Skip to content

Cloud Ops overview

Owner: Cloud Ops

This guide deploys Arbium into a customer-owned cloud account and exposes one customer-reachable HTTPS endpoint. The application layer is the same on every supported cloud; only the managed Kubernetes, PostgreSQL, secret store, ingress, and workload-identity wiring differ.

  1. Prepare access and common prerequisites.
  2. Choose and provision a cloud: AWS, Google Cloud, or Azure.
  3. Configure and install the shared Helm chart, using the preset and provider overlay from Step 2.
  4. Verify the deployment with common health checks and the checks for your cloud.
  5. Hand the endpoint to IT.

The flow forks only for Step 2 and then converges again:

┌─ AWS: EKS + Aurora + Secrets Manager + ALB ─┐
Common prerequisites ────┼─ GCP: GKE + Cloud SQL + Secret Manager + LB ┼── Shared Helm ── Verify ── IT handoff
└─ Azure: AKS + PostgreSQL + Key Vault + nginx ┘
AWS Google Cloud Azure
Kubernetes Amazon EKS Google Kubernetes Engine Azure Kubernetes Service
PostgreSQL Aurora PostgreSQL Cloud SQL for PostgreSQL PostgreSQL Flexible Server
Secret store AWS Secrets Manager Google Secret Manager Azure Key Vault
Workload identity IRSA GKE Workload Identity AKS Workload Identity
Ingress and TLS AWS Load Balancer Controller + ACM GCE Ingress + Google-managed certificate ingress-nginx + cert-manager
Default embedder CPU or optional GPU pool CPU or optional GPU pool CPU; optional GPU pool can be added
Helm preset values-aws.yaml values-gcp.yaml values-azure.yaml

Do not mix provider overlays. The common application values file is portable, but the provider preset and provider values file must match the foundation that Terraform created.

IT can start Entra app registration while Cloud Ops provisions the foundation. Before Helm configuration, Cloud Ops needs the customer’s Entra tenant ID and Arbium Agent client ID. If the fleet console is enabled, Cloud Ops also needs the Arbium Admin application values and the SCIM group IDs.

Start with 1. Prerequisites & access.