[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip / qa] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/g/ - Technology


Thread archived.
You cannot reply anymore.


[Advertise on 4chan]


File: o-HUG-facebook-4124043824.jpg (488 KB, 2000x1000)
488 KB
488 KB JPG
I love you, home server general anons edition

previous: >>101023485 #

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

>NAS Case Guide. Feel free to add to it:
https://wiki.installgentoo.com/wiki/Home_server/Case_guide

/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?
/hsg/ is about learning and expanding your horizons. 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://gitlab.com/awesome-selfhosted/awesome-selfhosted
RouterOS's: https://wiki.installgentoo.com/wiki/Home_server#Custom
https://reddit.com/r/datahoarder
https://www.labgopher.com
https://www.reddit.com/r/homelab/wiki/index
https://wiki.debian.org/FreedomBox/Features
List of ARM-based SBCs: https://docs.google.com/spreadsheets/d/1PGaVu0sPBEy5GgLM8N-CvHB2FESdlfBOdQKqLziJLhQ
Low-power x86 systems: https://docs.google.com/spreadsheets/d/1LHvT2fRp7I6Hf18LcSzsNnjp10VI-odvwZpQZKv_NCI
Cheap disks: https://shucks.top/ https://diskprices.com/

Remember:
RAID protects you from DOWNTIME
BACKUPS protect you from DATA LOSS
>>
(actual previous edition: >>101058868 )
>>
is this baker the same as /fglt/?
>>
aaah post racks
>>
File: 20240619_162214.jpg (3.44 MB, 4000x3000)
3.44 MB
3.44 MB JPG
Epyc, threadripper, or xeon for homeserver/nas that needs pcie lanes?
>>
>>101089896
>>101089521
Epyc.
>>
>>101089668
Since android tv started being a thing
>>
There's a lot of discussion regarding one big server with zfs etc, but does anyone have experience with many small servers for storage?
Think git-annex, seaweedfs, ceph, lizardfs/moosefs, garage s3, minio etc
>>
>>101090405
I'm running 4 mini pcs, 3 of those have USB hdd enclosures. I have a mix of hdds/ssds of varying sizes. Running k3s with amongst other things rook/ceph. works well for me. It's not enterprise but that's why my important data has 3 (in some cases) 4 replicas with a host failure domain.
>>
>>101090405
Unless you specifically need distribution and high availability, not worth it. zfs on a single machine with a ton of drives is so much simpler.
>>
>>101091223
yeah, i suppose, but i got a bunch of nucs and external drives

and zfs got autism wrt usb drives I've learned
>>
>>101090457
interesting...
>>
>>101091305
>and zfs got autism wrt usb drives I've learned
If you're not using the drives for anything right now, I'd plug them all into one NUC and give it a shot anyway to see how it behaves, I doubt it's that bad unless the drives or their USB to SATA chips are garbage. You can even fill the pool with data and then try replacing one of the drives with itself (google how to do that) to see how long resilvering takes.
>>
>>101091305
>>101091407
Oh and I forgot, make sure to use /dev/disk/by-id when creating the pool to avoid stupid shit if they get rearranged on boot.
>>
>>101089673
no we are not gay
>>
>>101091423
thanks
you run it on baremetal or in proxmox?
>>
>>101091426
wew thankfully
>>
>>101091456
If you mean running the pool inside a VM in proxmox, avoid it unless you have specific reasons for doing so, just run the pool in proxmox itself since it has good zfs support (ubuntu server also has good support if you dont need proxmox features)
>>
>>101091305
>and zfs got autism wrt usb drives I've learned
it's just generally a bad idea to use USB for storage
it'll work, theres nothing about ZFS that changes this, but it suffers from many issues euch as but nit limited to:
>chink controllers shitting out at high throughout
>power saving modes spinning disks down and not spinning them back up
>power saving modes shutting off connections
>drive order changing at boot
do this >>101091423 always, even SATA/SAS drives can rearrange themselves, but the ID is always unique
>USB HDD's controller shitting the bed
Plenty of people are able to run USB storage, and look, I won't lie to you - it'll probably be just fine. However, I'd never reccomend using this as your primary storage, as it's not very reliable. As a weekly backup, should be just fine, in fact I backup to an external USB HDD weekly and it's fine. But running 24/7 or even like 12/7 I just wouldn't trust it not to shit itself.

I've said it before and I'll say it again, USB went wrong when they made it for storage. We should have adopted eSATA when we had the chance
>>
hust some morning thoughts over coffee, can someone explain to me:
SATA/SAS are hot-pluggable
PCIe is not
NVMe is basically a PCIe subsystem
m.2 (PCIe), using NVMe, is not hot pluggable
m.2 (PCIe), using NGFF (SATA), is not hot pluggable
u.2 (PCIe?), using NVMe, is hot pluggable. how?
>>
>>101091580
did some reading, u.2's connector is based on SATA Express which has the right power staging for hot-pluggability
also hot-plugging was added to PCIe as an optional feature that few boards support, but it us possible I guess
>>
>>101091407
i'm using and i've used btrfs on these disks, and it's so simple to add or remove disks with btrfs
and that seems to be such a pita with zfs
>>
>>101091944
>and that seems to be such a pita with zfs
what's so hard about
>zpool add tank /dev/disk/by-ud/jeff12345
or
>zpool remove tank /dev/disk/by-ud/jeff12345
>>
Is it dumb to run my torrent client (qbittorrent, thousands of torrents) off a NFS share (machine, qbit is in VM hence NFS)?
>>
>>101091944
>and that seems to be such a pita with zfs
That's true, if you're going to use zfs you need to commit to the number of disks in your pool. You can use btrfs but just be aware that its not as reliable and the striping RAID modes are straight up flagged as not production ready by the developers.
>>
Just bought some 3.5" storage servers. I want to load them up with 12Gb SAS drives, 4-8TB/drive. 48 drives total. Any recommendations for where to buy drives in bulk? I should be aiming for $5/TB right?
>>
>>101091971
I can't tell if this is bait or not, but don't do this.
>>
>>101091971
all the resilvering and talk about not using raid1 because the resilvering might kill the drive so you should use raid2 etc

