Backlit edition

previous: >>105572884

READ THE WIKI! & help by contributing:
https://wiki.installgentoo.com/wiki/Home_server

/hsg/ is about learning and expanding your horizons. Know all about NAS? Learn virtualization. Spun up some VMs? Learn about networking by standing up a OPNsense/PFsense box and configuring some VLANs. There's always more to learn and chances to grow. Think you’re god-tier already? Setup OpenStack and report back.

>What software should I run?
Install Gentoo. Or whatever flavor of *nix is best for the job or most comfy for you. Jellyfin/Emby/Plex to replace Netflix, Nextcloud to replace Googlel, Ampache/Navidrome to replace Spotify, the list goes on. Look at the awesome self-hosted list and ask.

>Why should I have a home server?
De-botnet your life. Learn something new. Serving applications to yourself, your family, and your frens feels good. Put your tech skills to good use for yourself and those close to you. Store their data with proper availability redundancy and backups and serve it back to them with a /comfy/ easy to use interface.

>Links & resources
Cool stuff to host: https://github.com/awesome-selfhosted/awesome-selfhosted
https://reddit.com/r/datahoarder
https://www.labgopher.com
https://www.reddit.com/r/homelab/wiki/index
https://wiki.debian.org/FreedomBox/Features
ARM-based SBCs: https://docs.google.com/spreadsheets/d/1PGaVu0sPBEy5GgLM8N-CvHB2FESdlfBOdQKqLziJLhQ
Low-power x86 systems: https://docs.google.com/spreadsheets/d/1LHvT2fRp7I6Hf18LcSzsNnjp10VI-odvwZpQZKv_NCI
SFF cases https://docs.google.com/spreadsheets/d/1AddRvGWJ_f4B6UC7_IftDiVudVc8CJ8sxLUqlxVsCz4/
Cheap disks: https://shucks.top/ https://diskprices.com/
PCIE info: https://files.catbox.moe/id6o0n.pdf
>i226-V NICs are bad for servers
>For more SATA ports, use PCIe SAS HBAs in IT mode
WiFi fixing: pastebin.com/raw/vXJ2PZxn
Cockpit is nice for remote administration

Remember:
RAID protects you from DOWNTIME
BACKUPS protect you from DATA LOSS

>>105601412
>*technically* you can clone the MAC address of the ONT
assuming they only authenticate based on MAC address, they have other options besides that.

If I'm keen on running some some vidya servers, is Pelican still the best option for this? Or what's better?

Not (just) Minecraft servers, but other items as well.

>>105601497
true, depends on your ISP. YMMV.

>>105601509
YMUMUHHHHV!!

>>105601512
how are you planning on connecting 4x SATA HDDs to a beelink s12 which has no SATA expansion slots (or pcie slot to put a SATA card in)?

>>105601484 (OP)
>dual socket board
Nice

>>105601484 (OP)
Repost of >>105601125
Still trying to figure out the hardware / software for an iSCSI+NFS hybrid NAS:
>HDDs:
I'm settling on the new X24 (https://www.amazon.de/dp/B0DCZ6GXGR)
But I'm not sure whether to go all in with 4 drives or go with 2 drives.
Also not entirely sure about the Filesystem I should use, if possible I'd use ZFS in Mirror VDevs which are then striped. "ZRAID10" if you will.
>Optane:
Firm on the 750GB P4800X (https://www.ebay.de/itm/156183154122), I would love to use P5800X (https://www.ebay.de/itm/396417058881) but I'm not paying 10x for 50GB more capacity + PCIe 4 and even if, the 1.6TB version is cheaper per GB (https://www.ebay.de/itm/356727758876).
ZFS would be nice here as well as L2ARC / Metadata devices, if I can't use ZFS I'll just use bcache like I'm doing with my Desktop right now.
>SSDs:
Still on the fence, because of the mainboard / PCIe lanes.
>10G ethernet:
Same as above, still on the fence between the X550-T2 (https://www.ebay.de/itm/166643341272) and the X540-T2 (https://www.ebay.de/itm/196139352813). I think I'll go with the X540-T2 if I can find a cheap mainboard+cpu+ecc kit with a lot of PCIe lanes.
>Mainboard:
If I can actually bifurcate all PCIe lanes, I'll keep the X570 and use something like this as a CPU: https://www.ebay.de/itm/175275999106

Are my ramblings somewhat alright or am I barking up the wrong tree?

>>105601931
this entire setup is retarded, find some compromise
i'd start with dropping optane

>4 disk raid 10

>>105602071
No caching then?
>>105602090
Should I Raid 5 them instead?

>>105601564
USB enclosure for now since this is just for the media server. Frigate uses internal storage only since my retention rates are pretty strict. If it becomes a problem, then I did see online somewhere that someone bought an m.2 SATA adapter for their intel NUC. I suppose I could do that and leave the top off the beelink.

my home "server" has
>samba share
>qbittorrent
>jellyfin
>rclone backups
>internet-pi monitoring
Is there anything else that's useful or is this it?

LSI SAS 9300-16I for 50€
yes no?
its in it mode by default.

>>105602204
4 drives via USB?

>>105602090
I do that, it's great. 4x read speed, 2x write.

>mirrored mirrors of RAIDZ3 pools made up of mirrored RAIDZ3 pools

>>105602332
>jellyfin
>useful

>>105602332
I would say don't necessarily go crazy with adding stuff to your server. Let it come organically as you use it and you run into things you wish you had or wish you could do.
Sometimes I'll add something just to test it out and see if I like it, but most of my additions came after experiencing something and realizing I needed a solution instead of a momentary bandaid.
That's how I ran into adding DNSUpdater to my server, for example, after my public IP kept changing and screwing up my domain records, but I didn't want to pay for DDNS.

>>105601497
Can you list some? Even when I change my MAC address ISP gives me a 172.x.x.x address while the ISP provided cuckbox gets the real address.

>>105602332
autobrr, sonarr, radarr, whisparr, flexget or whatever is relevant for getting whatever you watch on jellyfin.
vpn client?

>>105602765
No bonarr?

>>105602766
bonarr is dead and replaced by whisparr...

anyone tried running the *arr suites through i2p?
i wanna give that a try since i heard torrenting speeds are actually acceptable nowadays.

Would your rather have your shit in a datacenter with 10gig access, or run everything locally at home with only 1 or 2gig access?

>>105602779
Bonarr was an actual thing? I was just making fun of the silly rr names.

>>105602765
>>105602766
If you go *arr I would add to get prowlarr so you don't go crazy managing all the separate *arr torrents.
Jackett was okay but I found Prowlarr to be better.

>>105602805
at home.
that being said. 10gig is accessible at home too.
a 10gig sfp+ nic costs like 20 bucks.
10g switches are in the 100 dollaridoos range nowadays.

>>105602832
I'm talking WAN not LAN speeds.

Assuming 10Gbps LAN and 1 or 2gbps WAN, vs a remote server with 10Gbps WAN, which would you prefer?

>>105602689
for ONTs?
serial number matching or logical identifier (LOID) auth are the other main kinds

>>105602826
I use it but all it does for me is manage indexers. I assume that's what you mean.

>>105602850
well you would need to have 10gig internet to take advantage of either locations or do you want ythe remote server to be used by lots of people?
if its just a few people then home all the way. if it is however some sort of platform for a lot of people let say 100+. then a remote server is better.

>>105602815
I download polish movies with Pizdarr

>>105602869
It would be me + 3 or 4 others only, so I'm leaning towards just hosting everything at my home.

>>105602805
i couldnt get away without ipmi so easily if my shit was in a datacenter
still, assuming the colo bill is the same as it costs to run my shit at home and the dc is within reasonable driving distance I'll pick that option

>>105602887
Cost isn't a huge factor. Datacenter WOULD cost more, but not by a massive amount, and while it's physically close (10-15 miles), the actual driving time is closer to 60-90 minutes due to roads and bridge access to cross the river.


I think I'll probably upgrade to 2gig internet at home and host locally.

>>105602850
NTA but I'd rather at home. WAN speed doesn't really make a difference when you aren't saturating a gb.
Actually having the hardware with you in your home is some of the fun.

The idea your shit is somewhere else also isn't a good feeling. Its doesn't feel private even if you are the only one with access.

>>105602919
>WAN speed doesn't really make a difference when you aren't saturating a gb.
I saturate 2gbit routinely. Though generally not for more than 10-20 minutes a day.

