The Linux Thread - The Autist's OS of Choice

[August Levasseur]

A Match Address statement before PasswordAuthentication yes. ...why do you want to do this?

Connecting arbitrary devices from the local network but only known devices from beyond? Seems reasonable to me.

 

[Betonhaus]

A Match Address statement before PasswordAuthentication yes. ...why do you want to do this?

So that I'm not locked out of my server if I don't have access to an authenticated device (like if my main computer does and I had just installed a new testing distro on my secondary.

 

[Mikoyan]

So that I'm not locked out of my server if I don't have access to an authenticated device (like if my main computer does and I had just installed a new testing distro on my secondary.

That would work but why not just keep your SSH pubkey on USB and copy that in (along with other home directory stuff) when using a new distro? Going semi-portable like that is useful if you're going to be distrohopping.

 

[acid_ra1n]

the humble qmlkonsole

 

[Dread First]

...terminal access

...through a browser

STOP this niggerlicious bullshit IMMEDIATELY and use SECURE MOTHERFUCKING SHELL (that's SSH if you really are a nigger and not just pretending). THIS is what it's THERE FOR, so you're NOT using some horrendous scarcely-used SECURITY RISK in its place. There is NO EXCUSE. Literally ANY CONNECTIVITY PROBLEM that you have can be resolved using SSH in order to connect via SSH securely.

So how do I set it up so that I can use a username and password to connect from the local network, but can only use a ssh key when connecting remotely?

Homie it ain't that serious, Cockpit requires sshd to be enabled or else you can't get it off the ground. If I'm already playing Jellyfin/Navidrome stuff while browsing FreshRSS on one browser window, it feels more natural to have terminal access via browser window as well. I wouldn't be using Cockpit over barebones SSH in a terminal window if Cockpit was some hideously insecure mess.

Is it niggerlicious to use a web browser for stuff that you self-host? Probably, but I'd argue it's even more niggerlicious to run Plex over Jellyfin just because the UI is prettier. I'd also argue that running Unraid/Proxmox/TrueNAS/etc is more niggerlicious than running barebones Ubuntu Server because the former 3 abstract away the nitty gritty instead of making you grapple with it.

To wit: huge companies in FOSS like Red Hat and Canonical are 100% purveyors and peddlars of the most niggerlicious, unintuitive bullshit possible. I won't ever deny that. HOWEVER, they're not all complete fuckwits. Linux servers get deployed all the damn time, there are tons of web portals for sysadmins and technicians to access stuff on, and they're all undergirded by SSH instead of TLS and JavaScript like a client side web page.

If big Linux companies treated the server ecosystem the way they treated the desktop ecosystem, Linux would've never mogged the shit out of Windows Server.

>get new external touch screen monitor for testing
>gnome just refuses to let it touch for some reason
>kde works
>kde has dogshit maliit osk that doesnt show up in installer
>plasma-keyboard works
>worried about kde being even buggier than gnome
>so far ran into like 3 severe bugs: broken copy paste, password failing, and kde being very laggy for some reason
so had to reinstall twice and still just getting through fedora's initial setup
here i am dreaming of shipping something and i cant even get a demo working lol

I'll spare you the pain: Plasma 6 is fucking worthless, regardless of whether you're using Wayland or Xorg, largely because there are an assload of persistent regressions that still never get fixed despite Red Hat forwarding the shit on Bugzilla upstream. GNOME Shell is worthless despite the touch UI focus because the GNOME team spent so long huffing their own farts like paint fumes that oxygen no longer reaches whatever's left of their brains. Plasma under Wayland with multiple monitors is abysmal because of mouse hangs when moving between monitors, generally poor performance despite massively overspecced hardware, etc. Oh and the general tendency to just fucking artifact all the goddamn time. The worst part? The X11 session is noticeably worse under Plasma 6 despite being fine like wine in Plasma 5, so they're just gonna shitcan it altogether come Plasma 6.7 instead of actually fixing the shit that they broke.

