Migrating from RPi4 to mini PC - hardware demands for HA

Hardware
1

Hi guys,

I would need some help please on choosing new PC.
Now I’m running HA on Raspberry Pi 4 (2GB ram only, boot from SSD) and my HA is growing, so RPi is freezing almost every day when I configure stuff (ESPHome, automations or configuration.yaml). I want to expand my HA setup also
I’m pretty tight on budget and considering these two options, but I’m also open for any other suggestion in similar budget class :slight_smile: .
First is refurbished: Fujitsu Esprimo q920

  1. Mini PC Windows 10 Pro, Intel Celeron J4125

Since I’m quite noob in hardware stuff, I tried to compare on PassMark and it’s obvious that Fujitsu with Core i5 4590T is superior to Celeron. But…

  1. is it OK to invest in used PC or should I go with new?
  2. Celeron (like J4125 with 8GB RAM) is enough for moderate HA setup?

Thank you for any suggestions and help!!

Regards,
Andrej

1 Like
2

Either of those will be more than sufficient.

3

There is nothing wrong with buying a refurbished pc. As nick said, both are good choices.

4

Thank you both for your quick answer.

What about ssd… I read here on forum that I should go with pro ssd…what is considered as pro ssd? The market is full of ssds and what parameters should I look for?
Thx

5

Good question. It is not that you need a datacenter ssd.

6

If it was me that mentioned “pro” ssd, I was probably referring to Samsung, which has various grades including “Evo”, “Qvo”, “Pro”. Otherwise pro may refer to something with powerloss protection, or something specifically designed for servers and workstations, like the Kingston DC500M (SATA drive).

The Samsung Pro model line is generally faster, but most importantly has MUCH higher reliability when writing small amounts of data at a fairly current rate. SSDs have something called write amplification, and it gets worse when you are dealing with small amounts of data, like continual writes to a database.

There are other factors as well, my current recommendation is the Samsung 970 Pro 1TB, assuming you have a system with a PCIe M2 slot available… I believe that Fujitsu is far too old to have an M2 slot. This is probably one of the last 2-bit per cell drives on the market, and even under much harsher conditions should last for many years.

I also recommend partitioning the drive to reduce fragmentation, for all my systems /boot, /var, /home, and swap get their own partitions, though most of that would not apply to a server for just HA, and if you are using HASSOS I do not know if it applies at all.

1 Like
7

Thank you, very helpful!
Just ordered refurbished HP mini PC with i5 7500T processor…and it also has M2 slot, luckily:) I decided for a little more powerful cpu (and less W power than fujitsu) because I will use it for HA, nextcloud and adguard.

And two more questions… should I go with proxmox or ubuntu server with docker?

I looked some yt videos about HA supervised installation and read some pages about it, but still can’t decide if this is the right path. Now I’m used to supervisor… Any suggestions?

thanks and kind regards

8

ubuntu server with docker will not give you the supervisor.

I use debian+supervised but proxmox+haos+ubuntu will work well if you want ubuntu.

1 Like
9

The size of your configuration does not affect the speed of the RPi, for sure it is not the reason for freezing (unless you have millions of integrations, devices and automations)
In other words, a faster computer will not solve this….
I suggest to fix the culprit before moving it to new hardware :wink:

1 Like
10

The problem causing freezing is probably either heat or power supply, a new computer probably will fix it but not in the way it was expected

11

Thanks. Yes, I’m aware of that. Db file is growing around 500mb per day and on new configuration will keep in recorder and history only energy sensors and sensors for heat pump…
My temp solution was to swap RPi4 (2GB ram) for another RPi4 (4GB ram) which I use for other purposes and the difference is very obvious (maximal temp of CPU is 58°C during back up process, on 2GB ram it was over 65°C), not to mention ram usage, and problem is also SSD usb disk (I have SSD boot) and when db file large (over 2gigs) it heats stays hot constantly. Since I’m running HA on 4GB ram I hadn’t have any freezes.

12

@richieframe
Which HA distribution do you use? Just asking cause I’m planning how to partition disk and how many ssds I need for the setup with nextcloud… on proxmox on debian.

13

I am using the docker install and managing it using docker compose, I scripted an automatic backup into the upgrade process, so if something goes wrong I can rapidly downgrade no matter what the new version did to the db.

Docker stores its files in /var, so that needs a bunch of space, I am using 24GB but it has gone up to around 40, the more containers you run the more space you need, 64GB should be minimum. More space also allows you to keep prior image versions for a longer period of time without purging.

for /boot I suggest 2GB if you are regularly upgrading kernels or playing with multiple kernel versions, otherwise 1GB, I use BTRFS and NFS so I need 5.11 or later for certain performance and reliability features, but also keep at least one prior kernel release AND the distro LTS kernels (5.4 right now).

for / I am using 20GB, so 64GB gives plenty of room to grow

for /home I am using tons of space on the computer with interactive login, 36GB. This is where I do code compilation and includes various SDKs and dev environments, browser cache, “snaps”, downloads, etc. If you are using the machine only for remote terminal work or running dev work in docker, then you probably do not need a /home partition, or you can keep it small like 16GB. I have more than 300000 files in there, so the separate partition helps the filesystem management.

Using Proxmox requires precise and thorough evaluation of the requirements for the underlying virtual machines, you could probably run nextcloud in docker instead, you will get better performance and it will be easier to manage. I would suggest an additional SSD in the system for daily backups of your important data, combined with a backup to a secondary on-site system (the pi if you can fix the freezing?), and have that manage offsite backups.

Estimate that for a docker host you should allocate at least 170GB of space

2 Likes
14

Thank you @richieframe ! Very much appreciate your suggestions and insight, you posted just in the right time! Now I have some information to process…:slight_smile: