What's your Linux server setup like?

[MC Morrado]

I am currently studying Kubernetes, Docker, and a bit more about Linux through The Linux Foundation and want to know more about how you guys do things in the real world either on your home servers or something else. I plan on using an old PC with an Intel i3 and 8GB of DDR3 to self-host a couple games at a time on Left 4 Dead as a little project to learn more about running a server.

 

What's your Linux server setup like? (operating system, software used to get server running, etc.) and what are okay-ish specs for self-hosting servers for old Source Engine games?

[Donut417]

30 minutes ago, MC.Morrado said:

I am currently studying Kubernetes, Docker, and a bit more about Linux through The Linux Foundation and want to know more about how you guys do things in the real world either on your home servers or something else. I plan on using an old PC with an Intel i3 and 8GB of DDR3 to self-host a couple games at a time on Left 4 Dead as a little project to learn more about running a server.

 

What's your Linux server setup like? (operating system, software used to get server running, etc.) and what are okay-ish specs for self-hosting servers for old Source Engine games?

My Plex server is an i5 3570K, 16 gigs of DDR3, HD7950. I run the desktop version of Ubuntu and just have Plex installed on that. With my media currently siting on a QNap NAS. It works for what I need it for. Part of the reason I run a dedicated Plex server vs running it on my NAS is due to the fact I need transcoding. My plex server is used to watch and record over the air Live TV, so I dont have a say in what format that media is in. 

[MC Morrado]

4 minutes ago, Donut417 said:

My Plex server is an i5 3570K, 16 gigs of DDR3, HD7950. I run the desktop version of Ubuntu and just have Plex installed on that. With my media currently siting on a QNap NAS. It works for what I need it for. Part of the reason I run a dedicated Plex server vs running it on my NAS is due to the fact I need transcoding. My plex server is used to watch and record over the air Live TV, so I dont have a say in what format that media is in. 

What's Plex? And I didn't know it was possible to record live TV on a computer like a VCR.

[Donut417]

Just now, MC.Morrado said:

What's Plex? And I didn't know it was possible to record live TV on a computer like a VCR.

It a media server. 

 

 

This goes in to it a bit, plus it talks about some open source alternatives. The reason I dont use the alternatives is that the guide data for Live TV is provided due to the fact I paid Plex $119 one time fee for the Plex pass. 

[RONOTHAN##]

I have a Proxmox server with a 5900X and 64GB of RAM (I'm looking to upgrade this) in a 2U Supermicro case. There's a Truenas VM with a SAS HBA passed through for NAS functionality, a Jellyfin VM with a Quadro P600 passed through for video encode, a VM dedicated to Docker, a VM dedicated to nextcloud, and a couple other small things. All VMs besides Truenas run Ubuntu Server. 

[AamirA]

I turn truenas scale with all my docker apps in a ubuntu vitual machine in truenas. All the docker containers have their own compose file and I manually update them after reading release notes. The important files of the apps is either on a nfs share from truenas, or replicated by borgmatic to other places on a schedule. If the borgmatic backup doesn't work, it sends me a alert to my phone. I access the machine though a combination of tailscale for stuff only I access and frp running on a digital ocean server proxying my web traffic to my internal caddy server

[MC Morrado]

What I'm trying to do is much more simple on just bare metal. Maybe Ubuntu Server hosting something like Minecraft or Left 4 Dead?

[Bdavis]

TrueNAS scale

[Kilrah]

ITX 7700K with 32GB running Unraid with 2x500GB RAID1'd SSDs for apps and 4x16TB drives for storage one of which is parity protection.

[jfmonty2]

My personal server is a Ryzen 5800X with 32GB of RAM and 12 4TB spinning disk drives (plus an NVMe drive for boot) for a total of 32TB usable storage (ZFS with 2 6-disk raidz2 vdevs). I run a few things on it, including a couple of gameservers (Valheim and Veloren), eventually I will probably end up hosting Minecraft as well.

 

I run a single-node Kubernetes "cluster" which is massive overkill of course, but I wanted to play with Kubernetes so here we are. Single-node Kubernetes is nice, though, because it frees you from the burden of having to figure out a clustered storage solution which is absolutely the Achilles heel of self-hosted Kubernetes.

 

At work I'm responsible for a 3-node Kubernetes cluster where each machine is an Epyc Rome 8-core (I forget the exact model) and 64GB RAM. The storage situation there is a mess, though, I'm using Longhorn but really don't like how it behaves when a node fails, so I'm thinking of looking for something else.

 

Quote

What I'm trying to do is much more simple on just bare metal. Maybe Ubuntu Server hosting something like Minecraft or Left 4 Dead?

Ubuntu server is nice, that's what I typically use. I don't like RHEL-derived distros because I don't like yum/dnf, not to mention their recent shenanigans with CentOS, and I prefer Ubuntu over plain Debian because Ubuntu is a little more relaxed about what they're willing to include (like ZFS), which is useful to me.