If you really want a "clean" Fedora experience, use a spin like Xfce, Cinnamon, MATE, or LXQt. Those ship with Xorg by default, no Wayland fuckery, and they actually fucking work properly. Idk about your touch monitor business, but it might be worth checking out. You got nothing to lose at this point, right?

 

[analrapist]

Homie it ain't that serious,

Realistically, I was trying to emulate SCHIZO RHETORIC. The point is to encourage use of the command line, free of a gigabyte of memory to render a tab in Chrome. Works fast everywhere. People just have bizarre notions of what they want automated. SSH runs on potatoes.

Products I have zero idea of the use case for in your post include: Cockpit, Plex, Jellyfin, Unraid, Proxmox, TrueNAS. I think some of them are storage related from their names but IDFK. I think Plex is media related? My concession to "normal" media systems is that I have a copy of minidlnad that runs, but again, I stream all my media over SSH. That DLNA server is for normies.

Unraid/Proxmox/TrueNAS/etc is more niggerlicious than running barebones Ubuntu Server

You bet. You're being endlessly sincere here, but I'm happy to have a serious Ubuntu user if only to balance out the Mint and Debian factions.

I mean, I'm so brainrotted that I unironically use Gentoo. Go figure, right?

 

[Betonhaus]

You bet. You're being endlessly sincere here, but I'm happy to have a serious Ubuntu user if only to balance out the Mint and Debian factions.

I mean, I'm so brainrotted that I unironically use Gentoo. Go figure, right?

I got to be honest I don't think anyone in the Mint faction is trying to use Mint as a server. I'm using Ubuntu for my server, but could just as easily be using Debian with no noticeable difference - just I needed support for my GPU and didn't want to deal with a backported kernel at the time.

 

[AnOminous]

I got to be honest I don't think anyone in the Mint faction is trying to use Mint as a server. I'm using Ubuntu for my server, but could just as easily be using Debian with no noticeable difference - just I needed support for my GPU and didn't want to deal with a backported kernel at the time.

FreeBSD or NetBSD make better servers (imo). Mint is just a really good Linux distro for using it as a daily drive. The BSD-based stuff is for a headless server I place under the bed and only use via a shell.

 

[Betonhaus]

FreeBSD or NetBSD make better servers (imo). Mint is just a really good Linux distro for using it as a daily drive. The BSD-based stuff is for a headless server I place under the bed and only use via a shell.

a quick google search indicates that migrating all of my docker apps to BSD will be difficult and unsupported.

 

[Dread First]

The point is to encourage use of the command line, free of a gigabyte of memory to render a tab in Chrome.

I'm a diehard Firefox autist. Not anywhere near as nice on memory usage as it once was, but it's not like I keep a Cockpit tab open all the time. I just log in to run something like an update or to fiddle with my docker-compose.yml configs, and then close out once I'm done. I lose more memory keeping Firefox with GameFAQs open on one monitor while playing Persona 4 Golden in the other window, and reading the guide on how to do all the social links on a single save file without going into New Game Plus. At least the GameFAQs tab is staying persistently up.

Works fast everywhere. People just have bizarre notions of what they want automated. SSH runs on potatoes.

Yes, I do agree that SSH runs on potatoes and works fast anywhere. I am also perfectly aware of how "automation" is poorly defined. That said, it seems like you're conflating two distinct issues that aren't quite aligned.

SSH works great insofar as giving you access to whatever computer you wanna log into. I can also attest to it running on worse-than-dogshit hardware because I've been using SSH intermittently since the days of jailbroken iPhones running Cydia and rooted HTC phones that sideloaded pirated apps through adb. But that's all SSH is capable of: giving you access to a computer you need to fiddle with. SSH in itself doesn't automate anything.

