gitops-principles

name

gitops-principles

description

Comprehensive GitOps methodology and principles skill for cloud-native operations. Use when (1) Designing GitOps architecture for Kubernetes deployments, (2) Implementing declarative infrastructure with Git as single source of truth, (3) Setting up continuous deployment pipelines with ArgoCD/Flux/Kargo, (4) Establishing branching strategies and repository structures, (5) Troubleshooting drift, sync failures, or reconciliation issues, (6) Evaluating GitOps tooling decisions, (7) Teaching or explaining GitOps concepts and best practices, (8) Deploying ArgoCD on Azure Arc-enabled Kubernetes or AKS with workload identity. Covers the 4 pillars of GitOps (OpenGitOps), patterns, anti-patterns, tooling ecosystem, Azure Arc integration, and operational guidance.

GitOps Principles Skill

Complete guide for implementing GitOps methodology in Kubernetes environments - the operational framework where Git is the single source of truth for declarative infrastructure and applications.

What is GitOps?

GitOps is a set of practices that uses Git repositories as the source of truth for defining the desired state of infrastructure and applications. An automated process ensures the production environment matches the state described in the repository.

The OpenGitOps Definition (CNCF)

GitOps is defined by four core principles established by the OpenGitOps project (part of CNCF):

Principle	Description
1. Declarative	The entire system must be described declaratively
2. Versioned and Immutable	Desired state is stored in a way that enforces immutability, versioning, and retention
3. Pulled Automatically	Software agents automatically pull desired state from the source
4. Continuously Reconciled	Agents continuously observe and attempt to apply desired state

Core Concepts Quick Reference

Git as Single Source of Truth

┌─────────────────────────────────────────────────────────────────┐
│                        GIT REPOSITORY                           │
│  (Single Source of Truth for Desired State)                    │
├─────────────────────────────────────────────────────────────────┤
│  manifests/                                                     │
│  ├── base/                    # Base configurations             │
│  │   ├── deployment.yaml                                        │
│  │   ├── service.yaml                                           │
│  │   └── kustomization.yaml                                     │
│  └── overlays/                # Environment-specific            │
│      ├── dev/                                                   │
│      ├── staging/                                               │
│      └── production/                                            │
└─────────────────────────────────────────────────────────────────┘
                              │
                              ▼ Pull (not Push)
┌─────────────────────────────────────────────────────────────────┐
│                      GITOPS CONTROLLER                          │
│  (ArgoCD / Flux / Kargo)                                       │
│  - Continuously watches Git repository                          │
│  - Compares desired state vs actual state                       │
│  - Reconciles differences automatically                         │
└─────────────────────────────────────────────────────────────────┘
                              │
                              ▼ Apply
┌─────────────────────────────────────────────────────────────────┐
│                    KUBERNETES CLUSTER                           │
│  (Actual State / Runtime Environment)                          │
└─────────────────────────────────────────────────────────────────┘

Push vs Pull Model

Push Model (Traditional CI/CD)	Pull Model (GitOps)
CI system pushes changes to cluster	Agent pulls changes from Git
Requires cluster credentials in CI	Credentials stay within cluster
Point-in-time deployment	Continuous reconciliation
Drift goes undetected	Drift automatically corrected
Manual rollback process	Rollback = git revert

Key GitOps Benefits

1. Auditability: Git history = deployment history
2. Security: No external access to cluster required
3. Reliability: Automated drift correction
4. Speed: Deploy via PR merge
5. Rollback: Simple git revert
6. Disaster Recovery: Redeploy entire cluster from Git

Repository Strategies

Monorepo vs Polyrepo

Monorepo (Single repository for all environments):

gitops-repo/
├── apps/
│   ├── app-a/
│   │   ├── base/
│   │   └── overlays/
│   │       ├── dev/
│   │       ├── staging/
│   │       └── prod/
│   └── app-b/
└── infrastructure/
    ├── monitoring/
    └── networking/

Polyrepo (Separate repositories):

# Repository per concern
app-a-config/          # App A manifests
app-b-config/          # App B manifests
infrastructure/        # Shared infrastructure
cluster-bootstrap/     # Cluster setup

Multi-Repository Pattern (This Project)

