Engineering note
Argo CD bootstrap: what installing the chart does, and what it does not
Installing Argo CD gives a team a controller. A usable delivery workflow also needs an application definition, repository access, promotion rules and an agreed recovery path. I keep those responsibilities separate when reviewing a platform.
- Terraform foundation
- Helm installation
- Argo CD controller
- Application configuration
- Reconciliation + recovery
Start with the checked-in evidence
The public EKS repository installs the Argo CD Helm chart from the add-ons Terraform root and configures an internal ALB ingress. The reviewed source does not include an app-of-apps repository or a complete staging-to-production promotion workflow. Those would be additional delivery configuration.
- Document which resources Terraform owns and which Argo CD owns.
- Avoid letting both tools reconcile the same application resources.
- Check the repository contents before describing a bootstrap as complete GitOps delivery.
Resolve access before exposing a dashboard
The example sets server.insecure for the Argo CD server. This needs an explicit TLS termination and authentication design; an internal ingress is not a substitute for either. Review certificate handling, SSO or account permissions, and which network paths can reach the service.
- Verify encryption at the intended boundary and restrict dashboard access.
- Limit repository credentials and deployment permissions to the required scope.
- Confirm that the team can recover access without relying on one person.
Test the failure path as well as the happy path
A useful acceptance test changes one service, observes reconciliation and checks what happens when health checks fail. Decide whether rollback means reverting Git, selecting a previous artifact or following an application-specific recovery procedure. Database changes may need their own plan.
- Use a small staging application with a known healthy revision.
- Observe a failed rollout and verify the documented recovery procedure.
- Record the starting conditions before comparing deployment durations.
A handover your team can use
I would hand over ownership boundaries, the application onboarding procedure and the recovery steps alongside the chart configuration. The repository is evidence of the bootstrap layer, not proof of a particular release frequency or recovery time.
Checks to run in your environment
Read-only checks after installation; application resources appear only after they are configured.
kubectl -n argocd get deployments
kubectl -n argocd get applicationsThese are suggested verification steps, not a record of a production test run. Check prerequisites and resource access before using them.
Source and references
Project links point to the exact source revision reviewed for this note.