this post was submitted on 04 Jun 2025
4 points (100.0% liked)

Self hosted

424 readers
1 users here now

A place to share, discuss, discover, assist with, gain assistance for, and critique self-hosted alternatives to our favorite web apps, web services, and online tools.

founded 3 years ago
MODERATORS
 

cross-posted from: https://sh.itjust.works/post/39436160

cross-posted from: https://sh.itjust.works/post/39436154

Hello everyone, I'm building a new server for the house, it will act as a NAS for everyone and host a few services like paperless, immich, baikal, jellyfin, syncthing probably navidrone, etc. The main reason I'm building a new one is that my current one is a HP prebuilt with a 3rd gen i5 and 8GB ram that is slowly beating the bucket, my 4TB HDD is completely full and there's no more sata ports nor space in the case.

I am fully psychologically prepared to be 24/7 tech support, but after all I already am, and in this way I have to support services for which I know how they work (and that I trust!) and not some strange Big Tech service whose UI and inner workings changes every other day.

For reference my new build is:

  • CPU: Ryzen 5 PRO 4560G + stock cooler. Has integrated graphics, can use it for Jellyfin transcoding.
  • RAM: Corsair Vengeance 2x8GB (from my desktop before I upgraded to 64GB RAM. If needed in the future I will upgrade the capacity and probably switch to ECC, I've chosen the CPU since it has support for it)
  • SSD NVME (boot+VM storage): Verbatim VI3000 512GB
  • Storage (SATA): 4x12TB Seagate Enterprise (White label) to use ZFS and Raid Z1 + 1x512GB Samsung SSD as cache.

I'm planning on using proxmox on bare metal and spin up VM/containers as needed, for which I'm wondering:

  • I know proxmox can manage ZFS arrays, is it better to create the array via proxmox, then share it as needed via something like openmediavault in vm/container OR to create a TrueNAS VM and passthrough the SATA controller to it, then manage everything via TrueNAS? I've done the latter in the past on another server, it's holding strong

  • I don't know if exposing the server to the open internet is a good idea (of course with fail2ban and a firewall properly configured) or to just keep a VPN connection to the server always open. I think the latter would be more secure, but also less user-friendly for parts of the family. I'm using wireguard currently to remote into my server when needed, and sometimes networks like eduroam in my university block it completely.

    • Self signed SSL certificates might also be a problem in the latter case
  • Since I will experiment with this server a little bit, I was thinking of keeping:

    • One VM for services for the family (exposed to internet or VPN)
    • One VM for services I still want to expose (I currently expose a couple websites for friends with data archived in my NAS)
    • One VM for me to experiment with before going in "production" for the family

Each VM would host its services using Docker+Portainer. My question is: is this too convoluted? Should I just use proxmox's LXC containers (which I have no experience with) and host services there?

I was also thinking of spinning up a pfsense/opnsense box and put the server into a separate VLAN from the domestic lan. But that will be a project for a second time. Unfortunately the way ethernet is wired in my house and for the physical space I have available prevent me from separating the networks by physically using another router.

Thanks!

you are viewing a single comment's thread
view the rest of the comments
[โ€“] RedBauble@sh.itjust.works 2 points 1 month ago (1 children)

This is basically what i ended up doing in one year of experiments, building and rebuilding.

Proxmox with unpriviledged LXCs. The only VM runs OPNsense, isolates services into 3 different vlans: internal for LAN/VPN only services, external for a couple public websites I host, and auth for authelia and lldap. Only internal can reach the others always through a reverse proxy lxc, the other vlans are blocked in opnsense and can only reach outside internet for eventual updates or api calls. No services can reach the 192.168.1.0/24 subnet with all the home devices (no smart home stuff).

Bulk storage is on a 4x12TB raid z1 array. The NAS dataset can be accessed via a cockpit lxc, with samba, the only one that sits on my LAN. Some services (like forgejo, immich, paperless) use their own dataset on the same pool.

Apart from cockpit, only opnsense sits on my lan as well. All the lan devices pass through it+reverse proxy to reach the other services.

Snapshots and backups are handled with zfs_autobackup and proxmox backup server, which runs on the repurposed old server, on which i also backup a subset of important files.

oh, and at least for the home one: put a raspberry pi (or similar, maybe even nanoKVM) into your rack, and connect it. it's awesome for maintenance. I have a rs232 connected pi and can control from bios, bootloader, shell of my proxmox.

I was eyeing the nanokvm, pcie version, for that exact reason. My Pi runs pihole, but is otherwise separated from the server.

every lxc runs in its own /30 subnet, only having the firewall local. There are a few bigger networks at home, iot (mainly wifi, client isolated), family for our laptops/tablets/mobils, guest network, tor-access wlan and other stuff

I was also looking at setting up a router/firewall with another minipc and a 2.5G pcie ethernet switch. Kinda long term project

[โ€“] nibbler@discuss.tchncs.de 1 points 1 month ago

Oh lol, didnt notice the age of your post :D makes the community look rather dead :(

Congrats on the setup, sounds solid!

I allow my internal api calls to be clear text & direct... Not perfect but its a tradeoff. Allows me to not let the proxy access everything.