ALERT_DAEMON

Action deployed successfully.

Homelab Infrastructure

PRODUCTION
Category: Infrastructure & Virtualization
Proxmox VELXCTailscalePrometheusGrafanaNginxNextcloud

// Infrastructure Path Flow

Tailscale Mesh NetworkNginx Reverse ProxyLXC Container EcosystemGPU Passthrough VM (LLM Inference)

Project Overview

This project focuses on the architecture and deployment of a self-hosted enterprise-grade homelab designed to consolidate local development, media, gaming, and AI exploration into a single, power-efficient hardware footprint.

The core motivation was to break away from restrictive cloud subscription models and public hosting limitations while solving two major engineering challenges: maximizing hardware utilization through bare-metal virtualization and securely accessing internal resources from anywhere without relying on a public IP address.

System Architecture & Flow

The infrastructure leverages a layered approach to separate edge networking, containerized core applications, and heavy compute workloads.

  1. Edge Networking & Ingress (Tailscale & Nginx): Inbound traffic bypasses strict ISP CGNAT limitations via an encrypted Tailscale mesh network. Remote administration requests are routed securely to an Nginx reverse proxy LXC container, which handles internal DNS mapping and SSL termination.
  2. Service Ecosystem (LXC Containers): Resource-efficient Linux Containers (LXCs) host isolated microservices. This includes a centralized MariaDB database, a Nextcloud/Samba network-attached storage pool, and dedicated game servers for Minecraft and Unturned.
  3. AI Inference Engine (VM & Hardware Passthrough): Heavy compute workloads are completely isolated inside a dedicated Virtual Machine. An AMD RX 580 GPU is directly mapped to this VM using PCIe hardware passthrough, bypassing hypervisor overhead to provide raw Vulkan-accelerated compute for local LLM inference.
  4. Observability Stack (Prometheus & Grafana): A continuous polling loop driven by Prometheus scrapes system-level metrics directly from the Proxmox hypervisor, container runtimes, and host hardware sensors, outputting real-time health data to Grafana dashboards.

Key Technical Specifications

Deployment & Usage

While the entire cluster is managed via the Proxmox VE web interface and automated via custom shell scripts, individual services or infrastructure components can be replicated or provisioned via CLI.

Sources available under public GPL v3.