Home Server and Self Hosting General - Technological Self-Sufficiency

  • 🇵🇦 Nuestro primer dominio localizado está en español en kiwifarms.pa. Our first localized domain is on Spanish on kiwifarms.pa.
  • Want to keep track of this thread?
    Accounts can bookmark posts, watch threads for updates, and jump back to where you stopped reading.
    Create account
Proxmox is great if you know Debian already. Otherwise, especially coming from vSphere, it can feel a bit rough and clunky in the beginning.

Make sure to check out all the available plugins, though. I use Proxmenux for a number of tasks that otherwise would require a lot of clicking. It runs on the console, so you can just ssh into it instead of having to use a GUI with a web browser. An extremely handy timesaver.
Thank you, I had no idea plugins were even available. So far the only time I’ve had to dip into the console has been switching to the free update repository, otherwise I’ve been pleasantly surprised how much you can get done in-gui.
It’s not too bad coming from vsphere, but the acid test will be when something goes wrong. I’ve already noticed vi is ACTUAL vi on the host, not vim as I’ve become accustomed to on basically every Linux install ever. It’s the small details like that that end up befuddling me in a crisis.
 
Thank you, I had no idea plugins were even available. So far the only time I’ve had to dip into the console has been switching to the free update repository, otherwise I’ve been pleasantly surprised how much you can get done in-gui.
It’s not too bad coming from vsphere, but the acid test will be when something goes wrong. I’ve already noticed vi is ACTUAL vi on the host, not vim as I’ve become accustomed to on basically every Linux install ever. It’s the small details like that that end up befuddling me in a crisis.
It being Debian underneath makes it easy to customize to fit your needs (I installed vim and set it the default for example). Of course with that flexibility you also get the ability to shotgun yourself in the foot if you're not careful but I always prefer the option vs being completely locked down.
 
@22 Rimfire If you haven’t found it already there’s - https://community-scripts.org/ . I found it really helpful for getting all my containers etc set up.
It does fuck up GPU setups a lot, had to do it manually for jellyfin and ollama, though strangely the immich one just werks.
I do recommend studying the command flow of the scripts and also doing some setups manually if only to pad out your knowledge.
 
Dumb question. I'm trying to run a home ai lab setup. My parts are pretty old except for my GPU (rtx 3060) and it does a good job with web ui. Trying to set up hermes agent and thinking my GPU is going to be limited on that.

Would just updating my GPU be enough or do I have to go through the whole process of selling my kids for ram? If I update my ram, I will likely have to update my cpu which means new motherboard etc. So ideally just have to bite the bullet for one part rather then an entire computer replacement.
 
Dumb question. I'm trying to run a home ai lab setup. My parts are pretty old except for my GPU (rtx 3060) and it does a good job with web ui. Trying to set up hermes agent and thinking my GPU is going to be limited on that.

Would just updating my GPU be enough or do I have to go through the whole process of selling my kids for ram? If I update my ram, I will likely have to update my cpu which means new motherboard etc. So ideally just have to bite the bullet for one part rather than an entire computer replacement.
The rtx 3060 should be plenty good enough to run a reasonable model with Hermes. Get it all running before you go crazy chasing upgrades, chances are you’re overthinking it.
 
It does fuck up GPU setups a lot, had to do it manually for jellyfin and ollama, though strangely the immich one just werks.
I do recommend studying the command flow of the scripts and also doing some setups manually if only to pad out your knowledge.
What problem did you have with your GPU? The jellyfin and tdarr scripts coped fine with my B50 with sriov_numvfs 2. Seriously, fuck intel for gimping the B50s SRIOV capabilities post launch.
 
What problem did you have with your GPU? The jellyfin and tdarr scripts coped fine with my B50 with sriov_numvfs 2. Seriously, fuck intel for gimping the B50s SRIOV capabilities post launch.
sriov works fine, but i prefer using LXCs because of the memory pool it's sharing. Ollama and jellyfin simply wouldn't recognise my B50 without serious tinkering with the mapping permissions. ComfyUI is a lost cause, it simply won't detect any card even if i install it on the host itself.
 
It being Debian underneath makes it easy to customize to fit your needs (I installed vim and set it the default for example). Of course with that flexibility you also get the ability to shotgun yourself in the foot if you're not careful but I always prefer the option vs being completely locked down.
Is there something supposed to go wrong? Proxmox is a very solid and stable distro. By default, it will induce a two-staged reboot to do an update, for maximum stability.

There's a recovery console it boots into if something fails, so you can always fix things.
 
Is there something supposed to go wrong? Proxmox is a very solid and stable distro. By default, it will induce a two-staged reboot to do an update, for maximum stability.

There's a recovery console it boots into if something fails, so you can always fix things.
I'm talking about if you go ham doing a bunch of goofy shit that would fuck over any Debian install if you're not careful; it's plenty stable in normal use. It just doesn't prevent you from doing those things unlike something like modern TrueNAS which is heavily (frankly to a ridiculous extent) locked down.

As an example (this was on a test box): when initially setting it up I only had a single drive so the boot/os wasn't redundant. I got an extra one after the fact and did some low level cli stuff to then mirror it. Worked the first time; second time I forgot a step and made it unbootable. I had to boot in with a recovery usb and chroot in/fix things.

Should you ever do something like that in first place in production? Absolutely not. Is it useful to be 'able' to if necessary? Hell yes.
 