zfs seems extremely picky about controllers, but btrfs is like sure bro and not a hazzle at all
>>
>>101091996
>That's true, if you're going to use zfs you need to commit to the number of disks in your pool
That's not true, you can add more drives to a ZFS array, it just doesn't recalculate the parity for the existing data.
It [recalcing parity] is on the roadmap but who knows when that will be added
>>
>>101091996
yeah, i've settled for btrfs raid1 meta and raid1 data, as that seems to be pretty stable
>>
>>101092024
Can't you recalculate parity by simply creating a new dataset and moving all your files to it? Regardless, it's a relatively new feature and I don't trust it that much.
>>
>>101092013
>don't add drives to an array
why?
>>101092018
>all the resilvering
reslivering is part of any RAID implimentation
>talk about not using raid1 because the resilvering might kill the drive so you should use raid2 etc
do you mean RAID5/6 aka RAIDZ1/RAIDZ2? This concern is valid for every implimentation of RAID5/6, but it's not like your drive WILL die in a RAID5 rebuild, I've done plenty without issue. BTRFS will also have the same "issues" if you use the same RAID level
>zfs seems extremely picky about controllers
never heard this before, for all software RAID you need a controller that supports IT mide, aka passing the drives through to the OS

not trying to say one is better than the other, but all the issues listed so far are with RAID in general and not specific to any implimentation
>>
>>101091580
pcie can be hot pluggable, the controller and connectors need to support it

even thunderbolt is hotplug pcie
>>
>>101092085
>>don't add drives to an array
>why?
Because adding a single drive to a zpool is actually adding a new vdev with a single drive and no redundancy, and if that drive fails the pool is gone. You can add a new vdev with multiple drives though.
>>
>>101092083
>Can't you recalculate parity by simply creating a new dataset and moving all your files to it
yes but that's just making a new array then
hardware controllers have supported this for a while and BTRFS has a method of doing it. it's been talked aboyt by the ZFS devs for a while, apparently it's in an a beta state and "works" but needs more testing before being released
>>
>>101092161
>yes but that's just making a new array then
Not really, because it means you don't need double the space of the pool so you can copy all the data off the pool and back.
>>
>>101092176
hang on, can you explain that to me?
>>101092143
didn't know this before, I assume these are related

so you are what, adding the new drive to the zpool, making a new vdev across the entire zpool, moving the data to that vdev and deleting the old one? that doesn't make sense to me, am I misunderstanding how vDevs work?
>>
>>101092003
>I should be aiming for $5/TB right?
Haha what
>>
>>101092003
>should be aiming for $5/TB right?
maybe on tape lmfao
HDDs you'd be lucky for $30/TB
>>
help me not get filtered by opnsense

I'm trying to setup the following: modem -> opnsense -> old router

I have a brocade switch and unifi wap to eventually do this "the right way" but thats a lot of work and everyone in the house is already connected to this router. anyway the docs tell me to setup a "transparent filtering birdge" but it didn't work. am i a tard or am i going in the wrong direction?

https://docs.opnsense.org/manual/how-tos/transparent_bridge.html
>>
>>101092195
>so you are what, adding the new drive to the zpool, making a new vdev across the entire zpool, moving the data to that vdev and deleting the old one? that doesn't make sense to me, am I misunderstanding how vDevs work?
No, I meant literally just reading and rewriting all the files one by one back to the pool to recalculate parity, since parity is recalculated on "new" data. According to the docs, removing vdevs from a pool that uses raidz isn't even possible at all.
>>
>>101092254
ok, that makes sense to me but that doesn't get around this issue then >>101092143
can you expand the vdev over the new drive at all?
>>
>>101092253
>transparent filtering birdge
Your overcomplicating it . Do a regular OPNsense setup . Plug modem into OPNsense WAN port then connect the LAN ports of your old router and OPNsense machine.
You will also need to put your old router in AP mode .
>>
>>101092272
Those are different things.
Running "zfs add tank /dev/whatever" adds a new top-level vdev with a single drive, which is something you probably don't want to do and I'm pretty sure zfs warns you about it if you try.
For the relatively recently added "raidz expansion" feature, you want to add a drive into the existing raidz vdev, which is done with the "zpool attach" command instead.
>>
>>101092224
>>101092248
I'm obviously talking about used drives retards, even though I've seen some of the higher-capacity drives go for $15/TB new
>>
>>101092338
>For the relatively recently added "raidz expansion"
ah, see I had heard about this and I assumed it was added to the "zpool add" command
good to know. so this will extend the vdev but srill doesn't recalc the existing parity at the new rate, meaning you'd want to rewrite all the data to do that, yeah?
and the syntax is still the same, just with "attach" insyead of "add", right? so technucally I'm still not wrong that it's as easy as one command (even if it's not as good as how other software does it)...?
>>
>>101092369
>and the syntax is still the same, just with "attach" insyead of "add", right?
It's similar but you need to specify the vdev to add the disk to. As usual, read the manual first before executing any commands on your zpool if you care about your data.
>>
>>101091423
You can also make aliases for your disks and reference those for simpler pool creation names:
cd /etc/zfs/

