In this third installment of my lab overhaul series, I’ll cover some of the most significant upgrades and migrations that I’ve implemented to enhance the performance, reliability, and versatility of my home lab environment. From flipping to Proxmox VE and leveraging TrueNAS with ZFS over iSCSI to migrating my smart home setup from HomeBridge to Home Assistant, these changes have not only increased the efficiency of my system but also opened up new possibilities for automation and control.
Let’s dive into the details of each of these upgrades:
1. Flip to Proxmox VE
The decision to move from VMware to Proxmox VE was driven primarily by changes going on with Broadcom and the price of a VMUG Advantage subscription to continue using vSphere in my lab. In my case, I’m using the community edition of Proxmox, meaning my cost has gone from several hundred dollars per year to zero.
a. TrueNAS and ZFS over iSCSI
One of the key components of my new Proxmox setup is TrueNAS, providing ZFS over iSCSI for high-performance storage. Using TrueNAS for file-sharing with ZFS gives me redundancy, scalability, and performance, with all the benefits of ZFS like snapshotting and efficient storage management.
For this setup, I’m leveraging the freenas-proxmox project, which allows seamless integration of TrueNAS storage with Proxmox via iSCSI. This solution ensures that I get the best of both worlds: reliable, high-performance storage with ZFS, while maintaining the flexibility of Proxmox for virtualization.
b. ProxLB: Load Balancer for Proxmox
For high-availability setups and VM load balancing, I’ve adopted ProxLB, a lightweight load balancer designed specifically for Proxmox. It’s an essential tool for my lab, as it allows me to load balance the VMs in my Proxmox cluster based on their RAM consumption, similar to functionality in vCenter that is not in Proxmox. In my case, I’m running the project inside a VM, but you can also run it on the hosts directly. You can read more about ProxLB and look over the code by visiting the GitHub page here.
2. Obico: AI Failed Print Detection for 3D Printing
I’ve integrated Obico for AI-based failed print detection on my 3D printer. Obico enables me to remotely monitor my 3D printer, detect failed prints in real-time, and receive instant alerts through Discord, email, and SMS. This setup has significantly reduced the need for constant manual checks and allows me to quickly respond to printing issues, ensuring my prints run smoothly.
Obico can operate in the cloud or on a self-hosted server. In my case, I am self-hosting the server using GPU passthrough on my main Proxmox host to present a Nvidia GeForce RTX 4070 Ti to a VM that runs multiple Docker containers for things such as media streaming and AI. You can learn more about Obico here.
3. PiKVM: Remote Access Powerhouse
For remote access and management of my Proxmox compute hosts, I do not have an out-of-band system like iDRAC, so I’ve deployed PiKVM on a Raspberry Pi 4B. This provides a KVM-over-IP solution, allowing me to manage my Proxmox and TrueNAS environments from anywhere. PiKVM connects seamlessly to my hardware, enabling me to control it remotely, even when the OS is not booted, thanks to a KVM switcher I incorporated which allows me to connect to 4 systems at once. All of this is housed in a 3D-printed 1U rackmount case to fit nicely into the back of my rack under my NVR.
4. AI VM with GPU Passthrough and Watchtower/Portainer
An exciting addition to my lab was setting up an AI VM, which features GPU passthrough for intensive computing tasks like deep learning. I’ve also integrated Watchtower for automatic container updates and Portainer for managing my Docker containers.
Watchtower and Portainer: These tools work together to automate the management of Docker containers, keeping my environment up-to-date and easily accessible via a web interface.
GPU Passthrough: Achieved through Proxmox’s PCIe passthrough capabilities, giving the VM direct access to a GPU for compute-heavy tasks.
Containers I run on this VM include:
- Ollama – For running llama 3.2 3B for Home Assistant voice pipelines
- Obico – The aforementioned AI failed print detection system for my 3D printer
- Spoolman – For managing and tracking filament usage for 3D printing
- Jellyfin – A late 2024/early 2025 project, as I decided to use the 4070 Ti’s horsepower for transcoding instead of my gaming desktop
- Wyoming Whisper TRT – A project I built through 2024 and early 2025 to accelerate speech-to-text on Nvidia GPUs and present it to Home Assistant using the Wyoming Protocol. You can read more about it in a dedicated blog post here.
- Wyoming GLaDOS – Another project I built, mainly by making a few code tweaks and updating Python packages for an existing project. This uses a custom voice model for text-to-speech to imitate the character GLaDOS from the Portal videogames, once again wrapped by the Wyoming protocol for consumption by Home Assistant. You can view the GitHub page here.