My setup is pure manual. I'm a weirdo who manually triggers sudo apt update && sudo apt full-upgrade -y once a day. I never bothered to learn how crontab works, and I never bothered learning how to set up scripts for consolidating minor things I do on the regular. This is about as manual as it gets for a Linux system. I honestly could get away with accessing my home server with SSH on my current home desktop PC, and in fact, I had to do things like that before I got Cockpit up and running. I just prefer using the terminal window in Cockpit through Firefox instead of SSH via GNOME Terminal in Fedora Cinnamon because it "feels" nicer. "Feels" is highly subjective but FWIW, I "know" the "correct" answer is that GNOME Terminal with SSH is more efficient.

Products I have zero idea of the use case for in your post include: Cockpit, Plex, Jellyfin, Unraid, Proxmox, TrueNAS. I think some of them are storage related from their names but IDFK. I think Plex is media related?

In absolute fairness: these are home server tools, not desktop tools. I started learning about home server Linux stuff ever since I built a new PC and needed to put the old one to good use. Let's just rattle off what you listed and explain what they all do.

- Cockpit: an admin control panel for your home server, accessible through your web browser. Gives you at-a-glance information about core system metrics like CPU temps, network traffic, disk utilisation, hostname, processes running, memory used, etc. Also gives you tools to view journald logs, storage devices via lsblk, create and manage your KVM virtual machines and LXC containers, create and manage user accounts, and a terminal window. A brainchild of the Red Hat ecosystem, universally available on just about any Linux distro nowadays. Debian, Ubuntu, and assorted clones have it in their repos, as does SUSE and Arch Linux's repos (side note: Cockpit assumes systemd, so Artix and Devuan are hard no-gos). Abstractions are minimal and tasteful; they map cleanly to systemd functionality like journald and networkd. The terminal in Cockpit is 100% SSH, just as fast and snappy as if you're running it in GNOME Terminal or Konsole or whatever. An excellent tool for home servers, absolutely worth setting up.

- Plex and Jellyfin: They're both home media servers, Plex being proprietary and Jellyfin being FOSS. You load up on "legally acquired" digital media like your TV shows, movies, and music, shove it on a server box, install either of them on that server box, run through the initial setup, plug in your LAN IP and your port number into your browser's URL bar or the Plex/Jellyfin app, and boom: you can watch your "legally acquired" digital media in a browser, on a smart TV, streaming dongle like your Roku or Fire Sticks, your phone, etc. Think Kodi but you actually use it for local media playback instead of sideloading a pirated streaming repo and then getting mad at why none of the stuff works because you forgot to pay for a debrid service. Plex has more polish, but it's not truly self-hostable because you need to make a Plex account, you need to pay a subscription fee for basic functionality like hardware transcoding, it's just enshittification in self-hosting, which negates the premise of self-hosting entirely. Jellyfin is wholly FOSS, the UX is surprisingly refined despite blemishes here and there, all the important stuff that Plex makes you pay for is gratis in Jellyfin, all accounts are 100% local on your server box with Jellyfin, blah blah blah.

- Unraid, Proxmox, and TrueNAS: these are all server-side operatiing systems. You run these on bare metal as your primary host operating system. All of these operating systems allow you to spin up containers or virtual machines for the services you wish to run. If you're running an enthusiast homelab setup or a wholly sovereign cloud, all three of them would've crossed your radar at some point. Unraid does containers really well, has tons of integration wtih Docker, virtual machines exist but they're just "okay" from what I can tell. Proxmox's specialty is spinning up virtual machines, you have granular control over everything, containers exist if you need them but virtual machines are what it excels at. TrueNAS is a NAS operating system first and foremost, but it has limited functionality for VMs and containers.

The biggest use my home server has right now is media playback via Jellyfin. Technically, Jellyfin exists in Ubuntu's repositories, but I opted for Docker Compose.

TLDR: it's a "declarative" text file where you outline all the services you want to run together in whatever configuration you see fit, then run docker compose up -d in your terminal window, fancy shmancy computer wizardry happens, and then you're able to access whatever service you just spun up.