open the vdev config file:
sudo nano vdev_id.conf   

then enter the aliases for your disks:
alias 01        /dev/disk/by-id/ata-HGST_HDN724040ALE640_PK1334PEJ09X0S
alias 02 /dev/disk/by-id/ata-HGST_HDN724040ALE640_PK1334PCJZJWLS
alias 03 /dev/disk/by-id/ata-HGST_HUS724040ALE640_PK1334PAKUA6US
alias 04 /dev/disk/by-id/ata-HGST_HUS724040ALE640_PK1334PBHB2JNS
alias 05 /dev/disk/by-id/ata-HGST_HUS724040ALE640_PK2334PBHE2WXR
alias 06 /dev/disk/by-id/ata-HGST_HUS724040ALE640_PK1334PBH9WYYX

refresh what your OS is seeing for vdevs with
sudo udevadm trigger

Then just build the pool like
sudo zpool create HDDPOOL1 raidz2 01 02 03 04 05 06
>>
>>101092520
>if you care about your data.
and if I love my data?
>>
>>101092861
Memorize it
>>
>>101092861
if you love your data you shouldnt put it in a situation where one mistyped command can obliterate it from this world.
>>
>>101092888
but on any OS, any number of single commandscan obliterate it from this word
observe
>dd if=/dev/urandom of=/zpool-name
>rm -rf /zpool-name
>zpool destroy zpool-name
>wipefs /zpool-name
and so on
so are you saying I should store my data on an immutable device such as CD-R?
>>
>>101093136
...you should backup your data, anon
>>
>>101093136
No, he's saying you should have copies of your data on other machines
I wish I wasn't poor so I could back up my 50+tb offsite, but I am so I have to cope
>>
stop locking yourself out of ssh dummies
>>
you know you can have 2 nodes right? even if you're really scraping the bottom of the barrel, you can certainly set up -something- you can dump your data on if/when you need to expand/delete your pools.
10g nics are like $30. Stop smoking weed/drinking for a week and sort your shit out
>>
I tried installing NZBget and Headphones, but the usenet and indexers I've tried don't seem to have the music I've been trying to download on them (Metallica and Queen, I know I can find them elsewhere but its a test) so I'm at a bit of a loss as to whether I should really bother with it?
Do any anons use NZB's and such? I feel like most if not all of the things I'd use this "easier/automated" service for I can find via google anyway, maybe I should download a program that just organizes files I download myself instead or something.
>>
>>101092253
disconnect old router completely
connect wan from modem to opnsense
make sure opnsense has internet connectivity.

connect lan from opnsense to old router
disable dhcp in the router, it is redundant
enable dhcp in opnsense
put old router in bridge/ap mode

this isnt an opnsense problem.
>>
How are you supposed to edit containers in Cockpit? There is no option in the UI to edit the podman template, like if I wanted to add hardware acceleration.
>>
>>101094207
Actually nevermind, Podman and Cockpit are fucking garbage and I'm not wasting any more time on this.
>>
my 2u PSU takes 60mm fans, how can I determine if it's a PWM one or not?
I'm going to buy a memetua fan to replace the stock one, but in 60mm they offer 12 PWM or 12v FLX, usually the PWM ones are fine but I've never done a PSU fan.

Probably not the right thread but w/e /pcbg/ is full of retards that can't do anything outside of PCPartPucker
>>
>>101094221
>>101094207
>he fell for the GUI meme
normal people just edit the file in a text editor
>>
>>101094207
>>101094221
>he doesn't use portainer
you did this to yourself
>>
>>101094230
What does this have to do with servers?
>>
Opinions on XCP-NG vs Proxmox? I've been running KVM on Debian for years and I've been happy for my own personal uses however my job is wanting to move off of VMware for licensing costs but I'd never suggest they switch to standalone KVM so I'm wanting to lab using a virtualization platform to give an informed opinion on business ready alternatives. I've used proxmox a little but it never tickled my fancy so I always moved back to my KVM workflow. Any informed opinions welcome, especially if you specifically have any experience with either of their business support.
>>
>>101094230
Actually just put the new fan in, see if it works and return it if it doesn't. You could waste your time trying to find out if it wont work before buying but it isnt worth the amount of time it would take.
>>
>>101094447
it's for a server power supply - nobody else on /g/ knows or cares about modifying power supply fans, especially industrial ones
>>
>>101094468
Proxmox' support is only on European hours without partners, not 24/7 hours, and actually more expensive to license for homelabs than VMware's.
>>
>>101094506
The fans are meant to have high static pressure. Replacing them with high airflow fans like Noctua would be counter-productive.
>>
>>101094516
>Proxmox' support is only on European hours
Big fat pass, wtf why?

I'd be using community edition for either one I set up for my own lab, I'm more interested in opinions on their support in a business setting since I've not seen either's support talked about much online.
>>
>>101094230
>how can I determine if it's a PWM one or not?
what equipment do you have on hand? circuit components? multimeter? if you dont have anything this is could be difficult.
>>
File: Capture.png (5 KB, 638x570)
5 KB
5 KB PNG
>>101094497
You also got memed into overthinking about fans and cooling power units. If you are getting some minimal airflow over a power cooling components you are getting major benefits. Buying some super expensive fan gets you so little extra cooling for components that give zero fucks about being 50C versus 55C. Servers arent supposed to be quiet anyway. The fan noise is comfy.
>>
>>101094468
Somewhat related question. How's the Terraform provider for XCP-NG? Been using Terraform at work for years now with AWS but after setting up Proxmox for homelab shit I've found the Telmate provider sucks ass. There being no way to import cloud images is retarded for any form of automation.
>>
>>101094665
I'm the original asker of that question, and I'm super interested in the answer to this question as well. I've never used terraform but I've been looking into it recently and want to give it a shot, it would go a long way for me if the terraform provider is good.
>>
>>101094665
https://registry.terraform.io/providers/bpg/proxmox
It's still ass, less than Telmate. But it supports importing disk images, with a PAM account using SSH.
>>
Is there anything wrong with this container script? I want to run a media server with the iGPU passed through.

podman container create \
--device /dev/dri:/dev/dri \
--publish 10.10.99.32:1900:1900 \
--publish 10.10.99.32:7359:7359/udp \
--publish 10.10.99.32:8096:8096 \
--publish 10.10.99.32:8920:8920 \
--pull=newer \
--restart on-failure \
--volume /home/main/appdata/jellyfin:/config:Z \
--volume /home/main/jellyfin-cache:/cache:Z \
--volume /home/main/media:/media:z,ro \
docker.io/linuxserver/jellyfin:latest
>>
>>101094783
That is at least a bit better. Still annoying that it's through SSH but I guess that's more Proxmox's fault than the provider's.