>>105602919
Yeah thats where I'm leaning, the only time I could even realistically need more than 1 or 2gbps would be if everyone I expect to use the server were ALL trying to download from it simultaneously, and even then it would at most be ~4-5Gbps, MAYBE 6gbps if some of us upgraded to 2gbps home internet, and even then the odds of all of us trying to access downloads simultaneously is pretty low.

>>105602866
Yeah. It doesn't even matter for me that much because I also grabbed a usenet subscription and set all of my *arrs to prefer to download from usenet first. Its much faster and more reliable. I'll select a movie on Ombi and by the time I bring Jellyfin up its already in the library, its that fast.

>>105601507
Why not just make a VM and install your dedicated server like a normal person?

>>105602942
I was more thinking of the 95th percentile useage, which is no where near a gb.
Its easy to saturate a gb or two intermittently and 10gb might mean you get things faster
But you also aren't losing anything with 1-2gb, things just take slightly longer to download on something that is probably pretty short already at gigabit speeds.

>>105603025
NTA, but I downloaded a 140GB game in about 13 minutes last week. Even with faster WAN connection I don't think the server I was downloading from had too much left to offer.

>>105602990
for me I definitely need at least 1 torrent tracker in addition to usenet.
>inb4 why don't you have emp
you haven't invited me yet ;)

>>105603040
Of course that's another thing, your not going to get peered much faster with 10g
Maybe a colo has better routes so you can get more speed but I doubt its that big of a deal unless your ISP's internal network is shit.

>>105602862
OpenWrt wiki says my ISP should not do any of this, not even MAC cloning. Any idea why it I'm getting a private IP address?

>>105603083
If there isn't any public information about your ISP and how they do their authentication then your best bet would be to call them up and try to talk to one of the network engineers as the regular technical help at call centers wont know what the fuck you're talking about.

>>105602571
Huh, so these modern mini-PCs can really be that power efficient? Were you measuring that at the wall, or does it have some power measurement sensor built-in?

>>105602090
Try 10 disk RAID-10

>tfw you call up your isp and the support agent knows more than you
she sent an email with the pppoe credentials while i had next to no idea what i was even asking for

cold day in hell before i switch from this isp, they're great.
good luck speaking to anyone who knows anything besides "turn it off and on again" at a big isp

>>105603231
if your ISP uses PPPOE then i'd expect the call center to know what it is, they're used to giving people that information and explaining what it is.

If your ISP uses a whitelist or some other identifier tied to their supplied equipment, then their call center has no training on authentication methods and what is required to authenticate with non-ISP supplied equipment.

>>105603258
>If your ISP uses a whitelist or some other identifier tied to their supplied equipment, then their call center has no training on authentication methods and what is required to authenticate with non-ISP supplied equipment.
If they do that then the PPPOE details are usually irrelevant in the first place because they already know how you are based off of the ONT. My ISP publishes credentials but they actually don't matter, it will accept anything you throw at it.

>>105602104
>>105602119
Got it, thanks anon.
So then if I have an L3 switch but also an OpenWRT machine, should I just avoid using the L3 features as a general rule of thumb? Seems kinda weird then for enterprise switches, the kind that can probably expect to saturate a substantual chunk of their 24 or 48 port capacity, to also be the ones most often including L3 features if those are going to be bottlenecks and better served by a dedicated router (something an enterprise rack can also likely easily afford to fit).

>>105603319
>Seems kinda weird then for enterprise switches, the kind that can probably expect to saturate a substantual chunk of their 24 or 48 port capacity, to also be the ones most often including L3 features if those are going to be bottlenecks and better served by a dedicated router (something an enterprise rack can also likely easily afford to fit).
It's not that you definitely will saturate it, it's just that I'm assuming you are buying cheap L3 switches which almost certainly will run into issues when routing/forwarding traffic. They only include these features to tick a box most of the time.

Switching is fast
Routing traffic is not
When possible always put your machines on the same network segment so you don't have to route the traffic but people like their VLANs, etc, which sometimes necessitates the mess of intra-VLAN routing, etc.

>>105603386
Hmm, makes sense. Again though L3 switches seem almost exclusively targeted towards the enterprise market and come with a big upcharge over non-L3 capable ones so it seems surprising that there's market demand for ones that just "tick a box", but if that's what it is then makes sense.
>you definitely will saturate it
Oh I definitely won't be saturating anything, my home network is tiny and I'm buying a 12-port switch anyway. I'm just trying to understand the intended use and the best practices in theory.

>>105603073
>NZBGeek
Yup, that's what I went with as well. Really strong index and its rare I don't find what I want from there.

>>105602815
cute fox. My server is keeping backups of her.

Thoughts about privileged containers?

thoughts on webdav+syncthing as a Dropbox replacement?

>>105603699
just do rclone on both ends?

>>105603699
>syncthing
good software. Dunno how well it will work for live collaborative work which is what you are trying i think but i can vouch for syncthing.

>>105603258
>if your ISP uses PPPOE then i'd expect the call center to know what it is, they're used to giving people that information and explaining what it is.
anon most isps expect 99.99% of their customers to only change the passwords on the provided wireless router, if that.
I get that its different in the USA where people have motivation to avoid a $40/month rental fee but we don't have such cuckery. I think those practices might be illegal here.

>>105603728
>>105603765
i want to share a synced dir among computers and phones, i thought just have webdav and syncthing working on the same directory. what does rclone offer?

>>105603781
>i want to share a synced dir among computers and phones
That is syncthing's usecase, yep.

>>105603781
server

>>105603781
>>105603818
rclone mount works with most of those as a client. even on wangblows.
but windows has a built in webdav client.

>>105603083
you might be cgNATted

>>105603877
I'm not behind CGNAT and ISP provided router gets the IP address correctly.
>>105603115
I guess I'll have to do that, not sure if even they'd know.

>>105602805
Locally obviously. What possible need could having my "home" server in some random data center? The entire point is that it's local.

>>105601484 (OP)
Am I getting ripped off paying ~$10/TB for "Refurbished" drive? Every thing I seeing that's cheaper either have bad return policies/support or is being sold as is.

>>105604270
12 and under is good

raids should always use the same model of drive because the slowest one will bottleneck all others right? I have two drives and I'd like to get another one so i can switch to RAID 5 from RAID 1 so do i just get the same model even if the others are used? Does that have an impact on anything and should you be a puritan or is just matching the model enough?

>>105604579
Is 12€/TB for brand new drives good?

>>105604769
As long as you get the same speed (e.g. 5400RPM vs 7200RPM), type (SATA vs SAS), and technology (CMR vs SMR, cache vs no cache, number of heads) it's fine. It doesn't need to be the exact same model.

>>105601484 (OP)
The office I work at has some "e-waste" PCs form the Win 7 days. Would they be good for a home server? They are so old I'm worried I could have problems attaching the amount of drives I want. I'm looking for a cheap NAS/Syncthing node.

>>105604769
RAID should be the same specs, not model, bought from different manufacturing batches. Don't go to the store and buy 4 of the same. They'll all fail on the same day.

>>105604769
>the slowest one will bottleneck all others
not always but usually, yes
>o i just get the same model even if the others are used? Does that have an impact on anything and should you be a puritan or is just matching the model enough?
just match the capacity and sector size* and don't buy drives with worse specs.

Any advice on buying drives and used ones specifically? So i just enter whatever size and i speed i want into ebay, check the listing to see if its working and check the model for yearly failure rates? I can see some drives that are faster and a lot cheaper than my current ones per tb and i can't say im not tempted to get some.
And some say that they have defective sectors but that "its fine ;)", i should avoid them right?

>>105605310
just buy the shit with a warranty
if they dont offer a warranty walk away

How do I router port forward using the device Hostname (not its local ip address)?

I've set the device to DHCP so the ip address can change and I don't want to set a static ip (don't ask why)

>>105605310
Depending on your raid level, and the price of the drives, it could be worth it. I'd say go for a wide raidz2 or raid6 if you're going to be dabbling in known bad drives. Remember to do a long smart test on them as well, and get a return if they're not as advertized.

>>105601484 (OP)
Is there a gallery software for auto image classification? For example I want to be able to search for picrel like "reaction". Even only OCR is fine.

>>105601484 (OP)
Is there a more autistic way of doing things like jellyfin?Like instead of using some web interface and an app you like Remote Desktop to view something or is that pointless/impossible

>>105605577
You should be able to make a "reserved" ip on the dhcp server so your ip don't change even if your computer has dhcp enabled.
Even cheap isp modems has a way to enable this.

>>105605575
Its what i did for my first two drives but i feel like they are being wasted in raid1 so i want to get them some friends. But some used ones cost less than half of them so its something to consider.