Jellyfin is the "anchor" to my setup, where I run Radarr and Sonarr for metadata management, qBittorrent for obvious reasons (it also has a web UI that looks identical to the normal graphical application), and Gluetun to get my VPN connectivity to work in a containerised environment. All of these services are spun up together because I "declared" them and all of their parameters in a single docker-compose.yml file. I have hard drive UUIDs pegged at the appropriate mountpoints, VPN encryption keys listed, web console UI logins declared, the whole nine yards, all of that in a single text file. I used to think "reproducibility" was some gay ass buzzword, but no: I can take this file, make a dozen backups of it, and have them at the ready in case the worst happens and I need to get my shit back up and running again.

End result? I torrent the DVD rip of Jackie Chan Adventures via qBittorrent (hooked up to VPN), I copy the rip over to my media folder so I can seed the original, use Sonarr to fix the metadata from "Jackie.Chan.Adventures.DVD.Complete.Series.XViD" to "Jackie Chan Adventures (2001-2005)" and have all the seasons correctly detected. Once it plays nice on Sonarr, I update my library in Jellyfin so that it gets detected. End result? I get to enjoy Uncle hopping on one leg while chanting a chi spell on my phone, my PC, my smart TV, etc. Best part? I can actually persistently seed stuff and have good seeding ratios instead of getting flagged as a leecher all the damn time.

Mind you, this is all "baby's first home server" that I'm talking about. Far easier and much less intrusive to make a Jellyfin server with assorted goodies than it is to convert my previous gaming PC into a single-purpose network firewall router appliance thingamajigger with pfSense or OPNSense. I'll save that pipe dream for when I live alone and have the freedom to fuck with the network without borking everyone else's WiFi for hours at a time.

My concession to "normal" media systems is that I have a copy of minidlnad that runs, but again, I stream all my media over SSH. That DLNA server is for normies.

Homie, I completely forgot that DLNA even existed; haven't thought about it since the days of HTC using DLNA functionality as a selling point. Jellyfin doesn't use DLNA at all. Technically, a plugin exists for it, but it no longer ships with DLNA functionality OOTB.

I'm happy to have a serious Ubuntu user if only to balance out the Mint and Debian factions.

I'm very much in the Mint camp, I haven't wholly trusted Canonical Ltd since 2012 when they did that shady Amazon partner data sharing thing with Ubuntu 12.10. The catch is that I just needed a server OS. Ubuntu Server was basically the "easiest" option for me to pivot to because I've spent over a decade intermittently using Ubuntu either directly (re: 10.04-12.04 days) or indirectly (re: Linux Mint).

I must say: Ubuntu Server is pleasant to use and a genuinely refined experience. I would expect as much from their server offerings because Canonical more or less abandoned the desktop wholesale for the server/enterprise market. They're doing so well that they're actually toe to toe competitive with Red Hat... and absolutely eating Red Hat's lunch imo, but different story for a different time.

I got burned hardcore when fiddling with Fedora (separate screed for another time), so Fedora/RHEL was a no-go. Debian is "okay," but Trixie's only supported for like 2-3 years tops as stable, and the further down the old/oldold/oldoldold stable pipeline you go, the less pleasant the experience is. I haven't considered SUSE because I forgot that they exist, really need to give SUSE a whirl one of these days. Side note: SUSE managed to outlive Attachmate and Novell, and they're once again wholly independent. Good for them.

I mean, I'm so brainrotted that I unironically use Gentoo. Go figure, right?

Nigga you're posting on Kiwi Farms in the Linux thread. You ain't so much brainrotted as you are inflicted with the tism. Look at who you're talking to!

FreeBSD or NetBSD make better servers (imo). Mint is just a really good Linux distro for using it as a daily drive. The BSD-based stuff is for a headless server I place under the bed and only use via a shell.

I really wanted to make FreeBSD work, but so much stuff I wanted to run basically needs either LXCs or full-on Docker shenanigans, and both of them are 100% married to Linux because of kernel-specific implementation stuff. Jails and bhyve are awesome, they're much more mature than LXCs and Docker, but such is the environment we live in. If I were ever hosting static web pages or my own blog, I'd probably use FreeBSD for that.