>>101094468
Back to you, from what I've experienced so far Proxmox is fine for home use once you have a template in place.Spinning up VMs from templates and managing hardware passthrough, networking, storage, etc, is painless enough through the web UI. The main pain is in automating any of the above. In short it seems the only thing Proxmox really exposes through an API is spinning up/killing VMs or LXC containers. Anything else you're essentially interacting directly with the underlying KVM running on the Proxmox host. That may be acceptable to you given your existing KVM workflow. I don't have any experience with Proxmox's support, so can't comment on that.
>>
File: Screenshot_6.png (520 KB, 1086x609)
520 KB
520 KB PNG
I'm in the market for an epyc and there's a whole bunch of combos for supermicro h12ssl-i motherboards and rome chips. Should I buy one?
>>
>>101095229
Why not DDR4 3200 MT/s ECC REG?
>>
>>101095264
because 2133 is what they're bundling. It's just under $300 for 256 gigs.
>>
File: IMG_1677.png (829 KB, 720x960)
829 KB
829 KB PNG
>>101095264
I also don't think this would hurt my too much as this is primarily a file server, serving 10gig. A ram cache is going to saturate that easily anyways. If something about what I'm, saying sounds retarded let me know. I'm not very versed with server harware.
>>
>>101094541
Doesn't matter, my chassis fans provide pretty substainstial cooling over the PSU anyway in near silence (Noctua do make static pressure fans and they far outperform the stock fans this 3u chassis had)
this is the last fan and it makes more noise than everything else in my rack combined, I suspect it is also broken
>>101094619
>Servers arent supposed to be quiet anyway.
who cares, it's my server, I want it to be quieter so I can hear the HDDs thinking
>>101094497
I'll try this, theres no risk of damaging the fan? as long as it's the right voltage (I have a multimeter to check)
>>
>>101095403
>I'll try this, theres no risk of damaging the fan? as long as it's the right voltage (I have a multimeter to check)
It genuinely doesn't even matter if it fucks the new fan, go through amazon, essentially guaranteed returns "i put the fan in my PC and it didn't even spin, defective unit, please refund" and send the fucker back.
>>
>>101095443
I refuse to buy anything from amazon, also amazon shipping and prices to where I pive in rural aus would make a single noctua cost more than food for a fortnight, let alone return shipping (which I gauruntee they'd make me pay and it's like $100 to send a fucking DVD anywhere let alone a fan)
theres a local store I buy from and they actually sell basically at MSRP which for australian, let alone rural australian, stores is fucking insane
>>
>>101095475
>I refuse to buy anything from amazon
I get it, don't want to feed the beast, but you have to realize I'm less saying to give your business to amazon, and more saying to give your risk to amazon. Feel free to order from Amazon, and return the thing whether it's broken or it works, that way you stiff amazon out of a sale, the price of shipping two way and if you do zap the fan, they lose the fan, but if you don't zap the fan, they have to put money into finding out if it's actually working or not then they'll need to mark it down for being used. Plenty of ways to fuck the man on this one, then you get to test the fan without much thought or time and you can keep the relationship with the local store there good. If it works from amazon, return it and buy from the local store, if it doesn't work, just don't buy the fan.
>>
>>101095757
>don't want to feed the beast
I also just don't want to have an amazon account but yeah all that too, I guess
but yeah, to get a noctua fan delivered from amazon (I wouldn't get a prime subscription for this alone let alone that prime doesn't work here) would cost me nearly $180 aussie dollars, or about $120 USD. for a fan I can buy down the road for $40 and return just as easily because we have consumer laws here and don't need to scam businesses just to have ethical rights
>>
>>101095757
>>101095870
like, to me it's less about not giving amazon money and more about not letting amazon know I exist
but it's also aboyt not giving them four times as much money as anyone else for the same product

The guy who runs the store's name is Jerry, I could probably just ask him to borrow one
>>
I love you, home server general anons
>>
>>101095890
i love you like I love my data
>>
>>101089602
Is anyone using calibre to store ebooks? I'm trying to make use of the send to kindle feature to get ebooks over to my wife's kindle automatically, but I can't figure out an email provider that will work with the feature. It suggests gmx but all of my account creation attempts get flagged as suspicious. And I made a throwaway gmail but it looks like gmail did away with allow third party services to send emails. Am I going to have to make an outlook account?
>>
Question anons:
Has anyone here used a Banana Pi board to run OPNsense? I'm looking for a good option with a small form factor.
>>
File: 1687187944560373.png (72 KB, 1105x1053)
72 KB
72 KB PNG
>>101095890
>>
File: 1630098628113.gif (1.37 MB, 430x360)
1.37 MB
1.37 MB GIF
>>101089602
Does anyone use matterbridge? What is it exactly? Is it a standalone chat server? Or is it like a bridge in the matrix sense where you need a separate chat server to use with it?
>>
so, either:
1: my oldest drive (6 years 281 days) died today
2: I didn't plug it back in all the way when I cleaned the bays yesterday
it's 9PM and the server is 45 minutes away. It can wait till morning.
I have two spares anyway.