Separates infrastructure from values for security boundaries:

infra-team/                    # Base configurations, ApplicationSets
├── applications/              # ArgoCD Application definitions
└── helm-base-values/          # Default Helm values

argo-cd-helm-values/           # Environment-specific overrides
├── dev/                       # Development values
├── stg/                       # Staging values
└── prd/                       # Production values

Benefits:

• Different access controls per repo
• Separation of concerns
• Environment-specific secrets isolated

Branching Strategies

Environment Branches

main ────────────────────────────────────► Production
  │
  └──► staging ──────────────────────────► Staging cluster
         │
         └──► develop ───────────────────► Development cluster

Trunk-Based with Overlays (Recommended)

main ────────────────────────────────────► All environments
  │
  ├── overlays/dev/       → Dev cluster
  ├── overlays/staging/   → Staging cluster
  └── overlays/prod/      → Prod cluster

Release Branches

main
  │
  ├── release/v1.0 ──────► Production (v1.0)
  ├── release/v1.1 ──────► Production (v1.1)
  └── release/v2.0 ──────► Production (v2.0)

Sync Policies and Strategies

Automated Sync

syncPolicy:
  automated:
    prune: true       # Delete resources not in Git
    selfHeal: true    # Revert manual changes

Manual Sync (Production Recommended)

syncPolicy:
  automated: null     # Require explicit sync

Sync Options

Option	Use Case
CreateNamespace=true	Auto-create missing namespaces
PruneLast=true	Delete after successful sync
ServerSideApply=true	Handle large CRDs
ApplyOutOfSyncOnly=true	Performance optimization
Replace=true	Force resource replacement

Declarative Configuration Patterns

Kustomize Pattern

# base/kustomization.yaml
apiVersion: kustomize.config.k8s.io/v1beta1
kind: Kustomization
resources:
  - deployment.yaml
  - service.yaml

# overlays/prod/kustomization.yaml
apiVersion: kustomize.config.k8s.io/v1beta1
kind: Kustomization
resources:
  - ../../base
patchesStrategicMerge:
  - replica-patch.yaml
images:
  - name: myapp
    newTag: v1.2.3

Helm Pattern

# Application pointing to Helm chart
spec:
  source:
    repoURL: https://charts.example.com
    chart: my-app
    targetRevision: 1.2.3
    helm:
      releaseName: my-app
      valueFiles:
        - values.yaml
        - values-prod.yaml

Multi-Source Pattern

spec:
  sources:
    - repoURL: https://charts.bitnami.com/bitnami
      chart: nginx
      targetRevision: 15.0.0
      helm:
        valueFiles:
          - $values/nginx/values-prod.yaml
    - repoURL: https://github.com/org/values.git
      targetRevision: main
      ref: values

Progressive Delivery Integration

GitOps enables progressive delivery patterns:

Blue-Green Deployments

# Two applications, traffic shift via Ingress/Service
apiVersion: argoproj.io/v1alpha1
kind: Application
metadata:
  name: app-blue
---
apiVersion: argoproj.io/v1alpha1
kind: Application
metadata:
  name: app-green

Canary with Argo Rollouts

apiVersion: argoproj.io/v1alpha1
kind: Rollout
spec:
  strategy:
    canary:
      steps:
        - setWeight: 10
        - pause: {duration: 5m}
        - setWeight: 50
        - pause: {duration: 10m}

Environment Promotion (Kargo)

Warehouse → Dev Stage → Staging Stage → Production Stage
    │           │              │               │
    └── Freight promotion through environments ───┘

Cloud Provider Integration

Azure Arc-enabled Kubernetes & AKS

Azure provides a managed ArgoCD experience through the Microsoft.ArgoCD cluster extension:

# Simple installation (single node)
az k8s-extension create \
  --resource-group <rg> --cluster-name <cluster> \
  --cluster-type managedClusters \
  --name argocd \
  --extension-type Microsoft.ArgoCD \
  --release-train preview \
  --config deployWithHighAvailability=false

# Production with workload identity (recommended)
# Use Bicep template - see references/azure-arc-integration.md

Key Benefits:

Feature	Description
Managed Installation	Azure handles deployment and upgrades
Workload Identity	Azure AD authentication without secrets
Multi-Cluster	Consistent GitOps across hybrid environments
Azure Integration	Native ACR, Key Vault, Azure AD support

Prerequisites:

• Azure Arc-connected cluster OR MSI-based AKS cluster
• Microsoft.KubernetesConfiguration provider registered
• k8s-extension CLI extension installed

See references/azure-arc-integration.md for complete setup guide.