>>105605596
Same thing i thought. I still have to do any actual math to compare long lasting drives with warranty to cheaper less reliable ones but if its in a proper RAID what is there to loose? You should do backups anyways.

>>105605626
I know but I just cant be bothered with that as it adds extra administration. I will need to reserve it based on the mac address of the device. And Ideally dhcp mapping isn't something I wanna used over setting it in the device static IP instead

>>105605664
But dhcp reserve is the ideal scenario, 1 minute or less and you don't have to deal with it for months or years.
With static ip assigment on the device you can cause collisions and conflicts if another device request that ip.

>>105605602
you could do that but it'd be strictly worse to use

>>105605645
You should have backups. Personaly I do raidz2 because I don't like the idea of spending a week rebuilding the entire pool from a backup. Even if it takes a week to resilver a drive, the entire pool is still useable during that time, and still has redundancy.

>>105605691
Don't bother arguing with him he is beyond hope if he thinks that DHCP is a chore he can't be bothered with.

Next we'll be hearing about the evils of DNS.

>>105605664
>>105605577
You have three options.
1) Set a static IP, ideally outside of your DHCP range
2) Add a DHCP reservation for your device
3) Leave it with a dynamic IP address, which means your port forward will break if the IP address changes.

Stop being a lazy shit. Spend 30 seconds adding a reservation or setting a static IP.

>>105605577
I will never understand why port forwarding can't be done with host names. Who the fuck still uses static IP in LAN? That's for public sites.

>>105606052
NAT doesn't understand hostnames, nor should it.

>>105606060
Learn to reading comprehension, ESL-kun.

>>105606114
I can read perfectly fine, but you should stop talking about things you're clueless about.

>>105606120
Suck my fat TCP stack nigger. I haven't had to assign static IP in over ten years. As far as port forwarding goes, I have written a custom script that maps magic IP addresses to hostnames and resolves them as real IP addresses so I never have to bother with static IP assignments. Are you even a real network admin?

>>105606243
>>105606114
>chatgpt pajeet is back reciting from gpt conversations and calling everyone who disagrees a nigger
Shit thread.

>>105606243
I just finished writing a script that builds an inventory of 4k+ MPLS SIDs, verifies uniqueness, and then verifies that local eVPN prefixes are advertised correctly to core routers via MP-BGP. But go off, I guess.

Port forwarding is NAT+ACL. NAT doesn't work with hostnames because FIBs don't use hostnames. They're not going to either, and they shouldn't.

>>105606280
*node SIDs
*RIBs/FIBs

>>105606052
Because assigning names to addresses+ports (services) is an application layer concept. You'd still need some lower-level in the stack proxy like Haproxy that understands these higher level concepts and can do SNI inspection and proxy to some other device based on this.

>>105605597
Immich
https://immich.app/

No server expertise here, so bear with me.
I was thinking about repurposing some old hardware for a home server with proxmox, a rarr stack and jellyfin, small NAS, small web servers with databases that only need to be up most of the time, not 99%.
I'll still need to buy some storage devices and don't know which filesystem to use yet.
If i get a large HDD for my media files and a smaller one for my NAS and server data, can I get a RAIDZ1 with ZFS that mirrors the smaller HDD onto part of the larger HDD and have the rest of the large HDD just be media files? So a 1TB drive and a 8TB drive and the 8TB drive is divided into a regular 7TB logical volume and a 1TB logical volume that mirrors my small physical 1TB drive.
Will ZFS allow me to do that or should i just use LVM.
Also i was gonna get 2x 128GB consumer NVMe for the boot drives. Idk if that is fine.

>>105606669
128GB is more than plenty for Proxmox itself, but you probably want to put your guests' OS on SSD as well. Not sure what OS you're planning to use on the guest VMs/containers, just make sure you have enough storage for that.
As for the large storage, you can set up ZFS on them and then use mountpoints if you want the large storage for your guests to be directly accessible from Proxmox.
I also would not recommend RAIDZ for performance reasons. You'll get better performance out of a mirror and there's less risk of a second failure during rebuilds as well.

>>105605691
ok what if I connect a unmanaged switch to the router, then the router goes down but the switch is still up

in this scenario I cannot connect to device 1 with reserved DHCP, because it relies on the router for the IP address, but if I set a static IP then I can.

Static IP are more reliable

>server went down
>manually restart it
>it's not booting into the os, boot drive not detected
>shit, crappy kingston ssd must have died.
>need to go inside the rack, pull the server out and remove the ssd, which requires disconnecting the internet because of the way the cabling is setup
>test ssd on my pc to make sure its dead
>it works
>wtf
>mfw already ordered a replacement ssd anyway
>in the process of removing the server, pulled the pins off the header of my fans
>spend 30 mins trying to fix it. had to end up taping it to keep the pins touching
>internet stopped working
>in the process of removing the server, i somehow snagged the power cable of the router
>plug it back in
>internet still not working
>ughhhhhhhhhhhhhhhh
>in the process of removing the server, i got the lan and wan cables to the router mixed up
>fix it
>put old ssd back in server to see if maybe it'll work again
>it works
>just wasted 6 hours doing literally nothing
LOL

>>105606805
thanks for the info. can i mirror the small drive onto part of the large drive with ZFS though? if so, will that require any special setup or will it work like just creating 2 partitions on the large disk and selecting one as the mirror?

>>105606945
I have no less than 3 different machines that occasionally throw random, unpredictable boot/cpu errors that always go away after a cold boot or manually selecting the boot media in the bios. You aren't alone.

>>105606951
I had missed that part. I don't think ZFS supports anything like that.

>>105606970
I've noticed the boot problem tends to happen when my server has been powered off for a long time(multiple days). I keep it off if im on vacation. This isn't the first time it has happened. Perhaps it's a sign of a failing SSD. Event logs in the bios showed there's definitely something wrong. (Not my picture, but same error).

I'm more annoyed at the other issues that happened because of how cluttered the inside of my rack is. I regret buying a 15U rack. I should've just went for 42U. It's a nightmare doing any kind of troubleshooting and I can't effectively organize it due to the sheer amount of shit inside it.

>>105607004
Forgot error pic

>>105607004
>I've noticed the boot problem tends to happen when my server has been powered off for a long time(multiple days). I keep it off if im on vacation. This isn't the first time it has happened.
Same situation here for at least 2 devices, one is retired though. I'll go fishing in my bioses for the errors later, but I wouldn't be surprised if it has something to do with the boot media as well.

How important is HDD cache size for a media server? I'm looking at some drives and I could get the Exos E with 256MB cheaper and from a better vendor than the Toshiba MG10-D with 512MB. is the upgrade worth it?

>>105606970
>work for company with hundreds of shit ass old servers that will just randomly die on boot/power off
>a couple dozen a week will have their RAID card lock up requiring a reboot to function again
>cold boot just increases chances of complete failure
>"""temporary""" replacement has a RAID0
It can always be worse

>>105607155
Yeah that's unspeakably fucking grim, I'll count my blessings.

>>105607140
I don't think it really matters.

>>105607140
Anything more than 128MB is fine in 2020+.

>>105606669 (me)
>>105606951
>>105606992
I might just get 2x 2TB drives for the data i want mirrored then.
how much does ZFS really eat through SSDs? they're basically the same price/TB than prosumer HDDs right now.
the only other options would be refurbished enterprise HDDs. those cost like half.
what's my best pick here? i think they'll mostly be used for NAS.

>yt-dlp + rsync + navidrome
I will have the Spotify slop experience for free so help me God.

>>105607430

Anyone use something like Xubuntu for a server? I've been trying to use Debian but apparently the Dell PC I got hates Debian because it needs non-free firmware that won't install despite Debian's wiki saying Debian 12 should come with them and install automatically.

>>105607488
Checked. I have happily used Xubuntu as both my main os and as my server's os for years. The server features:
>zsh
>zfs (1 raidz2 pool of 6 HDDs for storage and 1 stripped mirror (think raid 10) pool of 8 SSDs for streamed services - all the hardware's enterprise grade)
>nfs with autofs to deal with mounting
>grsync built on top of that, with the ability to rsync via bash remotely
>smbd for the gaymen machine
>plex (no, you don't need to pay, just use a vpn)
>deluge daemon/web service for torrents, which feed into the pools directly
>audiobookshelf
>ipmi/kvm support because I have a server mobo
>ssh/sftp (how I set everything up)
>tailscale (don't forget to self sign users/devices/2fa/config access list)
>nextcloud
>navidrome
It's fucking gas. I've run Xfce on my main machine ever since unity "dropped", and I've never looked back. Canonical deserves no small amount of criticism and makes plenty of retarded decisions, and snaps are generally gay, but if you want something you can run on a rock (I started ages ago on am3 gaming mobos like Asus m5a97 r2.0/900fx r2.0) that can be configured by anyone who can follow a guide with some copy+pasting terminal commands with solid support for both cutting edge and ancient trash, look no further.