>inb4 >RAIDZ1
Yeah, I never cared enough to buy another drive nor did I have enough bays on the previous server to support both the capacity I needed and the drives I'd need. It'll all be moved to a RAIDZ2 array soon anyway, provided it lives that long... which it will. Not gonna bother a week-long rebuild and risk another failure when in 2-3 weeks I'll have bew drives anyway
>>
>>101097704
good luck, anon
>>
https://www.youtube.com/watch?v=GDjVUtY-0Xk
>>
are those usb bays for hdds are any good for a small local homeserver?
>>
>>101098237
People will tell you that they aren't, I have 2x5 and 1x4 used with 3 SFF pcs. They work "fine", they aren't great, for example if i put a non functioning disk in then none of the disks will be recognized.
>>
>>101098237
I use those only for backup of my server
>>
i have a useless w510 stinkpad sitting and collecting dust, can i use it for a homeserver if my only usecase is jellyfin and nextcloud? maybe hosting email and a website
>>
>>101098467
wouldn't use an old laptop if you want 24/7 uptime
just get a used minipc/sff
>>
>>101098467
jellyfin will benefit from a CPU that has hardware transcoding, in Intel's that's called "Quick Sync" and there are various versions with various capabilities. Quick Sync was introduced circa 2011 which i think is the generation of CPUs after what you've got in that thinkpad.
>>
>>101098516
thanks. any recommendations for cheap mini pcs? or what to look for
>>
>>101098593
I've seen a lot of people use the elitedesk or a thinkcentre, be mindful of what this>>101098571
anon says and check the processor gen you'll need 7th gen or above to be certain for transcoding
>>
>>101098593
>>101098606
I have a HP Elitedesk mini and Fujitsu Esprimo which i picked up cheap used off ebay and have run both jellyfin & nextcloud on both of those. As the anon says these are various models so you can find much older and newer ones. My Elitedesk is an 8th gen i5 and Esprimo 10th gen i5.
>>
>>101098237
see >>101091522
also add in that a lot of those USB docks have their own RAID controllers that can cause a lot of issues, for example by not allowing you direct control of the drives for checking SMART status, or by setting RAID level uses physical dip switches that you better hope you never change (or it's byebye data)

it'll work, but theres enough downsides for people to say "keep them away from important data". keep a backup anyway, in fact, these things are fine for backups
>>
if I were to buy a NAS do I have to directly connect it to my server or is connecting it to my router fine?
>>
>>101099106
if you directly connect it to the server, you'd need a dedicated network port for that
if going through the router, you'd want the router to be able to handle the throughput (generally fine)
many people use their NAS as their server, so if you're in the planning stage be aware that a NAS is just a computer with a network connection and storage
>>
>>101098467
It's not a server and it's trash, toss it.
>>
>>101098593
>>101098606
>>101098634
These are not servers either. Mini PCs are on-topic in /hsg/ for their intended purpose, as thin clients, for VDI you build at home.
>>
>>101099106
Neither. Connect end-user devices to your access switch(es).
>>
>>101099355
cool story bro
>>
>>101099355
you're enterprisefag aren't you
>>
File: 1683771410384101.jpg (164 KB, 1079x1042)
164 KB
164 KB JPG
Someone I know has gifted me 25 used 500 GB 3.5'' HDDs from the company he works at and is willing to give me more. I have 3 servers right now and storage is not really a concern right now, plus 12~ TBs is not that much anyways.
Any interesting projects I could do with these drives or should I just tell him to not keep giving me e-waste? My current servers basically serve Syncthing, Plex, a few web apps and nothing else.
>>
>>101100201
Trash.
>>
>>101100201
i still have a couple of these kicking around taking up unused bays in my servers and honestly i can't wait to get rid of them.
>>
>>101100201
who the hell is only just decommissioning 500gb hdds in current year
>>
>>101100201
Backups?
>>
>>101100201
do a crazy RAID config for experimenting
>>
it's ok and on topic to talk about your server mini pc issues but let's keep mini pc suggestions to /pcbg/. My 2 cents
>>
>>101100490
suggestions were for use as home servers boomer.
>it's only a server if bigco says it's a server
>>
don't reply to enterprisefag
>>
You'll hurt yourself by not having a serial out or BMC management interface.
>>
You'll hurt yourself by not having hot swappable front disk trays, or at least capacity for six storage devices (two for boot, four for VMs).
>>
You'll hurt yourself by using ZFS, instead of LVM with ext4 or XFS capable of Network Bound Disc Encryption (NBDE) using clevis & tang.
>>
>>101100201
break them open for a set of fancy shiny coasters
>>
You'll hurt yourself by using Proxmox VE technologically by not having a web UI or API to upload cloud image templates/disk images and financially for servers with more than one socket for licensing.
>>
>>101100201
archive for /t/ megapack
>>
What's the best way to place multiple running laptops together? Should I stand them up like books? Something tells me I shouldn't rest them flat on top of each other.
>>
>>101100910
What does this have to do with servers?
>>>/g/sqt
>>
>>101089602
>Learn about networking by standing up a OPNsense/PFsense box and configuring some VLANs.
This meme also needs to stop. CML and EVE-NG are much better ways.
>>
>>101089602
hey /hsg/, what do you recommend for tape drives? I'm suddenly realizing that I need to archive every youtube video I watch and I do not have cheap HDs.
>>
>>101095229
>H12SSL-i
>Dual gigabit LAN via Broadcom BCM5720
Trash. You may want to spend money on a better NIC, and faster memory.
>>
File: file.jpg (22 KB, 382x550)
22 KB
22 KB JPG
Are NAS branded HDDs a meme or are they worth the extra money?
>>
>>101100917
they're running servers, obviously. I thought that would be the only reason to have multiple running but closed laptops
>>
>>101101052
Check compatibility with your NAS appliance. They're typically fine for archival and backups, not meant for VM or random access workloads.
>>
>>101101069
I don't have a NAS appliance, I'm thinking of building a server to do double duty as a Jellyfin thing and network storage.
>>
>>101101039
What makes you think tape is the best option? I ask this as somebody who has never worked with tape.
>>
>>101101042
I've been doing more looking and I was thinking about a 7f52, but I think I'm leaning toward picrel.
>>
>>101101090
SSDs for hot data.
>building a server
ISHYGDDT
>>
>>101101052
The main thing you need to avoid is SMR disks, manufacturers don't always make this easy to find out. Beyond that it really depends on your budget/risk/maintenance appetite. Are these disks part of a RAID/ZFS/Distributed storage system, can you handle downtime, are they backed up properly.
If the answer to these questions is, dunno, i'd suggest just buy the cheapest by density non SMR hdds you can find.
>>
>>101101111
cool digits
>>
>>101101111
>ĹŻ
nice binary digits
>>
It can take up to 5 minutes to restart a docker container. How can I debug why this is happening?
>>
>>101101103
Dell PowerEdge R7515 costs almost as much on britbong eBay as your setup, pre-built?