5. Migration from HomeBridge to Home Assistant
Switching from HomeBridge to Home Assistant was a significant step forward for my home automation. Home Assistant offers more flexibility, integration options, and local control. Here’s what I implemented in this transition:
a. Local TTS and ChatGPT/Llama3.1 Integration
I’ve integrated local speech-to-text (STT) and text-to-speech (TTS) functionality alongside an Ollama integration for enhanced AI-driven automation in my home. Now, Home Assistant can process voice commands without relying on cloud services. This project is still ongoing in 2025, but the plan in late 2025 or early 2026 is to replace Amazon Alexa throughout the house with this custom voice agent. So far, processing times are a bit slower than Alexa, sitting around 6 seconds versus 3-4, but all data is housed and processed locally, meaning I am not giving a company my data, plus it works even when my internet is down for most of the home. The biggest hurdles have been reducing processing time, which TensorRT for the STT and Llama 3.2 have helped with. As I go into mid-2025, I am hoping to reduce processing time on the TTS some more and potentially find a way of increasing the LLM performance, either with a different model or a different processing software. After that, it is a matter of further testing the Home Assistant Voice Preview Edition I purchased I received in February of 2025, or building my own voice satellites.
b. New Dashboards
The Home Assistant dashboards are now more personalized and feature a variety of widgets, including system stats, weather alerts, radar, and per-room smart home device controls.
c. Matter Fan Controls
I’ve implemented Matter over Thread fan switches from Inovelli to simplify fan control throughout the home. Fans can now be easily toggled and speed-controlled from mobile devices and voice assistants.
d. Water Meter and PowerWall Integrations
Integrating a Flume water meter sensor and my PowerWall into Home Assistant enables me to monitor water usage and energy usage in real time. As I continue to smartify the house, devices such as servers, plugs, and outlets allow me to see granular usage across the house to determine where most of my home draws occur. At some point in 2025, Inovelli will be releasing a firmware update allowing similar metrics on all my light and fan switches around the house.
e. Aqara Presence Sensors
With Aqara presence sensors in my master bedroom and garage, I can now automatically turn fans and lighting on and off as I enter and leave those rooms.
f. Meross Water Leak Sensors
Meross water leak sensors provide early warnings if there’s a leak in the home, helping to prevent damage and conserve water. Any time a water leak is detected, all Echos in the house announce it and a critical alert is sent to my phone. This can later be improved by adding a smart shutoff valve to my water main.
g. Meross HomeKit Garage Door Opener
I’ve integrated the Meross HomeKit garage door opener, enabling me to open and close the garage door via Home Assistant, Siri, or automations. I used to use a MyQ system but it is cloud-based and the APIs were heavily locked down. By presenting this device into Apple Home with the HomeKit bridge functionality and keeping a HomePod Mini at my home, I now get a handy widget to open and close the garage whenever I am near home on my CarPlay screen. I also have the garage door tied into various automations, such as turning on the garage lights automatically when it opens, and shutting the door automatically each night in case I go to bed forgetting to do so.
h. Automation Scripts
A variety of automations are now running seamlessly in the background. Examples include:
- Thermostat control based on whether Ring sensors detect doors and windows are open or closed in each HVAC zone.
- Auto-arming my security alarm at night and disarming it in the morning.
- Opening and closing blinds in the dining room and living room based on time of day or TV on/off status.
i. Ring Control of Sensors with MQTT Bridge
With the Ring MQTT bridge, I can now control Ring security sensors via Home Assistant. Fun fact: the MQTT bridge is maintained by Tom Sightler, a VP of Product Management at Veeam!
j. Govee Fridge/Freezer Sensors and Alerts
Using Govee thermometers, I now get real-time alerts about the temperature inside my appliances, ensuring that food is kept at optimal temperatures, since in 2024, I had two separate instances where my fridge stopped cooling while I was traveling.
k. Lab Power Control with UPS and PowerWall
With NUT and PowerWall integrations, I can now control the power usage in my lab. When the grid is down and PowerWall battery is getting low, I run a graceful shutdown script for my Proxmox and TrueNAS hosts using SSH. When grid power is restored and the UPS is fully charged, a combination of smart plugs and Wake -on-LAN/iDRAC commands power the hosts back on. Since my cameras, network switching, NVR, and firewall run off this same UPS, it significantly increases their run time in the event the grid is down long enough for the PowerWall to be exhausted.
7. Veeam 12.2 Support for Proxmox
Last summer’s release of Veeam Backup & Replication version 12.2 brought much-requested support for Proxmox VE. With its release, I performed my aforementioned migration from vSphere before my VMUG Advantage subscription came up for renewal in October.
Conclusion
The lab overhaul progressed well, with many of the core systems now in place and operational. From moving to Proxmox VE with TrueNAS for storage via freenas-proxmox to upgrading my smart home setup with Home Assistant, this is the next step toward a more automated, resilient, and scalable home lab environment. Stay tuned for the next part in the series, where I’ll go into some of the upgrades and changes I made in early 2025.
