I found an Excellent Raspberry Pi Replacement for Home Assistant / IOTstack (incl. Proxmox)

Video Statistics and Information

Video
Captions Word Cloud
Reddit Comments
Captions
Raspberry Pi boards are hard to get, probably also next year. So we have to have a strategy to survive without new Raspberry boards. Sometimes, a problem can be a chance. This is what we will see today. Stay tuned if you want to get a cheaper, faster, and prettier Raspberry replacement. Spoiler alert: It will be “Back to the future.” Grüezi YouTubers. Here is the guy with the Swiss accent. With a new episode and fresh ideas around sensors and microcontrollers. Remember: If you subscribe, you will always sit in the first row. Who invented it? No, not the Swiss. In my video about my Home Assistant setup, many viewers wrote that they left the Raspberries behind and went back to Intel or AMD chips. How comes? As shown in my last video, I use two Raspberry Pis. One to run Home Assistant and one for IOTstack, which contains several docker containers like Node-Red, InfluxDB, and Grafana. The viewers proposed another approach: Use a single Mini PC using Proxmox. That is what we will do today. In this video: - I will show you the catches I made on eBay and explain why I went this route - Install Home Assistant on the cheapest PC - Install Proxmox on “bigger” PCs - Install native Home Assistant on one virtual machine and - Install Debian, Docker, and IOTstack on a second VM on the same PC - Compare the speed with my Raspberries - Compare the power consumption and - Compare the prices - Of course, we do not compare availability. The Raspberry has none. At least if you are a Patreon on this channel, you have a headstart and will still find used Thin Client PCs on eBay. The others maybe have to wait for a little till new PCs appear on the platforms Let’s start with the hardware. Typical viewer recommendations were to use an Intel NUC with a current CPU. For example, with an N5100. While this is a good choice, its price plays in a different league, and its performance is also higher than needed for stuff that runs on a Raspberry. So I went the second proposed way: Used Thin Client PCs on eBay. You find a lot of them from different brands and with different configurations. They start as low as 10 Euros and go up to more than 200. So the first question is: Which ones do we have to choose for what purpose? If you just want to install Home Assistant, you do not need much more than the performance of a Raspberry Pi4. 4GB memory should be sufficient, and 8GB SSD should work, too. To go the Proxmox way, you need at least 8GB RAM and a 128 GB SSD. On my 34 Euro Fujitsu with 4GB memory and only 8GB SSD, Proxmox did not install. You could order a used 128GB SSD for 20 Euros if you want to try Proxmox. The 4GB RAM is the bare minimum, but ok for two VMs or a few containers. Its processor is an AMD “System on a chip,” which includes peripheral components to reduce cost and size. Its Thermal Design Power is 15W. What does that mean? TDP is the average maximum power a processor can dissipate while running commercially available software. Generally, processors with lower TDP consume less power unless we torture them. As a rule of thumb: Newer processors have smaller structures and dissipate less heat for a defined speed. For CPUs of a similar age, a low TDP usually means also low compute power. For comparison, I ordered two Lenovos: one with an I5, 4GB memory, and a 128GB SSD. The second with an I7, 8GB memory, and 240GB SSD. The TDP of these processors is 35W. Here is the CPU speed comparison based on Passmarks. The Raspberry is the slowest. But, because its CPU is new, it has a TDP of only 4W. The old I7, in comparison, is six times faster, and its TDP is 35W. The others are in-between. We will later compare the power consumption in a real Home Automation scenario. The Intel N5100 is the newest CPU. It offers a similar speed to the old I5 with a TDP of only 6W. Not bad, also compared with the Raspberry. Unfortunately, the prices for Mini PCs with this chip are pretty high. The same was true for the I5 or even the I7, BTW. We learn that the AMD will not perform well if we would use it to replace two Raspberries. For a Home Assistant server alone, it is ok. And very cheap if we compare it with a Raspberry. And here we come to an important fact: These Thin Client PCs are complete systems with a case, a power supply, large heatsinks, and so on. Look at this beauty! The build quality is excellent because they come from the commercial market. If we add all the needed stuff to make a home server out of a Pi, we quickly end up with 150 dollars. Compared with the 37 dollars for the Fujitsu. So you know what to look for if you shop for a PC. The brand is probably not too important. I