Verify RBAC permissions with kubectl auth can-i

Performance Debugging

• Use go tool pprof for CPU/memory profiling
• Monitor metrics endpoints (when enabled)
• Check resource utilization with kubectl top

Contributing Guidelines

1. Submit an issue describing proposed changes
2. Fork and develop following this guide
3. Add tests for new functionality
4. Run quality checks: make verify test-unit
5. Submit pull request with DCO sign-off
6. Address review feedback promptly

For detailed contribution guidelines, see CONTRIBUTING.md.

Resources

• Project Website
• Community Slack
• API Documentation
• Addon Framework
• User Scenarios