CD

Continuous Delivery is the ship and run half of CI/CD. We do it with GitOps using ArgoCD: the cluster’s desired state lives in Git, and ArgoCD continuously makes the running cluster match it.

Git as the single source of truth

Nothing is deployed by hand. The desired state — Helm/Kustomize manifests and image tags — lives in a Git repository that CI updates. That gives you:

• A complete, auditable history of everything that has ever run.
• Reviewed changes — deploys go through pull requests, like any other code.
• Rollback by revert — going back to a known-good state is just reverting a commit.

How ArgoCD delivers

1. Watch — ArgoCD watches the Git repository for the desired state.
2. Sync — it applies that state to the cluster, automatically or on approval.
3. Detect drift — it continuously compares desired vs. live state.
4. Self-heal — out-of-band changes to the cluster are reverted back to what Git says.

We use the app-of-apps pattern so a single root application manages the whole platform’s set of deployments declaratively.

Git (desired state) ──► ArgoCD ──► Kubernetes
        ▲                  │
        └──── self-heal ◄──┘   (live drifts back toward Git)

Safe deploys & automated rollbacks

This is the half of the platform that delivers the “Safe CI/CD with automated rollbacks” promise:

• Progressive delivery (canary / blue-green) via Argo Rollouts, so new versions are shifted traffic gradually rather than all at once.
• Automated rollback — if health checks or observability signals fail during a rollout, ArgoCD/Rollouts reverts to the last healthy version automatically.
• Self-healing keeps the cluster honest between deploys.

The result is what the offer promises: deploys are boring and safe, and recovering from a bad one doesn’t depend on someone remembering the manual steps.

Representative stack

ArgoCD is our default GitOps engine. Progressive-delivery strategy (canary vs. blue-green) is chosen per workload during the Launch.