 

[seri0us]

Buy an ad

 

[DavidS877]

I'm just surprised that anyone uses Cockpit for anything. If I wanted a clicky GUI I'd just run Windows Server. I guess I shouldn't be surprised as people still install a graphical desktop on servers too.

 

[Betonhaus]

I'm just surprised that anyone uses Cockpit for anything. If I wanted a clicky GUI I'd just run Windows Server. I guess I shouldn't be surprised as people still install a graphical desktop on servers too.

To be honest I barely use cockpit anymore, maybe I should just uninstall it to reduce my attack profile. I do use it to quickly reformat and add new drives but they seem to come out differently then if I do it by hand from the terminal

$ lsblk
NAME        MAJ:MIN RM   SIZE RO TYPE MOUNTPOINTS
sda           8:0    0  12.7T  0 disk /mnt/dsk/1
sdb           8:16   0  12.7T  0 disk /mnt/dsk/3
sdc           8:32   0  14.6T  0 disk
└─sdc1        8:33   0  14.6T  0 part /mnt/dsk/5
sdd           8:48   0  14.6T  0 disk
└─sdd1        8:49   0  14.6T  0 part /mnt/dsk/6
sde           8:64   0  12.7T  0 disk /mnt/dsk/4
sdf           8:80   0  12.7T  0 disk /mnt/dsk/2
sdg           8:96   0 223.6G  0 disk
└─sdg1        8:97   0 223.6G  0 part /mnt/backup
sdh           8:112  0 931.5G  0 disk
├─sdh1        8:113  0   100M  0 part
├─sdh2        8:114  0    16M  0 part
├─sdh3        8:115  0 930.9G  0 part
└─sdh4        8:116  0   509M  0 part
sr0          11:0    1  1024M  0 rom
nvme2n1     259:0    0 238.5G  0 disk
├─nvme2n1p1 259:9    0   100M  0 part
├─nvme2n1p2 259:10   0    16M  0 part
├─nvme2n1p3 259:11   0 237.9G  0 part
└─nvme2n1p4 259:12   0   509M  0 part
nvme0n1     259:1    0 953.9G  0 disk /mnt/wrk
nvme1n1     259:2    0 465.8G  0 disk
├─nvme1n1p1 259:3    0     1G  0 part /boot/efi
└─nvme1n1p2 259:4    0 464.7G  0 part /

All the drives that don't have a sdX1 partition were set up through the terminal. The next time I add drives I'll take the opportunity to reformat sdc and sdd to match up better.

 

[Dread First]

I'm just surprised that anyone uses Cockpit for anything. If I wanted a clicky GUI I'd just run Windows Server. I guess I shouldn't be surprised as people still install a graphical desktop on servers too.

Homie, Cockpit runs as a LAN web service. Ubuntu Server don't even ship with X11, Wayland, or anything else along those lines. You just download Cockpit from the repositories, enable it via systemctl, and then log in with your home server's LAN IP @ port 9090 using any web browser. No X11/Wayland dependencies, no GTK, no QT, it's just a local web service. All the services I run on my home server are ultimately graphical... on the client side using a web browser. Server side is still entirely text-based even if I plug in the DisplayPort cables on my home server.

I ain't a true text console chad; I'm just a peasant who's finally dabbling with home server stuff and seeing where the road takes me. I "know" that pure text console CLI is the "objectively" superior choice to GUI anything... but you try watching Cantonese arthouse movies in a text console instead of a graphical web service like Jellyfin. Will it work? Probably. Is it pleasant? Not until I get acclimated.

To be honest I barely use cockpit anymore, maybe I should just uninstall it to reduce my attack profile. I do use it to quickly reformat and add new drives but they seem to come out differently then if I do it by hand from the terminal

I just use it for at-a-glance system info like metrics, journalctl logs to troubleshoot, and rattling off commands to update the system and maybe adjust docker-compose.yml stuff should the need arise. The attack surface reduction is a valid point in favour of uninstalling Cockpit, but I doubt you'll seriously get any benefit from doing so unless you've got ports on your LAN directly exposed to WAN or anything else along those lines.