>>105607488
forgot to add here >>105607586 that I also obviously have yt-dlp tied into everything related to music ala navidrome, rsync etc as shared here >>105607430

>>105607429
As long as they're actually refurbished and not just used the refurbished drives should be fine.
ZFS is copy-on-write, so not great for SSDs. I've never personally tried it, and I don't think I would unless I absolutely needed the random I/O.

>>105607662
they're 40 bucks for 2TB so i am not expecting them to actually be refurbished to be honest. but at least they have the 1y amazon warranty.

>>105607586
>>105607602
Nice, thanks. My uni had a Ubuntu server so I was sure there'd be no reason it wouldn't work the same, but I just wanted to check.
I'm just dipping my toes in the water as this PC can only hold one drive, well technically two but I don't feel like removing the disc drive, and just want to try some things out. Xubuntu was my first Linux distro so I'm pretty familiar with it, but I wanted to try Debian for once. But apparently I can go fuck myself for wanting to try on this machine. I did already install Xubuntu on it so I know it works but sometimes I try to get familiar with other distros even if they all work about the same.

Should VMs be stored on a separate drive than the proxmox boot drive?
If yes, are consumer SSDs sufficient or should I buy used enterprise SSDs? Should I mirror both the boot and VM drive? Should I use ZFS for that?

>>105607429
im building a nas with 2 10 year old consumer 120gb zfs mirror ssds for the debian system partition
if they die within the week i will report back

I ran out of SATA ports. Do I really need an LSI HBA to expand or is a chinesium PCIe card enough? Any special chips to look out for for either?

>>105608083
A second hand HBA costs next to nothing.

Is there any good Windows compatible filesystem that I can back up my TrueNAS server storage to? One that lets me browse the files if the NAS fails, and that also has some kind of checksumming or data protection/corruption detection built in?

Doing file level copies to NTFS externals through Windows with freefilesync seemed fine at first but I started thinking about what ifs, and what if the NAS fails? What if I fix the NAS then copy corrupt files (from the NTFS externals) back to it without knowing?
Hashing every file seems inconvenient especially when files get updated.

I thought about connecting the drives to the server to back up directly but honestly I feel more comfortable with encrypted externals that I can plug into Windows PCs and mount directly with no complicated steps involved between me and my data.

I have a few backups so I'm not really looking for error repair but just notification if something is corrupt or can't be validated to be intact. If something gets flagged, I'm fine plugging in another backup to see if that copy is intact. I've generated some PAR files for some old truly archival grade folders/files that will never be modified but that's not really doable for everything.
Hashing files individually also adds lots of little files so I would want something with a database to have it consolidated somehow if that's a thing.

>>105608227
It isn't difficult to get smbd + zfs running on something like *ubuntu. Maybe that's the move?

>>105608188
the expensive part may be the power consumption, depending where you live

>>105604769
This is important >>105605083
Assuming HDD raid (who raids ssds?), just getting them all 7200RPM should almost always yield similar speeds. Make sure not to mix SMR and CMR drives (ideally don't buy SMR at all).

>>105608431
If you're worried about power consumption why are you running more hard disks than SATA ports? I don't think power consumption costs are a factor here.

recommended UPS to power my 550 W platinum PSU for up to 10 min?

>>105608575
only a sith deals in absolutes

>>105607430
Use soulseek too, youtube has slop tier quality.

>>105608670
>asks a question
>I don't think
>absolutes

Moron.

Does anyone else build dashboards to monitor their Internet?
I went a bit overboard with it.

>>105608083
nvm i overlooked 2 additional SATA ports on the mobo

>>105608914
I used ntopng in the past but realized I never looked at it and only set it up out of boredom.

>>105601125
>iSCSI
look also into nvme-of which is a better alternative
>10G ethernet directly to my PC for iSCSI
you can go 25, 40 or even 100 using cheap used mellanox connectx cards
>I've got an Asus Prime X570-Pro mainboard on hand, but I assume that it does not have enough PCIe lanes for everything and I don't know if I can even bifurcate every PCIe port
it supports x4/x4/x4/x4 and it seems it can also do x8+x4x/4 in the first two slots

I'm in the same boat as you but thinking of going deeper (something like 4 nvme raid, optane and 40gbps or 100gbps network) so I've already looked into the cheapest way to get more pcie lanes
you can either buy an old lenovo thinkstation p520 (5 pcie 3.0 slots, 48 lanes) or build a x399 gen1/2 threadripper system (3-4 pcie 3.0 slots, 48 lanes)
there's also cheap 1st gen epyc combos (search H11SSL-i in ebay), that will have like 6 pci slots and over 64 lanes

>>105605664
>I will need to reserve it based on the mac address of the device
Uh, yes... What do you think IP addresses get mapped to?
>I know but I just cant be bothered with that as it adds extra administration.
You don't want to locate your device under the DHCP leases tab and click "reserve" because that's too much work? Lmao, okay buddy.
>And Ideally dhcp mapping isn't something I wanna used over setting it in the device static IP instead
You literally said you don't want to do static IPs.... mongoloid

>>105601931
replied here for visibility >>105609105

>>105606881
This is only a problem if the router (DHCP server) goes down and stays down, at which point I'd assume getting it back up would be your main priority. Device 1 would also have to go down and have its lease expire, which can be set to 100+ years. But yes, if accessing that local device is more crucial than having a functioning router / DHCP server, then sure, you'd put a static IP on that device.

What are the practical differences between a DHCP reservation and a static assignment, anyway? When would it be best practice to use one or the other?

How bad is the interference on flat CAT cables? I need to lay a cable through my appartment for about 25m. I need to go under 2 doors, so I need flat cables at least here.
Is the interference negligible or should I run regular round CAT cables to and from the doors and only go underneath them using flat cables and then just get a bunch of couplers?

>>105609337
A fundamental rule is having 50% of your ips as static to balance the network

>>105609338
Cables being flat or round says nothing about interference. What determines sensitivity to interference is:
>frequency/line speed (the lower the line speed, the less susceptible it is)
>shielding (the lower the line speed, the less it matters)

>>105609376
NTA but I'm pretty sure a flat cable cannot be a twisted pair, and my understanding is that the twisted pair structure is key to cancelling out interference

>>105609338
I don't know the answer, but mind you ethernet cables are rated for 100m, so you're only doing a quarter of that, and even cat5e cables are rated for 2.5G at that distance. So if you're doing like a gig or less over 25m, and you buy a high quality cable, it might well work just fine. (I have no idea how flat cables are even rated, since they obviously can't adhere to the cat- standards.)

>>105609400
>a flat cable cannot be a twisted pair
They can be and they are. Of course, I can't guarantee that every chinkshit flat cable will be.

>>105609376
>Cables being flat or round says nothing about interference.
oh sorry, i thought flat cables in general were more susceptive to interference because they generally have less isolation around the cables since they need to stay flat.

>>105609400
>I have no idea how flat cables are even rated, since they obviously can't adhere to the cat- standards.
i'm seeing a lot of flat cables labeled Cat6 or Cat6a. is that just wrong marketing then? does the actual Cat standard mandate round isolation? sorry, i'm still a beginner.

>>105609436
>i thought flat cables in general were more susceptive to interference because they generally have less isolation
The thickness of the PVC coating doesn't meaningfully affect interference.
>Cat standard mandate round isolation?
It does not.
>is that just wrong marketing then?
It's a rating. But you can't always trust chinks with their ratings since there's no stringent enforcement, just like with widely available chink fuses that easily go over twice their "rated" current.

Are you looking for 1GbE or faster?

>>105609433
Literally how? If they are then wonderful but I just don't see how that's physically possible.

>>105609436
>is that just wrong marketing
It can be, if you're looking on amazon or whatever then buying ethernet cables is a fucking minefield. Mind you for home use they'll be good enough 90% of the time, but like the other anon said, there's no strong central standards enforcement. For example look up "cat6e" cables, there's always a bunch for sale, but cat6 never got an "e" extension (the improved version is called cat6a).

>isolation
Ethernet cables don't actually rely on isolation. It's the twisted pair shape (literally pairs of two wires twisting around each other, like a DNA strand) which allows interference to be cancelled out. The plastic around it is mostly just for simple contact isolation (so you don't physically short them together) and structural support, not shielding.

>>105609400
>>105609523
stop being so old and british, you're making my brain hurt.