Última edición:
I'm talking about if you go ham doing a bunch of goofy shit that would fuck over any Debian install if you're not careful; it's plenty stable in normal use. It just doesn't prevent you from doing those things unlike something like modern TrueNAS which is heavily (frankly to a ridiculous extent) locked down.
Well, TrueNAS is a NAS OS, Proxmox is a hypervisor OS. It is supposed to be open for experimentation because that is sometimes needed, depending on what exactly you use it for.
 
Proxmox is a hypervisor OS. It is supposed to be open for experimentation because that is sometimes needed, depending on what exactly you use it for.
So are ESXi and Nutanix but good luck being able to mung those to that degree (not that you should). I'm not really disagreeing with you but more praising Proxmox.
 
I'm a bit perplexed about the best way to bring my pile of docker containers over to the proxmox host. Looks like I have two options:
- Install docker in LXC on the host
- Install docker on a Linux guest VM

Searching around a bit seemed to suggest running docker direct on the proxmox host might be a bad idea, but the count of docker installs on https://community-scripts.org/ being so high suggests perhaps it's fine to do that.

I currently have them running in a hyper-v guest so importing that as a VM seems more straightforward than moving/reconfiguring all the containers onto the proxmox host.

Am I overthinking this?
 
I'm a bit perplexed about the best way to bring my pile of docker containers over to the proxmox host. Looks like I have two options:
- Install docker in LXC on the host
- Install docker on a Linux guest VM

Searching around a bit seemed to suggest running docker direct on the proxmox host might be a bad idea, but the count of docker installs on https://community-scripts.org/ being so high suggests perhaps it's fine to do that.

I currently have them running in a hyper-v guest so importing that as a VM seems more straightforward than moving/reconfiguring all the containers onto the proxmox host.

Am I overthinking this?
I wouldn't install Docker directly on the host as that could severely interfere with the network config/devices Proxmox does. You do have two options:

1. The simplest is just spin up a VM and install Docker in it.
2. You can run docker/container.io images directly as LXC but it can be a bit goofy (for one thing you might not have the container log output visible).

You can use the 'Pull from OCI Registry' and pull down Docker images and it will make an LXC template you can then create a container from. I have TvMosaic running for example:

Screenshot 2026-06-22 at 9.17.04 PM.png Screenshot 2026-06-22 at 9.17.23 PM.png

LXC is still a container sharing the kernel of the host but it seems to be treated way more like a 'light VM' than Docker containers are. It took me a while to get comfortable with it and honestly I'm still not sure I like them. Most of my containers were done from a base OS image and then I went in and installed things just like it was a VM; this one is running a private WoW server for example.

Screenshot 2026-06-22 at 9.28.07 PM.png Screenshot 2026-06-22 at 9.30.27 PM.png
 
Última edición:
but it seems to be treated way more like a 'light VM' than Docker containers are.
You are right on this. K8s/Docker containers are for single apps and microservices. LXCs are for entire OS environments sans kernel/filesystem/networks.
Am I overthinking this?
I have never had issues with running docker on the host, but i agree that it's bad practice. Unless you have specific needs like hardware device access (GPU) or for example BGP (reverse proxy with quic or http/3), then nesting docker in an LXC is painless.
 
OK, thanks. I'm gonna just try to import the vhdx disks and spin up a dedicated VM for hosting docker images similar to how I've been running it on Hyper-V.

Sounds like I should take the other route when I bring Plex in later, as I'll need to pass the iGPU through for transcoding.

Pleasantly surprised by Proxmox's capabilities so far. Way back when I administered a bunch of vSphere environments stuff like storage vMotion could be a bit precious to set up and use. This thing just moved a live Minecraft server with a few clicks and zero problems.
 
nesting docker in an LXC is painless.
I have Kubernetes (k3s) running in LXC and it works fine. For container nesting you do have to make the LXC parent container privileged which is technically a security risk (root in the container is the actual root on the host). I definitely wouldn't do that if I was hosting something on it that was accessible by the public but I'm tempted to at work for internal-only services.
 
The rtx 3060 should be plenty good enough to run a reasonable model with Hermes. Get it all running before you go crazy chasing upgrades, chances are you’re overthinking it.

Thanks mate. Trying to learn all this on the fly and maybe consulting ai to build ai might be leading be down rabbit holes that aren't needed.

Can get hermes running, trying to get an agent with ~~Soul~~ and able to write to it's memory is the current obstacle. Will give it to again.
 
Well, TrueNAS is a NAS OS, Proxmox is a hypervisor OS. It is supposed to be open for experimentation because that is sometimes needed, depending on what exactly you use it for.
Plus TrueNAS is pretty much a 'set and forget' OS where you load and configure your torrent client, Plex/Jellyfin etc. and then boom, you basically don't have to touch anything again and can just remote into the apps or use something like Filezilla to transfer data to it. Logged into mine for the first time in 3 weeks and everything is running as normal: disks are good, temps are good and scrubs have found no issues so far.

Yep it's limited in what you can do with it, but for most people wanting a basic fucking media server it's a godsend for being simple as fuck to set up and a billion times better than those Synology monstrosity OS's
 
I have Kubernetes (k3s) running in LXC and it works fine. For container nesting you do have to make the LXC parent container privileged which is technically a security risk (root in the container is the actual root on the host). I definitely wouldn't do that if I was hosting something on it that was accessible by the public but I'm tempted to at work for internal-only services.
?
It works fine w/o privileges as long as you keep the nesting flag enabled.
1782224215627.png
 
Atrás
Top Abajo