# Decision: VM vs LXC for Docker Swarm nodes

**Date:** 2026-01-10  
**Status:** Approved  
**Author:** Chester + FrankGPT

## Context

We need to run Docker Swarm manager and worker nodes on Proxmox VE hosts. Two options exist:

1. **QEMU/KVM Virtual Machines (VMs)**
2. **LXC Containers**

## Decision

**Use VMs for all Docker Swarm nodes.**

## Rationale

| Factor | VM | LXC |
| :--- | :--- | :--- |
| Docker support | Officially supported | Unsupported (requires hacks) |
| Stability | High | Medium (kernel updates can break) |
| Isolation | Full kernel isolation | Shared kernel |
| Resource overhead | Higher (~1-2 GB RAM baseline) | Lower (~256 MB baseline) |
| Maintenance | Standard Ubuntu updates | AppArmor/seccomp tuning required |

**Trade-off accepted:** We accept the higher resource overhead of VMs in exchange for stability and official Docker support.

## Specifications

| Parameter | Value |
| :--- | :--- |
| Base image | Ubuntu 24.04 LTS (Noble) cloud-init |
| Disk | 32 GB per VM |
| RAM | 4 GB per VM |
| vCPU | 2 per VM |
| Network bridge | `vmbr0` (bridged to LAN) |
| Storage pool | `local-lvm` |

## Capacity planning (per physical host)

- Physical NVMe: 512 GB
- Available in `local-lvm`: ~357 GB
- Initial allocation: 2 VMs × 32 GB = 64 GB
- Remaining: ~293 GB (room for 4+ additional VMs)

## References

- [community-scripts/ProxmoxVE docker-vm.sh](https://github.com/community-scripts/ProxmoxVE) — reference implementation
- Docker documentation on supported platforms