>>105609536
What the fugg how

>>105609495
>Are you looking for 1GbE or faster?
1Gbit should be enough.
Anything I need to look for specifically? Getting one from a cable specialised webshop isn't really that much more expensive than the chinesium on amazon.

>>105609400
>>105609523
>>105609548
not that guy
but
>I'm pretty sure a flat cable cannot be a twisted pair
Anon, stop thinking in 2D. We're in 3D.
They're flat, like plywood is flat.
Not perfectly flat or close to it like a flex cable is, but flat enough.
>If they are then wonderful but I just don't see how that's physically possible.
Twisted PAIR, with multiple pairs in the cable. With the pairs in a row, you get a flat cable.
See pic rel, it's really that simple.

Now you know.

>>105609337
static is more robust (not reliant on the router, only the switch working) but more harder to manage (more manual going into every device to change)

Is the TBW a reliable metric for SSDs? Is it worth investing extra into an SSD with higher TBW or is lower price per TB always better?

>>105609917
*a reliable metric for expected SSD endurance

>>105607488
>>105607586
wtf do you get out of running a de on your server

>>105609917
>>105609935
I believe it should be reliable however remember that it's an average. Whether it's worth it depends entirely on the price difference - your data storage should already be resilient to drive failure so it should purely be a calculus of how soon you expect to have to pay for a new drive vs. how much you want to pay extra for a longer-lasting drive

IMO the middle ground drives are fine, the real shitter chink SSDs are probably not worth it but the super high end "flagship" drives aren't worth it either as far as I'm concerned.

>>105608083
a chinesium pcie card is enough. dont buy one with a shitty chipset i.e. avoid jmicron
an lsi hba is much more reliable

>>105608908
ok retard you are a dumb faggot with zero sense of reflection
last you

>>105610026
>Can't use basic grammar
>Calls others retard

Dumb faggot.

>>105609623
Anon
The pairs are twisted around eachother in regular, compliant, cable.
Pairs are twisted
twists are twisted
cat5e and above also have a central core (you can argue that the pairs being in their own "pockets" fills this role)
cat6a and above require S/UTP, usually S/FTP.

What do people get out of big software packages and custom distros? I set up a multi functional server just using debian, docker, and some other simple (to me) tools. Just off the top of my head I came across lots of talk on things like unraid, proxmox, trueNAS, portainer being used in home servers when browsing guides/forums. idk it all seems like bloat to me, but maybe I'm missing out on something.

>>105610369
It makes things simpler to manage.
>log into webui for proxmox
>can see all my VMs, their status etc
>can deploy and reconfigure VM/LXCs/Proxmox hosts
Yes, it's an extra set of tools but it's akin to calling a chainsaw "bloat" when you can just use a hand chain and yank it round the tree yourself.

I'm guessing this setting is bugged if you set no ip address here?

I tried netgear.localdomain and netgear.local and it doesn't resolve unless I set a ip address field.
I thought it was supposed to just pull an ip adderss from the pool then resolve the hostname to that ip address but nope.

>>105610450
>I want to set up a static DHCP IP address
>why do I need to enter the IP address that I want to statically assign
anon I,,,,,

>>105610386
>Yes, it's an extra set of tools but it's akin to calling a chainsaw "bloat" when you can just use a hand chain and yank it round the tree yourself.

Say I already have a crosscut handsaw and I have one small tree to cut down. Why go out and buy a chainsaw, then put gas in it, then learn how to use it, when I can just go to my shed and grab the saw?

That's kind of how I see it. Chainsaws are great if you have a lot of trees you need to chop up, but people have gotten by with hand tools for centuries and they're readily available to me.


>>log into webui for proxmox
>>can see all my VMs, their status etc
>>can deploy and reconfigure VM/LXCs/Proxmox hosts

>something is wrong
>SSH in and check container status and logs
>edit docker-compose.yaml and restart containers
Again I don't really see what a webui and virtualization gives me other than another point of failure and another config headache.

>>105610469
DHCP, DNS, NAT.

All of that is the devils work.

>>105610469
I don't need it to have a static ip, just from the pool is fine. It literally even says if its not set it'll use the pool yet it doesn't work lol, clearly a bug.

Also what I want to do hostname > mac address > pool dchp works on adguard but for some reason opnsense is unable to do such thing.

>>105610514
Okay, so don't use it.
You aren't forced to use the tools that are available.
I use it because it makes it easier, I have 2x proxmox hosts, 9? LXCs and 4 VMs, I can see and manage all of them in one place.

>>105610518
If you actually look at how they're supposed to be used, they are fine.

>>105610530
see image
You just set unbound to register IP addresses from the DHCP server. If you don't want it static, don't set a static assignment.

>>105610514
The point of virtualisation is that in the good old days everything was bare metal and all servers had a specific role. So you had a physical server as a file server, another one doing DHCP, another one doing printing etc. With virtualisation you didn't need to buy hardware for each of those roles.

It gives you greater versatility as you can create a virtual machine for any new role you want your environment to do.

You don't have everything running on the same bare metal machine do you?

>>105610545
I have it enabled already
also it does say
>If this option is set, then machines that specify their hostname
so the machine does NOT specify their hostname, but I've done it through opnsnse dhcp static mappings so this does not work

>>105610530
>It literally even says if its not set it'll use the pool

Do you have a pool set up?

>>105610514
>Again I don't really see what a webui and virtualization gives me other than another point of failure and another config headache.
People aren't expecting things to break all the time to see everything as a point of failure.
IDK where this mindset even comes from.

Also the nice thing about virtualization is if you want to do something that isn't available as a feature in the middleware, just spool a VM.

I've been thinking about building an "everything" server that would fit into one compact-ish package (something like a Jonsbo N3 or N4)
By everything I mean I want it to have two "parts".
One is something like Truenas Scale, running Jellyfin, the various *narrs, nextcloud, MAYBE the occasional game server, mail server, typical homelab shit.
The other is essentially a machine I can SSH to to run machine learning using whatever Nvidia GPU I could get my hands on.

What's the most "humane" way to use two virtual machines like that with one GPU?
(not necessarily, literally two VMs, any other solution is welcome)
By that I mean that while I like having a server, I don't like maintaining it and tinkering with it beyond the initial setup, so I'm looking for a solution with minimal headache.

My initial thought was to use something like an Arc pro A40 for Jellyfin encoding and the other nvidia gpu for ML tasks, but considering size, power and money limitations it would make more sense to use just the one GPU for everything, especially since encoding videos isn't very demanding for even 3-4 concurrent viewers.
So that's option 2, but from what I've read about TrueNAS Scale if I want to pass a GPU to a VM it would require another GPU to be used by the bare-metal system.
So I guess Proxmox with two virtual machines?

I'm certain I'm missing something simple, which is why I'm asking.

>>105610566
works fine for me, are you deleting the old lease for the device?
set static mapping as "test"
delete old lease
when my phone acquires a lease it gets the hostname test

>>105610621
static mapping

>>105610553
He has everything running in docker.
>>something is wrong
>SSH in and check container status and logs
>edit docker-compose.yaml and restart containers
But why not just do all that in a vm under Proxmox? Muh bloat isn't a reason. Unless you're running your server on a 486, it's not going to impact your performance.
>Pseudo ipmi functionality
>Multiple nodes allows for live migration of virtual machines

It's a lot cheaper to run your simple debian server as a vm in proxmox than it is to buy some hardware with ipmi/kvm functionality.

>>105610621
>>105610628
Also works fine on my machine.

>>105610668
Why not just use LXC in Proxmox and have VMS and containers side by side all managed in one place? What could possibly go wrong with this instead of running it all bare metal?

>>105610553
>You don't have everything running on the same bare metal machine do you?

Why would that even be a problem? I'm just doing file storage, video streaming, and an NVR on my local network.

>>105610685
Nothing. Anon just likes hist autism box. I'm just giving him reasons that it would be better with proxmox, even without changing how he does anything.

>>105610685
I want to switch my brain off and use the pre-made docker compose scripts - I run both LXCs and a few docker hosts - but it's easier setting some things up in docker because that's how they expect you to use it

>>105610628
ok that works suddenly weird

>>105610713
you forgot to revoke the old lease didn't you~
:3c

>>105610705
Do you put all your eggs in one basket ?

>>105610705
>need to restart machine
>router is down
>torrent client is down
>jellyfin is down
I mean, yes that would technically happen if you had a single machine with no HA, but my eventual goal is to at least have my OPNsense VM be HA so I can restart proxmox hosts without knocking everyone else off the internet.

