Home Lab Infrastructure ↗
↳ View Home Lab Infrastructure case study →A continuously evolving environment where I operate real services and turn classroom concepts into practical experience across virtualization, storage, automation, monitoring, networking, and systems administration.
Hermes Agent extends the environment with a self-hosted AI operations layer for read-only-first investigation, infrastructure documentation, and approval-gated administrative workflows.
Proxmox Home Lab
A Proxmox VE environment hosting isolated virtual machines and Linux containers for DNS, monitoring, media, dashboards, and secure ingress services.
TrueNAS Storage
Bulk storage built around eight hard drives in a RAIDZ3 layout, with attention to redundancy, permissions, file access, and future backup improvements.
Infrastructure as Code
I use Terraform to define and provision Proxmox workloads as code, then Ansible from Alma-MGMT to apply baseline configuration, patch Linux systems, validate services, run health checks, coordinate backups, and perform controlled reboots. Inventory, playbooks, and infrastructure changes are version-controlled in Git so recurring maintenance stays repeatable and auditable.
Monitoring Stack
Prometheus and Grafana monitor infrastructure and K3s health, with alerting for the services that matter most. Uptime Kuma tracks availability, while Homepage brings system and service status into one internal dashboard.
DNS & Secure Access
Technitium DNS provides internal resolution and network-level filtering. Nginx handles reverse proxying and service failover, while Cloudflare Tunnel provides controlled external access without directly exposing internal services.
Self-Hosted Applications
Jellyfin, Immich, and internal dashboards run as managed services across Linux VMs, containers, and Kubernetes, giving me hands-on experience with deployment, permissions, updates, monitoring, and troubleshooting.
Hermes Agent
A self-hosted Hermes Agent runs on a dedicated Ubuntu Server VM with a cloud-hosted language model, private Discord access, and a self-hosted Honcho memory stack. A dedicated SSH identity connects it to Alma-MGMT for read-only-first investigation, weekly maintenance checks, service validation, security review, troubleshooting, and approval-gated infrastructure actions.
Jellyfin K3s Cluster
Jellyfin runs as a single active workload across a three-node K3s server/control-plane/etcd cluster. Longhorn keeps two replicas of application state, TrueNAS supplies shared media, and Nginx routes traffic between primary and failover nodes. I tested controlled transfers and automatic failover so Jellyfin can recover on another node without rebuilding its configuration.