would go with a small case called USFF (Ultra Small Form Factor), sometimes called 1-liter class. Motorcyclists like me probably would call it the 1000cc class. Why do I like them? First, they look nice. Second, they do not use a lot of space. And most important, they do not have unnecessary parts like beefy video cards, which consume power and are not used in server mode. Now we go on with the software. I promise you a good return if you stick till the end. There are quite a few small cliffs to circumnavigate on this journey. First, I will show you some tricks to installing Home Assistant on a Thin Client machine. They are not easy to find out. Ease of use is a significant advantage of the Raspberry Pi ecosystem, by the way. The first cliff: You cannot copy the Home Assistant image on the built-in SSD without additional hardware. If we do not want to purchase an msata adapter, we need to install and start a Debian Live on a USB stick. I'll leave you a link to where you get it. When started, you use Firefox to install Balena Etcher on the USB stick. After starting Etcher, you input this Home Assistant download URL into the “from,” select the internal SSD as a target, and confirm that you want to flash it there. Now, Home Assistant is on the SSD. But unfortunately, it did not boot, at least not with my Fujitsu S920. So you need to add these commands: Now your PC should offer a new boot option called “Homeassistant.”Select this one, and it should boot! After a few minutes, I had Home Assistant running on my cheap Fujitsu. An 8GB SSD is ok for the usual stuff. Right after the installation, we still have 66% or 4GB free disk. Result of this first “finger exercise”: We have replaced one Raspberry Pi for less than 40 bucks. Not bad if we consider that we bought a whole 4GB PC, not only a board. But let’s continue with the real stuff! On the next machine, I will use the Proxmox virtual environment because it is open source and can be used free of charge or, if you want to support the project, 95 Euros a year for one server. With Proxmox, you can run virtual machines as well as containers. Today I only will use two virtual machines: One runs Home Assistant with its OS, and the other runs Docker and IOTstack on Debian. I chose Debian because it is close to the Raspberry OS. Installing Proxmox is a piece of cake. Still, a few remarks: All my Mini PCs were in PXE boot mode. As usual, I had to press the F2 or F12 key till I got into the setup menu. There I had to: - Disable the PXE Boot Agent - Enable all virtualization stuff as well as maximum C-state support to save power - Disable secure boot - adjust the boot order to boot from the internal SSD Maybe you want to change other stuff you think is needed. For this step, a monitor with a display port connector is advised. To continue, we have two possibilities. Download the Proxmox VE installer ISO file and burn it to a USB drive using Balena Etcher. Or we copy the download URL into Etcher. I like this new and effective method. If you insert your USB dongle into a USB3 slot of your PC and switch the power on, you should see this screen. If not, change your boot order. Hit install, accept, choose the internal SSD as a target, enter your country and time zone and check the network settings. They should be ok if your PC is connected to your home network. Now hit install and wait for a few minutes until you get the address of your newly created server. After a final reboot, your server is ready. From now on, you can detach the monitor, keyboard, and mouse and use your browser. Enter the server’s address, including the port number. It might not work right away. So check if you use HTTPS and not only HTTP. If it works, you get a warning. Continue unsafe, and finally, you can log in using the user “root” and your password from before. Do not care about the nagging pop-up. We will take care of it later in the video. As you can see, we created a “datacenter” with one node and two storage devices. Next, we must create our first virtual machine and install Home Assistant. Here I leave you a link to an excellent video where “The Tinker Dad” shows how it is done. Including an essential trick for Proxmox version 7. Now we can go to the browser and connect to our newest instance of Home Assistant. Cool. Next, I want to create a second VM for IOTstack. As a base, I cannot use Raspberry OS because it would not run on an Intel platform. So I chose Debian in the AMD64 version. It also runs on Intel processors. Again, I leave you a link to a video for the Debian installation. It is very short. Just write this guy that he should stop the music! I chose 4 CPUs and 4GB of memory because I want both virtual machines to use all CPUs. But I am not sure if this is right. Maybe somebody can help me understand