 

Your i3 with 8GB should be ok for a basic Minecraft world, although if you do heavy redstone then it could start to chug. Also flying with elytra is surprisingly taxing on a server when it has to generate chunks as you fly.

 

Haven't hosted Left 4 Dead myself, so I can't speak from direct experience, but if you're talking about the 2008 game I imagine you're fine. Older games like that tend to be lightweights, compared to today's games.

[SpaceGhostC2C]

Peasant-grade Ubuntu Mate LTS sharing a RAIDZ2 storage pool through Samba with the rest of the house, plus a "PCie SSD" (it's two NVMes in raid0 on a PCB, but sold as a unit) on its own pool as faster, non-redundant shared storage. That's pretty much it. Might host a LAN Micreaft server in the future.

 

I started with an AM3+ setup, 16GB ECC DDR3, 4x4TB in RAIDZ1. As my computers' life cycle moved on, that's now semi-retired (have to fix it to host some backups) and everything is migrated to my former X399 PC upgraded to ECC memory (TR 1920x, 128GB ECC DDR4, 6x12 TB in RAIDZ2).

I feel better about RAIDZ2 than RAIDZ1, but RAID is not backup so I must get around getting the backup system in working condition (I have plenty free space, so I don't need as much storage for the backup for now).

[Bersella AI]

My server was built upon a refurbished commercial PC from HP, and is running the Linux-based TrueNAS Scale system. Now it hosts a dozen of services pretty well, while keeping efficient in power usage; no virtualization, no web or game hosting, in favor of simplicity. I have introduced it dozens of times here, but don't mind to showcase again.

• Chassis/Motherboard/Power: HP EliteDesk 800 G3 SFF, with 180W Platinum-certified PSU
• Processor: Pentium G4600
• RAM: Unbuffered Crucial DDR4-2400 8GB kits x2
• Storage: Seagate Exos X16 (16 TB), Seagate SkyHawk (4 TB), Intel Optane (16 GB) x3
  Note: No RAID, snapshots & periodic replication are taking care of precious data
• UPS: CyberPower UT650GC

It looks like this:

 

[JJ12415]

I have a Dell PowerEdge R620 (full specs below) running AlmaLinux 9. I use Docker for things like a mail server, a SIEM, Portainer, and a few other random things.

 

I use bare metal LinuxGSM for game servers, and Plex (for now), and Nextcloud. It's a neat little setup, excluding the weird ethernet issues I'm having atm, only real downside is power consumption...

 

R620 specs:

• CPU: 2x Intel Xeon E5-2650 v2 (16) @ 3.40 GHz
• RAM: 112GB DDR3 1866MHz ECC (with 512GB SD for Swap)
• Storage (all HDDs): 2x 1.2TB SAS in RAID 1, 6x 1TB SAS in RAID 6, ~8TB in external exclosures

 

Edit: Forgot Vaultwarden and HomeAssistant docker containers.

Edited October 16, 2024 by JJ12415

[digitalscream]

Main server is an i5-12400F with 128GB RAM, 12 x 1.6TB Intel DC SSDs (ZFS) and an RX 7800 XT with Ubuntu 24.04 installed, running media services (Rygel), Minecraft and Ollama/Open WebUI Docker images) and file services. All machines on the network backup to it as an archive (as does the forum I run), and then the offsite backups run from there. It's probably going to get a second GPU at some point, because 16GB isn't enough VRAM to run the models I need.

 

Secondary dev server is a Xeon E5-2650 v2 with 64GB RAM (needs an upgrade), 500GB boot drive and a 2TB Intel PCIE SSD for VM images, running ProxMox because life's too short for dealing with VMWare. Currently use it for testing deployment strategies for work, not sure what I'll do with it once that's finished.

 

Main server is always-on, dev server is only run on-demand.

[BabaBakChodiWale]

I currently have a I3 3220 16 gb ddr 3 and a h61 mother board with 4 sata ssd 2x1TB Curcial bx500 and 2x120 wd green running proxmox as hypervisor with 4 VM's and 2 LXC Container's in total.

one of the VM is running my website, another one is running my Nextcloud server and
LXC container's running Cloudflired-tunnel so that i can access the Proxmox Web UI form internet,
and another Container and a VM to try new things.

all the VM and Containers are using static Public ipv6 i am only using Cloud-Faired tunnel for the Proxmox Web UI

as its a home server i have also enabled "Restore power on AC" and also have enabled "start at boot" setting for all the VM's and Containers so if i ever loose power which happens frequently. the server just comes back on again

I am planning to upgrade it now trying to find a cheap i7 3770 around so that i can have intel VT-d support for PCIE passthrough since i have a 10 gig NIC which i acquired for free from my job place. but it is not working when ever i put this card my server just don't start at all no fan spins nothing. as if the system is dead and the moment when i pull the card out it starts working again I have already checked the card and it is working in other servers at my company but not at home. can any one help me with this??

I have also tried replacing the damn thing the replacement card also did the same. 

It is a wired 10 gig card which has 4 cuts in its 16x pcie connector. i have also attached the image if any one can help!!