>RAM
Sure, why not. It's not Supermicro's validated RAM under their own brand (which is also not too expensive), but it'll probably work.
>H12SSL-NT-B
That one also has Broadcom BCM57416 NICs. Unlucky.
>AMD EPYC 7402
The listed title doesn't say if the CPU is locked to Dell servers, or not locked to any OEM. Buyer beware.
>>
>>101101093
It has a massive storage potential for a very low cost.
It is TERRIBLE for read and seek, basically- it's something you set once and then never touch again in hopes you never need it.

If Youtube really does go to shit and starts embedding a metric shitton of ads in every few seconds of every 5 minute video I've ever watched, I want to have them downloaded and archived, but I won't necessarily need to watch them all of the time until that happens.

So my thinking on tape is that instead of getting NAS ready drives which are expensive, simply get Tape and archive what I need, and then prepare for a future where I may need it.
>>
File: 1695490825899868.png (339 KB, 881x765)
339 KB
339 KB PNG
>>101100335
All of my servers serve as backup nodes, so I got that covered. Thanks.

>>101100392
Honestly, this is the only interesting suggestion so far. If you see someone posting a 25-drive RAID 0 striping on /hsg/, it'll be your fault.
>>
>>101101224
Also tape has higher failure rates from even small environmental changes. To the point where some people suggest putting tape drives in the fridge when you're not using them.
>>
File: 1689236898004625.png (1.55 MB, 1453x2000)
1.55 MB
1.55 MB PNG
I think I might get my own home-server.
>>
>>101101195
That seems to be about what the 7402's go for so I think it's fine. Ebay buyer protection is also great. Are the nic's that big of an issue? I wanted onboard 10gbe to save pcie slots and not have to rig a fan onto a nic, but if they're not great anyways I could save some money by getting the verison without 10gig and an add in card. I couldn't find memory for the H12SSL but I imagine it's close enough.
>>
Hey question
There used to be a linux service called 'Privoxy', and one of its features was the ability to sit as a proxy server to a machine on a local network and act as its caching service, storing all cached content on a NAS and then serving it back when it detects the same cached content.

This, of course, is useless nowadays with SSL. But my question is: is there another service like this? I know it effectively has to do a MITM attack on your network in order to do caching now, but something that just caches my normal browsing habits in case something happens to the things I watch/read would be extremely useful.
>>
>>101101430
>I couldn't find memory for the H12SSL but I imagine it's close enough.
MEM-DR432L-CL04-ER32-BD
https://www.ebay.com/itm/394896841415
https://store.supermicro.com/us_en/32gb-ddr4-3200-mem-dr432l-cl04-er32.html

Other validated memory: https://store.supermicro.com/us_en/validatedparts/result/index/?cat=11&q=MBD-H12SSL-NT
>>
>>101101224
Personally for this use case I'd be wanting replicated data with active periodic scrubbing. How much data are you anticipating? I run a mix of SSDs and HDDs, my set up is pretty elaborate and i take comfort in the standard hardware. FWIW while i have a few 'NAS/Enterprise' HDDs i've only bought them when the deals made them price competitive, I don't pay more for it.
>>
>>101101530
>https://www.ebay.com/itm/394896841415
Oh that one wasn't OEM validated, but a third-party generic replacement.
>>
>>101101543
>How much data are you anticipating?
I do not actually know what the total footprint is of all the youtube videos that I watch and would like to keep.
>while i have a few 'NAS/Enterprise' HDDs i've only bought them when the deals made them price competitive
How price competitive are we talking?
>>
File: file.png (658 KB, 1024x468)
658 KB
658 KB PNG
>>101101376
So far, I've decided to not get a raspberry Pi, and pick up a mini-PC. Possibly pic related depending on whether I can get decent storage for it.
>>
>>101101554
My model is to have at least 3 replicas of everything i consider valuable distributed across at least that many hosts, and I have offsite backups. So for that I'm comfortable buying used enterprise (with appropriate SMART values) hard disks when i can find them cheaper per TB than new consumer disks.
>>
>>101101615
That's not a server, anon.

