Skip to content

Container Orchestration Strategy

Current State

Running: Atlantis, Cloudflare Tunnel (via Ansible + Podman + systemd)

Planned: Zitadel, monitoring stack, more internal tools

Decision: Nomad

Services Solution Complexity
1-5 Ansible + Podman/systemd Low
5-15 Nomad (single node) Medium
15-50 Nomad (cluster) Medium
50+ Kubernetes High

Why Nomad:

  • Lightweight (~50MB RAM vs K8s 1GB+)
  • Single binary, works on single node
  • Auto-healing, rolling deploys, health checks
  • HashiCorp ecosystem (aligns with OpenTofu)
  • Simple upgrade path to cluster when needed

Options Comparison

Aspect Ansible+Podman Nomad Kubernetes
Learning curve Days Weeks Months
Min resources Minimal ~50MB ~1GB
Auto-healing No (just systemd) Yes Yes
Rolling deploys Manual Built-in Built-in
Single node Yes Yes Awkward
HA cluster N/A 3+ nodes 3+ nodes 
Current: Ansible + Podman/systemd
    ↓ (at 5+ services)
Next: Nomad (single node)
    ↓ (if need HA)
Later: Nomad cluster
    ↓ (if 50+ services or client requires)
Maybe: Kubernetes

Trigger for Nomad: When managing 5+ services or Ansible playbooks become unwieldy.