>>105610668
Maybe proxmox is more simple than I first imagined, I'll admit that. I just picked docker for managing the few services I have because I'm familiar with it. Honestly it's less about containerization to me and more about having networking and storage mount configs in one clean tidy place.

>>105610747
>not just copy+paste fstab everywhere in your enviro

>>105610618
Don't run your VMs on TrueNAS, run TrueNAS on proxmox and pass the entire HBA through.
TrueNAS wasn't built as a hypervisor and those fucking retards keep breaking stuff every release

>>105610758
>TrueNAS wasn't built as a hypervisor and those fucking retards keep breaking stuff every release
What have they broken?
Genuinely curious as I've kept everything in truenas and use VMs

>>105610729
>>105610742
I'm the only client of this non essential software. These services are a luxury. If it goes down for a few minutes, who cares?

Also, my router is independent of my home server.

>>105610793
So you do put all your eggs in one basket.

>>105610793
Me. I see no reason to nuke all of my services because I need to reboot something.

>>105610784
They broke VMs on the latest release I think? I just saw a load of "warning warning this is unstable it will break bla bla bla" and just stopped reading. It works fine for me on Proxmox (though, I need to buy more ram since my host only has 16GB)

>>105610793
Then good for you! everyone has a different scale they need/want! you don't have to use things if you don't want to!
I ran everything on a single Optiplex 7040 Micro running windows server for like 2 years, now I have a 50TB NAS, 20TB of NVMe and run a bunch of bullshit.
do I /need/ all of it? not really. Is it cool and fun? yes uwu

>>105610742
>>need to restart machine
this doesnt happen

>>105610840
>he upgrades his ram and CPU with the machine running
Damn anon, share that mainframe with the rest of us!

It's crazy how much temps vary between HDD's. My highest HDD is 50C while my lowest is 31C.

>>105609597
If you're sticking to gigabit, then interference should not be a problem, meaning you should be fine with using regular unshielded Cat5e, which is usually more flexible and thinner than any of the higher Cats. This means you might be able to get away with using round cables even under the doors. In my experience the flat cables deteriorate very quickly and wires start poking through sooner rather than later. I've never seen this happen with regular round Cat5e, even when they get regularly squished by opening doors or machinery. Also I don't know if you've considered it, but you might need unterminated cable for those tight runs and terminate it yourself (in which case you probably don't want to deal with flat cables). 300m rolls seem overkill, but you might be able to find shorter lengths for sale.

>>105610865
My M.2 drive is around 48-50c, but all my HDDs are within 1-2 degrees of each other, between 28-30c.

>>105610742
For me it's:
>need to restart machine
>secondary DNS resolver is down
>dumb clients that don't also try the primary DNS are also down too now!
Oh, well. That's annoying at least it lasts like 5 minutes at most.

>>105610758
Are there any gotchas I should keep in mind as a newbie if I do this?
That is, having multiple VMs using the exact same hardware, like the same GPU.

>>105610816
>They broke VMs on the latest release I think?
They migrated to Incus, not necessarily broke VMs entirely.

>>105610876
i got home and measured the gap under the doors and i realised that it is so big that i can even fit a 6mm Cat6 under there. thx for the help. i've been making this way too complicated

>>105610878
I think that's near impossible with a Supermicro case. They're packed too close together and there is no way to properly cool them since the backplane is in the way.

>>105610903
I don't think you can pass the same GPU to 2 VMs simultaneously, without some funky enterprise SR-IOV shit, which you're not gonna have on your peasant hardware.

>>105611000
Most shelves are even denser then that.
Fill all the sleds and it'll probably even out.

>>105610848
damn you're right it does in fact happen once every 3-4 years
how will my users cope with an hour of downtime communicated well in advance

>>105611113
okay anon, I think you're really cool for running five 9's in your homelab, some people aren't.
it's YOUR homelab, and my homelab is MY homelab, these two are different things.

>>105611269
i dont have a home lab
i have a homeserver
it just so happens that i never need to reboot it since i set up live kernel patching and it already has ample hardware

>>105611351
That's cool anon!
some people like doing different things. It's cool that you have a homeserver and you're happy with it!

>>105611359
ok but why would you be rebooting your homeserver all the time
theres no reason to reboot ever except for kernel upgrades, if you're too lazy to setup livepatching (wouldnt blame you)

>>105611408
Changing random hardware
Needing to move them
windows update (yes, I run some windows servers, I'm a windows sysadmin fite me)
there are some servers that I don't need to restart, some of them I restart monthly for whatever reason

>>105611437
>I'm a windows sysadmin
I'm sorry to hear that anon

>>105611561
>forget to fully deploy RDS
>RDS grace period expires
>cannot RD into machine to fully deploy RDS
we also didn't domain join the machine
Windows Admin Center saved my ass, managed to DJ it using powershell and then got in with my DA creds.
It's fun + I can run my i9 7980xe HFT server as a game server in an office rack :3

ITS OVER 100!!!

>>105610742
Not everything needs virtualization and your services don't need to be all in one place, this doesn't make your life easier in any tangible way.
A router or firewall as a VM is the biggest meme.
Torrent clients are disposable, you could run a low cost, low power mini PC as one. In fact, for pretty much all things homeserver, modern mini PCs are more sensible choices than rack servers or "Optiplex" style e-waste.

>>105612077
>Don't use virtualization lets go back 20 years and do everything on bare metal
Router, vpn, firewall should be done on bare metal, but that's it.

>>105612077
>dedicate 4 mini PCs for your infra
>instead of just one or two that can run everything
Like, okay, if you want to do it the hard way.
I fucked up my OPNsense install one time, do you know what I did instead of reinstalling and loading from a backup? I clicked "revert to checkpoint" while I was laid in bed
but hey! like I've said a dozen times, you can do whatever you want! I'm not forcing you, or implying you should do it the way I am! have fun!
>>105612112
I use bare metal firewalls at work, but having a virtualised firewall still has some neat benefits. I can roll back a completely dead, unresponsive firewall from any system in my network
Also, all my other virtualised services never touch a physical switch, so it's a lot tidier IMO.
It also means I don't need to devote an entire mini PC to something that only really needs 2 cores and 2GB of ram. I have an entire NIC passed through, so my "WAN" side is for all intents and purposes bare metal, LAN is on a shared SR-IOV X710 that my other VMs share, my host has its own X710 port, but also shares the OPNsense X710 so again, it doesn't need to traverse a pair of switch ports to get to the internet.

>>105612188
How is the X710? I've been meaning to buy a proper Intel NIC for sometime but keep putting it off because I don't really need it yet (if I ever got dual WAN I would)

>>105612350
Its been rock solid so far, the MS-01 has an X710 built in, so that's why I'm using it. I have a Marvell 10Gbase-T NIC in my desktop and it is dogshit under windows, but the MS-01 is running windows server and working fine. SR-IOV support is nice (though, any modern NIC generally has SR-IOV anyway)

In the MS-01 specifically, I have one port of the X710 in a shared vSwitch that goes to OPNsense as a LAN interface, is used by the OS and some other VMs, then the other port is dedicated to the host only, so that in theory, if I'm doing big data moves at 10Gb it doesn't choke out my LAN traffic.
I also have the non-Intel ME I226 in a private switch for my WAN connection (1Gb symetric fibre) that goes to OPNsense, the other I226 is again just dedicated to the host. Since (currently) my MS-01 is my main server, there isn't that much traffic in and out of the X710, but I just built a NAS on a dell R240 (louder than my R220 :( but not unbearable if I keep my room cool) so I expect soon there'll be a lot more traffic passing through it as I move services around and they need to touch the network more.

Eventually, I plan to get another MS-01 and build an overlay network over the thunderbolt ports for VM migrations as well as general node-to-node traffic (which I an fairly sure is possible)

>>105612188
The problem that I've seen is people installing firewall VMs onto all sorts of hardware that has old inadequate CPUs, and weird hypervisor networking configuration that creates novel failure scenarios. And then those hypervisors never get updated or lag behind horrifically because the admin is scared of what could go wrong with routing if the host runs into issues.

What's more, sometimes the hypervisor is in a remote location and accessing its OOBM is done through the virtual router. This is how some people end up running ESXi 6 in 2025, or a 5 year old pfSense that they're afraid to upgrade.

>>105613281
I reserve my right to not comment on those specifics

Are those cheap intel optane nvmes actually more reliable or were they just "more reliable" at the time and newer stuff is just as good?

