S3 — Bring Your Own Cluster (BYOC)
Your team already runs Kubernetes (EKS, AKS, GKE, or self-managed). This scenario skips AWS provisioning and just installs the OpenBox Helm chart onto the cluster you already operate.
When to pick S3
- You have a mature K8s platform team already running production workloads
- You want to skip cluster provisioning + IaC (your team already owns those decisions)
- You need to reuse existing shared infra (RDS, monitoring, secrets management)
When NOT to pick S3
- Your cluster is < Kubernetes 1.27 — upgrade first or pick a different scenario
- You need an air-gapped install → post-MVP scenario S4 (not yet documented)
- Your cluster is missing an ingress controller and you don't want to add one → pick S2 Fresh AWS instead
Assumed environment
- Kubernetes cluster ≥ 1.27 with
kubectlcontext configured - A default
StorageClasswithReadWriteOncevolumes - An Ingress controller installed (Istio / NGINX / ALB Controller / Traefik)
metrics-serverdeployed (verify viakubectl top nodes)- (Optional but recommended) cert-manager, External Secrets Operator, Argo Rollouts
Full checklist in Prerequisites checklist →.
Overview of the flow
- Cluster requirements → — versions, addons, capabilities
- Prerequisites checklist → — 10-item copy-paste-friendly verification
- Helm install → — chart deps + install command + common overrides
- Ingress & TLS → — pick Istio / NGINX / ALB and configure TLS
Cost estimate
Your cluster already exists — incremental OpenBox cost is ~$500-1,500/mo (mostly ML model-hosts + guardrails GPU if enabled). See System requirements → Cluster sizing.