 

[analrapist]

I "know" the "correct" answer is that GNOME Terminal with SSH is more efficient.

GNOME Terminal and a lot of modern consoles, you may as well just use a fucking browser, frankly. A lot of the solutions today grow ever closer to just sticking Electron underneath the piece of shit and calling it a day. This is why us old console hags get so autistic about our terminals, because they say a lot about the feature sets that we find indispensible. (People look at me writing C# .NET on my lightly patched ST without even scrollback and correctly deduce that I'm some kind of Lovecraftian nightmare beast who thinks their program-waifu is shit.)

A lot of console-factionism comes down to saying "I don't need that" at features because other features fill them in adequately. When I started Linux, I renamed everything in a file browser like, aw fuck, I don't even remember the popular ones. These days I use PCManFM, though Thunar was better in a certain pathological usecase where I was trying to to a visual evaluation of ~300k images in a folder. These days, I even do file management in the shell. That takes getting familiar with wildcards. (PROTIP: echo test*filter*ext will let you "debug" using weird splatty wildcards.)

But I also don't give a shit about Scene garbage in my filenames, either. The last video I watched has the filename: /media/server/video/anime/onepunchman/s3/one-punch.man.s03e12.1080p.web.h264-sensei.mkv. Uploaded the torrent via transmission (GTK, in a concession to UI), it completes, I snatched it with rsync -Pa e:/srv/app/seed/misc/*Punch*e12*/*.mkv. All that deep Queearr autism just looks like black magic to me.

 

[DavidS877]

All the drives that don't have a sdX1 partition were set up through the terminal. The next time I add drives I'll take the opportunity to reformat sdc and sdd to match up better.

Personally I always partition drives. Just feels "cleaner". Also on my RAID systems I'll chop a bit off the end so if I have to replace the drive and the replacement is a few sectors short it will still work.

Obviously this is one of the downsides of GUI tools that they'll often do what they want instead of offering all the possible options which would make them impossibly cluttered for noobs to use.

 

[Ahriman]

Southern Linux Man (allegedly BBC enjoyer) weighs in on Wayland.

 

[August Levasseur]

Southern Linux Man (allegedly BBC enjoyer) *ways in on Wayland.

ftfy

 

[unaboomer]

I got burned hardcore when fiddling with Fedora

I've been bit by Pozman too. I've also been bit too many times by SELinux and FirewallD that I just turn that shit off as standard procedure now. The deal breaker for me wasn't Fedora's weird sugar-free substitutions, but its short 8-month lifespan. I'm not IaC-maxxed to the point where rebuilding once a year is anything less than a pain in the ass. I still stayed on the Red Hat plantation like a RHiggercattle though, and went with Rocky Linux for its long shelf life instead.

Far easier and much less intrusive to make a Jellyfin server with assorted goodies than it is to convert my previous gaming PC into a single-purpose network firewall router appliance thingamajigger with pfSense or OPNSense. I'll save that pipe dream for when I live alone and have the freedom to fuck with the network without borking everyone else's WiFi for hours at a time.

I was able to mitigate some disruption to the household by pre-configuring the interfaces to make it as "plug-n-pray" as possible. It's definitely nice to have if your situation allows for it. For instance, you can offload the VPN to your router and obviate the need to set up a client on your sneedbox.

 

[Voloromir of bad Yuristan]

Southern Linux Man (allegedly BBC enjoyer) weighs in on Wayland.

I thank God my autism isn't the one about spending 500 hours customizing a tiling window manager, because I would have to neck myself if I had to use wayland with that.
Personally been using wayland since kde 6 released whenever that was, it usually works fine for what I'm doing and hdr is nice, it's too bad that there's something wrong with my system so I can't use proton ge or cachyos to use hdr, I always have gpu resets with my 7900xtx.