What are you going to use this for?
How much are you willing to spend?
How are you going to add storage?
How are you going to manage this remotely and headless without serial out or BMC, if you're locked out of your computer?
Do you love your data?
You will have hard time building reliability or clusters of servers with only one NIC, and no redundancies.
Your options for hypervisors to run virtual machines are severely limited with a thin-client.
>>
>>101101648
>Offiste backups
Do you use a service for this?
>>
>>101101615
ignore enterprise anon, these SFF pcs can work just fine for server use cases.
Definitely makes a lot more sense that Pis
>>
>>101101703
I use backblaze b2 with restic
>>
File: 1669490660845726.jpg (102 KB, 623x579)
102 KB
102 KB JPG
>>101101680
>>101101705
My idea was just to buy a mini-PC + pre-built NAS and filled it with the data I love (porn)
>>
>>101101726
Yeah you'd literally be throwing away money buying enterprise hardware for this sort of use case.
>>
>>101101747
What if I wanted to start my own smut enterprise?
>>
>>101101749
>smut enterprise
If you want failover and don't need 100% uptime (e.g maybe a few seconds downtime when a node fails) I'd still suggest just buying a handful of these SFF machines used. I am biased because that's exactly what i have on the desk next to me.
>>
>>101101430
>Are the nic's that big of an issue?
No, but vendors like Starwind like to encourage to use Mellanox for inter-cluster data (> 10 GbE), Intel NICs for management (BMC NIC can be whatever) for the optimal setup.
https://www.vmware.com/resources/compatibility/detail.php?deviceCategory=io&productid=45597
https://www.starwindsoftware.com/v20/assets/pdf/system-requirements.pdf
https://serverfault.com/questions/28009/how-can-you-tell-which-nics-offload-processing-from-the-cpu
>>
Posted a while back about this and got no useful reply. Please help a guy out regarding his computer:
sudo btrfs fi usage /media/user/btrfs
Overall:
Device size: 14.55TiB
Device allocated: 8.32TiB
Device unallocated: 6.23TiB
Device missing: 0.00B
Used: 8.26TiB
Free (estimated): 3.15TiB (min: 3.15TiB)
Data ratio: 2.00
Metadata ratio: 2.00
Global reserve: 512.00MiB (used: 0.00B)

Data,RAID10: Size:4.15TiB, Used:4.12TiB (99.26%)
/dev/sda 2.08TiB
/dev/sdb 2.08TiB
/dev/sdc 2.08TiB
/dev/sdd 2.08TiB

Metadata,RAID10: Size:10.00GiB, Used:8.32GiB (83.17%)
/dev/sda 5.00GiB
/dev/sdb 5.00GiB
/dev/sdc 5.00GiB
/dev/sdd 5.00GiB

System,RAID10: Size:128.00MiB, Used:464.00KiB (0.35%)
/dev/sda 64.00MiB
/dev/sdb 64.00MiB
/dev/sdc 64.00MiB
/dev/sdd 64.00MiB

Unallocated:
/dev/sda 1.56TiB
/dev/sdb 1.56TiB
/dev/sdc 1.56TiB
/dev/sdd 1.56TiB

btrfs fi df /media/user/btrfs
Data, RAID10: total=4.15TiB, used=4.12TiB
System, RAID10: total=128.00MiB, used=464.00KiB
Metadata, RAID10: total=10.00GiB, used=8.32GiB
GlobalReserve, single: total=512.00MiB, used=0.00B

Some say I should rather store all metadata as RAID1 since it barely takes up space and the redundancy protects in case of a 2 drive failure. Do you recommend me doing so and can I simply rebalance with
btrfs balance start -mconvert=raid1 /media/user/btrfs
?
Also it is obvious there is 4x4TB = 16 TB installed. I would expect 8 TB of space. Is it correct that the discrepancy here between my expectation and the above reports is due to BTRFS allocating space as it is needed ? So if I now chose to dd 3 TB into a file on the drive it would start allocating the 3TiB dynamically ?
Thanks.
>>
File: 20240621_171749.jpg (90 KB, 742x1209)
90 KB
90 KB JPG
>>101101530
Well this is why I'm posting and not just buying the first thing that looks 'good enough'. Is there a large difference between validated and unvalidated memory?
>>
>>101102070
No, except cost. Unvalidated generic memory will most likely work (at JEDEC speeds and latencies) and will be cheaper, but the server OEM or motherboard manufacturer won't help support you if they find you to be using non-validated memory for the platform.
E.g. I use Supermicro memory in my Dell servers and have not encountered issues, but Dell obviously wouldn't support me (if I had a support contract) in this scenario.
>>
>>101094230
PWM fans work without a PWM signal too, it'll just run at full speed all the time.
>>
>>101095382
That system is insane overkill for a file server. You could run a file server off a quad core and that'd probably be overkill too.
>>
>>101102684
NTA. Not at all, the minimum is 24 cores per vSAN-Max-XS node for storage.
>>
^this is enterpriseschizo, don't reply to him
>>
>>101102740
is it some kind of obscure sexual fetish?
>>
>>101101747
No they wouldn't. Future proofing. I'm glad I got powerful equipment when I designed my server because it allowed me to run AI processing, heavy transcoding and other tasks long term without bottlenecking the rest of my services. Never know what may interest you 2-3 years down the line.

My server started off as storage for porn only and I got 100% deep into networking/servers/etc.
>>
systemd-analyze blame

14min 51.087s docker.service
1min 30.028s archive.mount
1min 30.026s truenas.mount
30.019s nut-server.service
5.802s crowdsec.service
4.421s crowdsec-firewall-bouncer.service
2.714s nmbd.service
2.711s nginx.service
2.441s webmin.service
2.033s postfix@-.service
1.398s nut-driver@ups.service
1.084s networking.service
879ms systemd-udev-settle.service


docker.service time is fucking 14minutes?
>>
>>101103305
You don't need enterprise to do any of that, okay SFF might not be able to do all of that but it's a great place to start and expand as you need.
>>
File: 1719084640847.webm (1.4 MB, 640x640)
1.4 MB
1.4 MB WEBM
>>101103316
>Docked
>>
>>101103492
Of course not, but buying used enterprise equipment off eBay is cheap, so why not. You can always swap parts out as needed.