[Bersella AI]

9 hours ago, BabaBakChodiWale said:

It is a wired 10 gig card which has 4 cuts in its 16x pcie connector. i have also attached the image if any one can help!!

I think it should work at PCIe 2.0 x8, and have special pins near its x8 interface, which makes it incompatible with regular x16 slots seen on an H61 board (likely to have only one x16 and one x1 slot). You have to search for a premium motherboard with a second x16 slot that is electrically wired as x4, i.e. no more pins can be found beyond x4 configurations, so that it would not electrically contact the special pins. Here's the example of such a slot, at the bottom of ASUS TUF B760M motherboard.

 

[BabaBakChodiWale]

7 hours ago, Bersella AI said:

I think it should work at PCIe 2.0 x8, and have special pins near its x8 interface, which makes it incompatible with regular x16 slots seen on an H61 board (likely to have only one x16 and one x1 slot). You have to search for a premium motherboard with a second x16 slot that is electrically wired as x4, i.e. no more pins can be found beyond x4 configurations, so that it would not electrically contact the special pins. Here's the example of such a slot, at the bottom of ASUS TUF B760M motherboard.

 

Thanks this seems promising. I will try this fix 

[digitalscream]

On 10/16/2024 at 4:53 PM, digitalscream said:

Main server is an i5-12400F with 128GB RAM, 12 x 1.6TB Intel DC SSDs (ZFS) and an RX 7800 XT with Ubuntu 24.04 installed, running media services (Rygel), Minecraft and Ollama/Open WebUI Docker images) and file services. All machines on the network backup to it as an archive (as does the forum I run), and then the offsite backups run from there. It's probably going to get a second GPU at some point, because 16GB isn't enough VRAM to run the models I need.

 

Secondary dev server is a Xeon E5-2650 v2 with 64GB RAM (needs an upgrade), 500GB boot drive and a 2TB Intel PCIE SSD for VM images, running ProxMox because life's too short for dealing with VMWare. Currently use it for testing deployment strategies for work, not sure what I'll do with it once that's finished.

 

Main server is always-on, dev server is only run on-demand.

Turns out, things move fast round here. Needed to spin up rather more instances for the work deployment testing than the old 8-core machine could handle, the deployments were timing out while the instances were building.

 

Dev server is now an old Fujitsu server I had lying around in the shed - 2 x 10-core E5-2470 v2 CPUs, 256GB RAM. The beauty of Linux...just pulled the drives from the old one, shoved them into the new (well, older) one, and it just launched everything exactly as before with vastly more resources.

 

Seriously tempted to chuck in both servers and build a monster of a Xeon Gold or EPYC machine. Resisting at the moment, on the grounds that it's probably just my "buying more stuff will be fun" instincts betraying me.

[MC Morrado]

10 minutes ago, digitalscream said:

Turns out, things move fast round here. Needed to spin up rather more instances for the work deployment testing than the old 8-core machine could handle, the deployments were timing out while the instances were building.

 

Dev server is now an old Fujitsu server I had lying around in the shed - 2 x 10-core E5-2470 v2 CPUs, 256GB RAM. The beauty of Linux...just pulled the drives from the old one, shoved them into the new (well, older) one, and it just launched everything exactly as before with vastly more resources.

 

Seriously tempted to chuck in both servers and build a monster of a Xeon Gold or EPYC machine. Resisting at the moment, on the grounds that it's probably just my "buying more stuff will be fun" instincts betraying me.

What if you took all the servers you have and run them as one big Kubernetes cluster so the load is distributed evenly?

[digitalscream]

23 hours ago, MC.Morrado said:

What if you took all the servers you have and run them as one big Kubernetes cluster so the load is distributed evenly?

Mainly because I don't want to be running all of them all the time - these aren't exactly lightweight machines, and electricity isn't cheap here in the UK at the moment. I keep the main machine running all the time, but it's surprisingly power-efficient - even with the 7800 XT in there, it idles near 40W. The others...double that, if not more.

 

Ergo, the main server stays on all the time, and I just need the ProxMox machine while I'm using it for dev testing for a few hours out of the day. Not only that, the testing I've been doing so far has proven that I can rebuild the entire machine in about half an hour from nothing (including the OS) if I need to - the data's important on the main server, but the dev server is entirely ephemeral, proof-of-concept stuff only and I don't really care about having the stuff on it being always-available; if I need it, it's just a question of walking to the back room and switching it on.

 

I should stress that this isn't really a homelab-type setup - it's all based around specific requirements.

 

That said, I've got an idea for a smart/magic mirror project based around a Pi, backed by AI services run on the main server, whereby it'll need to run different personalities based on who's standing in front of the mirror. I strongly suspect I'll need to add another GPU or two to the main server...and that means I'll be running out of PCIE lanes. If that happens, then I may just end up doing the EPYC upgrade...and at that point, it'll probably be sensible to just spend a bit more and add enough memory capacity that I can do everything on one box whenever I need a few extra VMs running.