ArgoCD & GitOps

ArgoCD is a Kubernetes-native continuous delivery tool that implements GitOps: a deployment model where a Git repository is the single source of truth for what should be running in a cluster, and a controller continuously reconciles the cluster’s actual state to match what’s declared in Git, rather than pushing deployments imperatively through a CI pipeline.

How It Works

ArgoCD runs inside a Kubernetes cluster and continuously monitors one or more Git repositories containing Kubernetes manifests (or Helm charts, Kustomize configurations, and similar). When it detects a difference between what’s declared in Git and what’s actually deployed in the cluster — someone merged a change updating an image tag, for example — ArgoCD automatically (or with manual approval, depending on configuration) applies that change to bring the cluster in line with Git.

Why GitOps Matters

This pull-based model, where the cluster reconciles itself against Git rather than an external CI pipeline pushing changes directly to the cluster, has a few concrete benefits: every change to production is captured in Git history with a clear author and commit message, providing a built-in audit trail; rolling back a bad deployment is as simple as reverting a Git commit, since the controller will reconcile the cluster back to the previous state automatically; and the cluster’s actual running state can be continuously verified against Git, catching configuration drift caused by manual kubectl changes that bypass the declared source of truth.

Where It Fits

ArgoCD is one of the most widely adopted GitOps tools alongside Flux, and it’s commonly paired with a separate CI pipeline that handles building and testing application code — CI produces a new container image and updates the image tag in the Git-tracked manifest, and ArgoCD takes over from there to actually deploy that change to the cluster.

ArgoCD’s Sync Strategies

ArgoCD supports both automated sync (where detected drift from the Git-declared state is reconciled immediately and automatically) and manual sync (where a human explicitly approves applying a detected change before it takes effect), letting teams choose the appropriate level of automation for different environments — automated sync is common for lower environments like staging where fast iteration matters more than caution, while manual sync approval is often preferred for production deployments where a human checkpoint before applying changes provides an additional safety net. This flexibility to dial automation up or down per environment, without changing the underlying GitOps model, is one of the practical reasons ArgoCD has become a standard choice for Kubernetes-native continuous delivery.