how Proxmox deals with cores and logical processors. And, of course, I name the server IOTstack and the user pi for the feeling. Later you will see that I also could name it “superpi.” Next, I use PiBuilder to prepare my machine. Unfortunately, copy-paste does not work in the browser. So I have to SSH into the machine. This, again, is not as easy as with the Raspberry Pi. We first have to include pi in the sudoer group with these commands: Now we can start the ssh server Now we can ssh into our server. In the first step, I tried to download PiBuilder. But before, I had to install Git. Now we can download the project and run all five PiBuilder scripts one after the other. No worries. The VM reboots after every script. And you have to enter the password a few times. But already now, you will understand why I could have named the user “superpi.” The boot process is mindboggling fast. If you want to use Zigbee2MQTT or Node-Red, you must do some additional steps. First, we must ensure that the Zigbee stick can be seen inside the VM. This is easy: We insert the stick into our PC, go to Proxmox→Hardware, and add a USB device. We select “Use USB vendor/device ID” and select the dongle, in my case, a Conbee II. After a reboot and lsbusb, we see the stick. With this command, Now we have to adapt these lines in the Node-Red and the Zigbee2MQTT definitions using And we have to delete or comment these two lines because they are Raspberry specific and do not work here. After starting docker, Home Assistant and IOTstack run without a Raspberry Pi. How cool is that! I backed up Home Assistant and IOTstack on my current machines, copied the files to my newly created Proxmox server, and imported them. After that, I had my complete environment, including all historical data running in the new environment. Two questions are left: How fast are these “new” servers compared with the old setup, and how much power do they consume? Do you remember: In my last video, I complained about the slow compilation speed in ESPhome. This is why I compared the Pi and my new machines using this compilation process. Remember that compilation includes downloading some ESP repo stuff, and the CPU never exceeded 25%. Still, I think this is an actual situation. As we see, the slow Fujitsu was similar in speed. As expected, the I5 is 3.5 times faster, and the I7 is the fastest, but not much. But what about power consumption? To get precise measurements also in the low power area, I recently purchased this costly meter, mainly to do a future video on measuring power consumption in our homes. Here are the results: My two Raspberries consume around 12 Watts, including power supplies. The Fujitsu with just Home Assistant running consumes around 14W for only Home assistant. The i5 runs at 16-17W, even during installations. It has spikes to 25W but never got below 14W. The I7 with double the cores runs in the same ballpark when idle. But it spikes higher than the I5. As said before, I have no N5100 PC. So I cannot judge how much it consumes. I would assume it is below 10W idle, but not much because of all the other parts on the mainboard. One additional advantage of the Thin Client PCs: Because of their large heat sinks, they run much cooler than my Raspberry Pi. So, what is the deal? Let’s assume the modern processors have an average 10W advantage over the second-hand thin clients. The difference in one year is 87.6kWh. Next year, one kWh costs 27 cents. So the difference is 24 CHF or Euros or Dollars. And I only pay half the amount. Because of my PV system, I sell my electricity for half the price. For a price comparison with current technology, I configured a new Lenovo M75 with roughly the same Passmark and 8GB/128GB. It comes to roughly 5x more than my used I5 or I7. So I think these thin clients are an excellent catch for a server. My viewers were right, and I am happy I listened to them. Just to be precise: The Raspberries will still be used here. Just not for servers. As a reward for the guys still here, I promised to show you how to switch this nagging Proxmox screen off. Go to “Proxmox Helper Scripts.” There you will find the “Proxmox VE 7 Post Install” script where you can disable this pop-up. And many more helpers. This was all for today. As always, you find all the relevant links in the description. I hope this video was useful or at least interesting for you. If true, please consider supporting the channel to secure its future existence. Thank you! Bye
Info
Channel: Andreas Spiess
Views: 700,676
Rating: undefined out of 5
Keywords:
Id: rXc_zGRYhLo
Channel Id: undefined
Length: 18min 4sec (1084 seconds)
Published: Sun Nov 20 2022
Related Videos
Note
Please note that this website is currently a work in progress! Lots of interesting data and statistics to come.