>>105612032
>105% cpu usage
peak optimism

>>105613281
>ESXi 6 in 2025
heh
i'm still running that, because later versions don't support the Realtek NIC I'm using :p

>>105613352
The Optane memory is exceedingly durable and with the M.2 sticks being low capacity and cheap, it just screams "cheap durable boot media"

But its literally impossible to say if they are going to be more durable then any random NVMe drive atleast for that usecase.

For the larger ones like the 905P, yeah they are probably going to be more durable than a NAND based SSD of the same size especially if used for some type of caching.

>>105613810
Loonix supports everything and, by extension, Proxmox supports everything :^)

>>105613969
i have thought about it, but i've decided to rebuild everything on rhel instead

>>105613992
Based and redpilled.

>>105613992
proxmox sissies btfo

>>105613992
>RHEL
>Not Rock or Alma

Unbased and bluepilled.

>>105609105
>look also into nvme-of which is a better alternative
I didn't even know of this, I'll look into it.
>you can go 25, 40 or even 100 using cheap used mellanox connectx cards
My 'puter has native 10G Ethernet which is why I was going after that. Also I live in a pre-internet flat in Germany, I cannot drill holes into the walls to pass SFP cables but I can use flat ethernet cables around my doors. Hell, they'll drill a hole from the staircase into my flat to bring fiber instead of doing it properly through the wall like they did with DSL.
Otherwise I'd contemplate running 100G easily.
>it supports x4/x4/x4/x4 and it seems it can also do x8+x4x/4 in the first two slots
Neat. Where did you find that chart?
>Epyc
That's a great hint, thank you very much!

are n100 chips still the best choice for mini pc servers? and which ones are good?

>>105614135
my desktop is currently on Rocky, but that's only because i didn't know i can use dnf --installroot to bootstrap a RHEL installation on a ZFS volume.

>>105610002
Oh I didn't go looking to do it. But it does help that I'm retarded, so having the ability to manually plug a screen in back in the old days, or now just turn on kvm, does help when I can't be fucked to fix things with a terminal.
That's about it, really.

post start pages

>>105614766
>Afternoon!
lol

>>105614766
What's that?

>>105614462
I hate that they don't offer ECC RAM. Single channel, alright, it's not supposed to be a gaming chip, but ECC is such a missed opportunity. And more PCIe lanes would be nice too.
They could carve out a real niche with that.

>>105614724
...
but you still have to fix things with a terminal

>>105614766
i'm into that minimal aesthetic

>>105614906
Which -does- work about 95% of the time. Alas, I am not clever enough to make the last 5% of problems go away with only the terminal.
I own my faults.

using OPNsense what is the actual difference between these 2 DNS server settings? in general and within the services IPV4 DHCP

I'm trying to setup my Adguard docker container IP address as the DNS server for all my devices on the router level

>>105615212
if you don't put in a dns server it will obviously default to your ip's dns.

>I'm trying to setup my Adguard docker container IP address as the DNS server for all my devices on the router level
is the docker hosted on the same machine as opnsense? typically to do that you leave the dns blank in opnsense and set the upstream in adguard to 127.0.0.1:adguard port like so. there are many tutorials on doing it

opnsense already has a plugin for adguard home so im not sure why you're running it in docker

>>105615212
right (System) are the DNS servers that OPN itself uses. additionally, other services may use these DNS servers for their own configuration, e.g. the Unbound DNS service might forward queries to these servers if you dont manually configure any forward servers yourself.
left are the DNS servers that the DHCP service tells clients to use once theyre connected. if you leave it blank, then DHCP will tell clients to use OPN's internal DNS (Unbound/Open/etc) as their DNS. if you leave it blank AND you dont have a DNS service enabled, then DHCP will tell clients to use the "global" DNS servers (the ones on the right in your pic).
>I'm trying to setup my Adguard docker container IP address as the DNS server for all my devices on the router level
according to the other anon theres a plugin so i would look into that first, but otherwise:
1) set up container
2) put the IP in the DHCP DNS settings
and youre g2g. you could also enable one of OPN's DNS service and forward all queries. not sure what the benefit would be tho.

>>105615285
you get a nice GUI with docker or am I wrong?

>>105615348
I was following a guide to set unbound as the upstream as well in adguard, but yeah looks like it all works until something breaks down the line

>>105615393
>set unbound as the upstream as well in adguard
i would do the opposite, i.e. unbound forwards to adguard. that way DHCP leases that are registered with unbound can be resolved without having to forward outside your network, as opposed to adguard trying to resolve vm06-loliarchive with its first upstream before it gives up and forwards it to unbound.
you could also try and figure out how to register DHCP leases from OPN to adguard, but i was never able to figure that out with pihole.

>>105615393
>>105615461
oh i see, Unbound is the ONLY upstream for adguard. i guess that does the same thing.

>>105615461
nah you're right I might just do that, I tested it out in OPNsense in DHCP setting 192.168.1.10 (adguard container ip) as the primary dns server then 1.1.1.1 as the secondary.

Then put my adguard offline and it failed to use 1.1.1.1 and I couldnt load websites. Might be I need to renew the lease after I changed the settings though

>>105615096
im struggling to imagine anything thats easier solved with a gui rather than through the terminal
its not meant as a slight

I have some ram, 2 NVME ssds, 2 full size HDD, and a mini HDD from a broken laptop. how do I check the case I'm getting can fit all of this?

>>105615532
None, taken, and I completely agree, however, as previously stated, I am a retard:
>server runs off of a Gigabyte MC13-LE3
>the IPMI/kvm connection to the host OS consistently drops input connectivity to host, while maintaining video feedback to KVM window after 2 hours of uptime - I can see the OS enviornment, but my mouse/keyboard imputs aren't reaching the server
>nothing fixes this aside from rebooting the host
>rebooting host has twice now caused the CPU fan to just stop spinning when the OS is loaded, which in both cases required me to unplug and replug case fans from the mobo while booted into the OS to fix, even without any funny Y-splitter on the cpu head
>so to avoid restarting the host, I just plug spare keyboard/mouse into server and run kvm input to tv mounted above the server
Why would I even need this functionality, you might be screaming at your screen?
>deluge daemon won't allow me to set download source location to anyting aside from the server's boot drive remotely (and by extension, the zpool where my served media is stored) so unless I just make a copy of the torrented media to the zpool, I have to interact with the host directly to actually move the download location of any given torrent where I want it to be, the server's streaming zpool
tl;dr I'm a retard who can't quite handle his shiny new server mobo yet, which represents the singular need I have to interact with the server's GUI.

Should i use T568A or T568B wiring when using a crimper?

>>105615609
And I guess by extension I should add the deluge dameon, but I've also only had that working for about a month now.
The real kicker is the post codes displayed in the ipmi are referenced on page 83 of the manual - there are only 82 pages in the official pdf on gigabyte's support page for my motherboard: https://download.gigabyte.com/FileList/Manual/sever_manual_MC13-LE3_e_v1.0.pdf?v=ead3ff48b5ebba4a2decb8891ec09b22

>>105615609
oh you're the anon who didnt do the incantation and wonders why his gigabyte board doesnt work

>>105615680
what

>>105615554
look for any tower matx case

>>105615461
ok so I've set queries to forward all to adguard docker container. But if it goes down where do I set the backup DNS server for example to 1.1.1.1 instead of forwarding to adguiard?

>>105615754
Just use 8.8.8.8 or 8.8.4.4 like a normal person.

>>105615690
you didn't perform the ritual properly and now your fans dont spin
sad, many such cases

>>105615609
>>deluge daemon won't allow me to set download source location to anyting aside from the server's boot drive remotely
ok i'll stop being silly
are you running deluged under its own user account with fucked permissions on your other mounts, by any chance?
or do you mean theres no option to change the download location?

>>105615832
Oh the option is certainly there, doesn't even throw an error - just doesn't complete. And let me be clear, this is for attempting to change the download location through the thinclient running on my workstation, which is linked to the daemon on the server, running under the primary user/group I built everything with:
sudo nano /etc/systemd/system/deluged.service.d

# Override service user
[Service]
User=anon
Group=anon

where the connection to the actual daemon is configured as:
anon@1.2.3.4:58846

And the actual daemon is configured as:
[Unit]
Description=Deluge Bittorrent Client Daemon
Documentation=man:deluged
After=network-online.target

[Service]
Type=simple
UMask=007

ExecStart=/usr/bin/deluged -d

Restart=on-failure

# Time to wait before forcefully stopped.
TimeoutStopSec=300

[Install]
WantedBy=multi-user.target

The only real customization was setting the UMask so Plex could see/scan the files, but that's the long and short of it. Both my server and workstation have the same user name and group, just swap that for anon and 1.2.3.4 for the server's actual IP.
So at this point, when interacting with the thin client, I can move the download folder to any directory referenced in the boot drive's filesystem (/...), but the issue is the file explorer that's referenced when doing so reflects the filesystem on my client! So any attempt to set the download folder to the pools necessitates me attempting to reach the target pools through their mounted location (/mnt/nfs/plex/...), which is what's failing to apply - not even throwing an error message.
What, precisely am I fucking up here?

