Thanks! Yeah i am already using a nginx reverse proxy in a docker container to expose my other docker containers so I was thinking two reverse proxies in a row might be too inefficient. Will definitely look into nftables. Nftable rules are temporary though right? What’s the correct way to automate running these rules on boot?

I was thinking the same thing regarding VPS and Wireguard. I use Wireguard personally to VPN into my home network for remote management, but I still haven’t looked up how to make a VPS as a proxy using it. I know they can join the same network and talk with each other but what’s the best way to route port 80 and 443 on the VPS to my server at home? Iptables?

Not OP, but I’ve been looking into Cloudflare tunnels on my end as well and ended up not going with them because you’re forced to use their own certs so they can decrypt and see the data. I mean most likely they aren’t doing anything untoward, but it’s still a consideration with regards to data privacy.

Yep! I was surprised at how power efficient the build was myself. It really pays to go with an APU both because it doesn’t go ham with the core count and clocks and also because you don’t need an external GPU. As long as you’re just doing light to medium loads and not transcoding at maximum speed 24/7, your power usage will be fine.

The reader itself leaves a lot to be desired though. There’s literally no UI besides the arrow keys and no way to configure font rendering etc. It’s cool that the functionality is there, but it needs work.

It idles anywhere from 28-33W, but when its doing heavy processing it spikes up to the full power consumption of the CPU (max I’ve seen is like 120W according to my UPS). I run it in Balanced performance profile so there’s essentially no limiter to the power consumption. I figured I spent all this money on a CPU, I might as well take advantage of its processing power when I need it.

Lately I’ve been running a 24/7 Palworld server, and that is constantly running at 65%-85% CPU (out of a possible 1200%). My UPS reports 45W.

If Palworld isn’t running and someone watches movies off of my Jellyfin, usage is around 40W-50W when doing transcoding, and 35W when doing direct play.

I went with Arch Linux, mostly because I am the most familiar with it, how barebones I can make it, and how rolling updates are generally easier for me to deal with than large break-the-world distro upgrades. All my services are running in Podman containers so they’re completely isolated from any library versioning issues.

DIY NAS all the way. I had a QNAP that had a known manufacturing defect in the Intel CPU and QNAP refused to provide any support or repair options despite knowing about the issue for a long time. I will never again bow down to silly corporate shenanigans when it comes to my data.

My DIY NAS is a bit…unconventional and definitely doesn’t fit in your budget requirement, but I’ll leave the parts list as an interesting thought experiment: https://pcpartpicker.com/list/Lm92Kp

…okay, look, I know its a bit crazy. No, its A LOT crazy. But I genuinely feel like it isn’t worth dealing with HDDs anymore when it comes to building a NAS. Back when I was using the QNAP, I had to replace each HDD at least twice and I spent $90-$100 bucks per drive. A NVMe SSD can easily outlast two or 3 HDD drives and you can get the MSI Spatiums on sale for $180, so in the long-term the costs even out. But the speed at which an NVMe array performs during scrubs and rebuilds blows a regular HDD array out of the water. Yes, its a higher up front cost, but an immensely better experience and the costs even out in the long run. Plus, a PCIe bifurcation expansion card is a hell of a lot smaller than 4 HDD drives, so it opens up your case selection for more compact builds.

I got the NZXT H1 because it was easy to build, came with cooler and PSU and just made things simple. It also goes on sale for around $180. You can definitely go with something else entirely. My thought process was that if I ever wanted a compact PC, I could possibly repurpose this case. This is just for me, it is not a hard recommendation.

I picked Ryzen 5600G because it was relatively cheap, decently powerful, and has HW h265 and h264 decode and H264 encode, which is basically what you need for Jellyfin or Plex. Just be aware that it only supports up to x4x4x8 PCIe bifurcation, so if you do go with a NVMe expansion slot, you can only put 3 on there and will have to use a mobo slot for the 4th. That’s how mine is currently setup and it works great.

Yeah, its crazy, and I am sure some people here will scoff at the build, but after using it for 3 years, I just can’t go back to regular HDD performance. An NVMe array just makes all of the services you host fly.

I run Nextcloud and two Jellyfin instances behind Nginx Proxy Manager. I also run a Palworld server. All of them are running under podman. I do use cockpit for checking container status, logs, and viewing the console for each container. I also use docker-compose to create all of my containers (using podman-docker of course). Unfortunately, all of them are running rootful instead of rootless, mostly because most proxies require root and setting things up for rootless like enabling low ports for regular users and allowing processes to run after logout are a pain in the ass.

My DIY NAS runs Arch

• LTS kernel
• BTRFS snapshots on root fs
• 4 drive NVMe array using ZFS raidz1
• podman for my docker containers

It’s been working fantastically so far.

I felt the exact same way. So many comments online told me that running Arch as a home NAS was insane, but after the Jupiter Broadcasting guys did it without much issue, I decided to give it a go and was pleasantly surprised. I think if most of your stuff is running in Docker and you have BTRFS snapshots for your root filesystem, the system’s pretty much bullet proof. The rolling updates also mean you’ll never have huge upgrade cycles that are a pain in the ass to migrate to. You’re always just dealing with small manageable fires instead of large complicated ones and that’s a plus.

Rust may be niche now, although it’s current momentum is huge, especially in the FOSS space.

Out of curiosity, what reverse proxy docker do you use that can run rootless in podman? My main issue, and feel free to correct me if I am wrong, is that most of them require root. And then its not possible to easily connect those containers into the same network as your rootless containers so then your other containers have to be root anyways. I don’t really want my other containers to be host accessible, I want them to be only accessible from within the podman network that the reverse proxy has access to.

And then there’s issues where you have to enable lingering processes for normal users and also let it access ports < 1024, makes using docker-compose a pain, etc. I haven’t really found a good solution for rootless, but I really want to eventually move that way.

I believe its https://browse.feddit.de/

LOL, I have a feeling you’re not wrong. I introduced Lemmy to a friend of mine and his first question was, “okay, what about the nsfw content though?”

Lemmy seriously needs better federated search. The fact that you need a separate website to find all the communities is just silly, especially if you’re on smaller instances, which everyone seems to encourage.

Apps do that, including some web app.

I don’t just mean grouping communities client side though. I mean super-communities that a mod or admin can curate and that other people can share and subscribe to. Out of curiosity, which apps have you been using that have this feature?

I also think instances just need to have more sane urls. It sounds silly but like it or not branding does matter. Instances like beehaw and sh.itjust.works just rub me the wrong way during first impressions. Even some of my friends who I’ve tried to introduce to Lemmy are just like…“what the fuck kinda website are you making me go to?” I am not surprised that lemmy.world is big, partly because the URL actually sounds official.

We also need super-communities that pull in content from multiple communities/instances, better multi-instance search, and a way to migrate between instances before the masses will be okay with smaller instances.

I believe Matrix already supports olm which is the same encryption technique used by Signal. The main issue with Signal becoming federated is that in order to make the federation work, a lot of metadata will leak and that could be a cause for concern when using Signal as a private messenger for important things like whistleblowing, etc.

Oh my bad then, ignore me hahaha.