---

Security Considerations

Secrets Management

Never store secrets in Git! Use:

Approach	Tool
External Secrets	External Secrets Operator
Sealed Secrets	Bitnami Sealed Secrets
SOPS	Mozilla SOPS encryption
Vault	HashiCorp Vault + CSI
Cloud KMS	AWS/Azure/GCP Key Management

RBAC Best Practices

# Limit ArgoCD to specific namespaces
apiVersion: argoproj.io/v1alpha1
kind: AppProject
spec:
  destinations:
    - namespace: 'team-a-*'
      server: https://kubernetes.default.svc
  sourceRepos:
    - 'https://github.com/org/team-a-*'

Network Policies

• GitOps controller should be only component with Git access
• Restrict egress from application namespaces
• Use network policies to isolate environments

Observability and Debugging

Health Status Interpretation

Status	Meaning	Action
Healthy	All resources running	None
Progressing	Deployment in progress	Wait
Degraded	Health check failed	Investigate
Suspended	Manually paused	Resume when ready
Missing	Resource not found	Check manifests

Common Issues Checklist

1. Sync Failed: Check YAML syntax, RBAC permissions
2. OutOfSync: Compare diff, check ignoreDifferences
3. Degraded: Check Pod logs, resource limits
4. Missing: Verify namespace, check pruning settings

Drift Detection

# Check application diff
argocd app diff myapp

# Force refresh from Git
argocd app get myapp --refresh

Quick Decision Guide

When to Use GitOps

• Kubernetes-native workloads
• Multiple environments (dev/staging/prod)
• Need audit trail for deployments
• Team collaboration on infrastructure
• Disaster recovery requirements

When GitOps May Not Fit

• Rapidly changing development environments
• Legacy systems without declarative configs
• Real-time configuration changes required
• Single developer, single environment

References

For detailed information, see:

• references/core-principles.md - Deep dive into the 4 pillars
• references/patterns-and-practices.md - Branching and repo patterns
• references/tooling-ecosystem.md - ArgoCD vs Flux vs Kargo
• references/anti-patterns.md - Common mistakes to avoid
• references/troubleshooting.md - Debugging guide
• references/azure-arc-integration.md - Azure Arc & AKS GitOps setup

Templates

Ready-to-use templates in templates/:

• application.yaml - ArgoCD Application example
• applicationset.yaml - Multi-cluster deployment
• kustomization.yaml - Kustomize overlay structure

Scripts

Utility scripts in scripts/:

• gitops-health-check.sh - Validate GitOps setup

External Resources

• OpenGitOps Principles
• ArgoCD Documentation
• Flux Documentation
• Kargo Documentation
• GitOps Working Group
• Azure Arc GitOps with ArgoCD
• Azure Arc-enabled Kubernetes

---

Gotchas

• selfHeal: true fights kubectl edit: Engineers patching live resources during incident response will see their changes reverted within seconds. Either pause auto-sync first or commit the fix to Git.
• Monorepo + ApplicationSet git generator scales the controller hard: Every directory change re-renders every app. A 200-app repo with a noisy main branch can pin controller CPU. Split repos or use webhooks instead of polling.
• PruneLast=true and Helm hooks collide: Hooks run during sync; prune happens after. Resources from post-install hooks get deleted on next sync because they're not in the rendered manifest. Annotate hooks with Prune=false.
• Drift correction depends on the controller noticing: Resources outside ArgoCD's tracked set (e.g. dynamic Secrets from External Secrets Operator) are NOT reconciled — drift in them is invisible. Use explicit ignoreDifferences.
• Sealed Secrets are cluster-scoped: A SealedSecret encrypted for cluster A will not decrypt on cluster B. Multi-cluster GitOps needs per-cluster keys or a shared key (which defeats the security model).
• git revert rollback assumes immutable image tags: Reverting a manifest reverts the tag string, but a mutable tag (e.g. :latest) now points at a different image. Pin to digests for true rollback safety.