>>105615990
>What, precisely am I fucking up here?
using deluge

>>105616011
Checked. Surely it can't be that cut and dry, right? What are my alternatives?

>>105615789
What can be done to save my soul?
>>105616011
Checked. Surely it can't be that cut and dry, right? What are my alternatives?

>>105615762
yes everyone listen to this fed

how hard would it be to make a primergy hold a matx server board, would an angle grinder be sufficient

im putting together a hsg faq

>why
shut the fuck up you stupid cunt

is no one concerned about their electric bill? on average im using 2.2kW a day. that's $236/mo @ 14.7 cents.

>>105608726
>it's just an appimage
Why the fuck isn't setting it to executable with proper permissions running it? What am I missing here

>>105615832
You know what's even funnier? I figured out that I -can- remotely set the torrents to the proper destination in the streaming zpool...but instead of manually moving the folder in the thingclient, I do it by copying the file path of the destination to the same operation window in the web client. That will cause the file to be transferred to the proper location.

>>105615832
You know what's even funnier? I figured out that I -can- remotely set the torrents to the proper destination in the streaming zpool...but instead of manually moving the folder in the thin-client, I do it by copying the file path of the destination from the thin-client's gui interface to the same operation window in the web client. That will cause the file to be transferred to the proper location without any problem.
Thus I have no further need to reboot, use the KVM or have a DE. The fans stay on and the server can run until the hurricanes get here for all I care now.

>>105617299
we've got options man

>>105617299
How the fuck are you doing that?

>>105617305
Use nicotine+ on your desktop, it's for finding music.

>>105617299
500w for a nas is nuts, thing must be glowing red

>>105617952
And just like that, we're in. What's proper upload/download etiquette here? Does that even matter? I have stuff I could share but very little of it is flac.

>>105617904
AI gen'ing so 3090 is using 100% power + 500TB server + Air conditioning

>>105618027
nah I just have about 40x hdds in it

>>105618028
Just keep your music in your public shared folder and you're fine. Don't put stuff in private and port forward if you can. Keep in mind that everyone can see your shared folder so don't put anything personal there.

I need an HBA. Is the IBM M1015 with the LSI SAS9200-8i a good choice?

>>105618127
how much are you earning with all that slop

>>105605042
Even core 2s from 15 years ago run well on sata drives, it's more of the enclosure's limit as most old pleb office towers have like 2 DVD bays and 2 or 3 SATA bays.

>>105605042
they should be fine.
the only problem with older CPUs is that they usually draw more power. although old office PCs are at least more efficient than old gaming hardware.
if you need more SATA ports for your drives, you can always get a used HBA on ebay for about 40 bucks. they go in a PCIe slot and let you add another 8 drives.

>>105616780
dremel with tiny cutting disk, wear full faceshield tho.
front io is a bitch, idk about primergy but lenovo front io is not simple usb header but has its own cpu and firmware.

>>105618127
>nah I just have about 40x hdds in it
What are you storing? Asking bcause my nas is just 4tb and it's not even half full. Have another 4tb nas but only for offsite backup.

Dear /hsg/, I fell for the mini PC meme because it's so cute and tiny and sits under my router, but what is the easiest way for me to add additional storage to turn it into a NAS? I'm thinking 4x10TB HDDs for a start.

>>105610618
>machine learning
>truenas

>>105617299
how is your NAS using a constant 500W?

>>105619661
hba for pcie/nvme slot, can use a riser low profile card on some mini pcs, might even get a 4e lsi to work with a disk enclosure, no matter how you get the sas/sata cable out power will have to come from an external source

>>105619661
external hdd bay? not like a mini pc could handle full sata speeds for all drives at once so youre not posing anything

What are some decent flex PSUs?
Any of the aliexpress ones any good?

>>105620101
just stick to silverstone or fsp

>>105619600
half is just mirrored so it's really only like 250tb usable. as far as storage, mostly porn and ai models. AI is taking up an increasingly larger amount of space.

>>105619105
earning? i don't make money doing ai it's for fun

Is there a way to limit specific docker container network access only to local network?
For example I don't want to configure anything for the container to not send logs or ping home, I just want it to not have external internet access at all

>$ ls /var/.zfs/snapshot/ | wc -l
>38
>$ ls -logdh --time-style=long-iso /var/.zfs/snapshot/*/lib/mpd/state
>...
>$ uptime
>... load average: 9605.67, 9599.18, 8862.55
wtf. Seriously, is listing a single file in 38 snapshots so much to ask? `ps aux` shows it forked hundreds of `mount.zfs` processes for a single snapshot and it slowed my PC to a crawl.
Wish ZFS had some CLI tool to get stat info of a single file in a snapshot without having to mount the entire snapshot.

>>105620455
On BTRFS you can just mount the root at least (and I don't mean the root as in / I mean the root as in the subvolume that contains the /@ subvolume - or whatever it's called - which is mounted at /). I don't know why ZFS has this weird "auto-mounting of snapshots" behaviour.

>>105620455
>$ cd /var/.zfs/snapshot/2025-06-16-00-00
>...
>picrel
it just forks the same mount process infinitely... i can't even `cd` into a single snapshot.
However, if i mount it manually, it works:
>$ mount -t zfs nvme/gentoo/var@2025-06-16-00-00 /mnt/zfs
Conclusion: don't use `.zfs/snapshot` directory, just do it the old "legacy" way.
>>105620480
>I don't know why ZFS has this weird "auto-mounting of snapshots" behaviour.
yeah, ZFS has tons of Solarisms and "not invented here" syndrome, the "legacy" `mount` + fstab is replaced with `zfs mount`, kernel cache is replaced with separate ARC, NFS and samba config options are specified through dataset properties, this entire `.zfs/snapshot` directory thing...
It's all so stupid, I wish ZFS on Linux ditched all those things and became just a regular filesystem like any other without all those things bolted on and reinvented. ARC being separate has bitten me in the ass multiple times too, i wish you could just disable it and use standard kernel page cache...

>>105620552
ARC is one of those things that can bite you but it probably should be separate to the kernels cache because of the way it works. It's adaptive and is a bit more clever than the kernels cache. It also can span multiple devices (e.g a cache device on an NVMe drive which is really useful and properly persisted across reboots too now. The native Linux solution to this is bcache which is really crappy and full of bugs. They're trying to replace it with pcache which will hopefully be better integrated. I'd really like to be able to use an SSD cache for a BTRFS RAID without having to do all of the crap bcache requires)

>>105620552
since you are using gentoo maybe there is an incompatibilty in the versions of kmod and tools you use (for instance). im glad zfs is the way it is.

>>105620647
good point, this might be it
>$ zfs --version
>zfs-2.3.2-r0-gentoo
>zfs-kmod-2.2.5-r0-gentoo

>>105620480
>I don't know why ZFS has this weird "auto-mounting of snapshots" behaviour.
I think it's consistent with overall ZFS "automounting everything that is imported" behaviour. Honestly, I never really understood or liked the constant need to manually mount and unmount things and set up mountpoints and automount configs in Linux. Like, I either want to be able to use the filesystem, or I don't want it anywhere near me. But yeah, just a preference and different design philosophy, I guess.

>>105614257
the chart is from here https://www.asus.com/support/faq/1037507/

>>105619600
why don't you have at least 3 months of porn stored?
you're being irresponsible.

>>105620728
zfs doesn't even require partitioning. it solves everything that's annoying about managing disks

>>105614257
I'd choose threadripper over epyc, those server bios are a bit lacking (e.g. no overclocking/undervolting, shitty fan control) and can be finicky with modern gpus

>>105621014
Btrfs does that too, but fair enough.

>ASUS B650M-A AX II
>Supports non-registered ECC DDR5
>CPU supports it
Okay, where in the fuck do I buy it?
I'd like 2x32GB

WHO makes it?
I'm thinking of just getting regular RAM and saying fuck it.

>>105621918
new

>>105619150
>>105619481
Good to know. My main concern is that the pci ports will be so old that it will be hard to find parts that will work. I don't think I use a long of power in my apartment, so I should be fine keeping it running all day. The newest machine I've seen is a 7th gen i5. I'm not sure how many drives I could fit into any of them.

stupid nes sale is bigly impacting my LSI sales