When I say "enterprise", I'm mainly just talking about the chassis + backplane anyway. I need a 24X+ bay chassis, and only enterprise equipment has that range.
>>
>>101103645
>why not
because the anon who was asking was already planning to have a separate NAS so clearly didn't need anything like a 24 bay chassis. Used enterprise equipment is more expensive, larger, older, noisier and more power hungry than a modern SFF which would suit anons use case. I didn't say there aren't ever good use cases for enterprise.
>>
>>101103753
>Used enterprise equipment is more expensive, larger, older, noisier and more power hungry than a modern SFF
If you call 29-34 dBa noisy, ok.
At scale of computing, even an 8-10 year old rackserver is more power efficient than a thin client. Race to sleep.
>>
>>101103817
I'm glad you're happy with your purchase anon but you're encouraging beginners to spend more money than they need to.
>>
>>101103817
Not everyone has room for a rackserver
>>
>>101103857
>you're encouraging
I'm not >>101103645, although people typically refer to me as the resident enterpriseschizo who contributes a quarter of posts in these threads. I actually agree with >>101101747, it's some other anon pushing for enterprise hardware for this use case. I stopped when that anon wanted a NAS and a desktop computer to use the NAS.
But I will continue to argue your thin clients are not servers, like there: >>101101680
>>
>>101103988
haha i think you're totally nuts, like two politicians arguing over the definition of a "woman" but at least you're nta
>>
>>101099878
pathetic europeans who have to pay out the ass for electricity who feel ego harm from the mere suggestion of something capable of IPMI.
>>
>>101103857
buy once cry once. buying something that's a piece of shit with limited features will pigeonhole your skillset.
>>
File: 1700417300060995.png (1.73 MB, 1200x675)
1.73 MB
1.73 MB PNG
Is 10GB MoCA a thing, or am I limited by the technology of my time?
>>
>>101103753
>more expensive
wrong for secondhand
>larger
this is what racks are for
>noise
not all enterprise devices are noisy, and if you have a garage or basement it doesn't matter
>more power hungry
yeah that's the difference between piece of shit equipment and good equipment

faget europoors
>>
/hsg/ I finally got torrents running on truenas with a VPN all in one container. My media server is DONE and I feel like I just snorted cocaine, the euphoria i'm getting from this satisfaction is like when Patrick Bateman kills those fuckers in the apartment from American Psycho. Thanks for all the help guys. May you all achieve some happiness in this shit world.
>>
File: 1718910482177053.jpg (80 KB, 1024x702)
80 KB
80 KB JPG
>>101104368
Nice anon!
>>
>>101104277
0.066 EUR (6.6 cents)/kWh or below for the past four months, Nordic average. I don't understand this meme.
>>
>>101104396
i like how you picked the most energy rich segment of all of europe to make your argument juden

average euro pricing is 0.26 EUR

i would tell you to kill yourself but you cant own any weapons whatsoever and you'd probably have to pay a somalian to do it for you
>>
>>101104498
Year 2023 prices?
>>
>>101101914
>Do you recommend me doing so and can I simply rebalance with
yes, you can also do raid1c4 for extra paranoia
>Also it is obvious there is 4x4TB = 16 TB installed.
your drives look like they are 2TB, and are nearly full. don't rebalance when it's that full
>>
>>101104368
you put all the arrs + the torrent client in one container?
>>
File: aaaa.jpg (183 KB, 3564x1629)
183 KB
183 KB JPG
I don't know if this is the right thread, but

My ISP provided router has a 2.5 gig WAN port. I confirmed it is receiving 2.3Gbps. Only problem is the provided router has only a 1 gig LAN port. Can I connect a 2.5 gig ethernet switch to the WAN port, and connect my PC to the switch so I can receive the intended speed (2.3Gbps)?
>>
>>101105292
Nah, I use a few databases and don't feel like automating that stuff. I'll install sonarr in a bit and maybe some others to get some labeling taken care of. Since plex kinda handles that labeling im not in a huge rush
>>
>>101105296
>is the provided router has only a 1 gig LAN port. Can I connect a 2.5 gig ethernet switch..
yeah as long as there isn't some weird setting or configuration for the WAN port(i doubt it) you should be able to just plug and play. That's kind of what I did for mine. My isp sends fiber in and I just use a multi-layer switch to route the cat6 wired throughout the house.
>>
>>101105296
What does this have to do with servers?
>>>/g/sqt
>>
>>101105296
>Can I connect a 2.5 gig ethernet switch to the WAN port, and connect my PC to the switch so I can receive the intended speed (2.3Gbps)?
You can but then you won't have a router/firewall, you should just get a router that can do 2.5g. and it's needs to be somewhat beefy to actually handle that much throughput.
The isp is dumb, they should be giving you a router that can actually route all 2.5g to a single computer.
Also does your computer actually have a 2.5g lan port? If not you will need to get an add in card or you will be limited to 1g anyway
>>
File: 1718565608637600.jpg (226 KB, 600x600)
226 KB
226 KB JPG
Why are epyc motherboards so expensive. It's lile twice the price of my cpu. $300 cpu $580 mobo.
>>
>Search eBay for Kioxia PM5-R (7.68 TB)
>$750 USD each
>Search the part number KPM51RUG7T68
>$600 USD each
Why are sellers like this?
>>
>>101105817
Different sellers targeting different audiences
>>
>>101105817
>Search ebay for 7.5 TB Enterprise SSD
>$1,500 USD
yup
>>
>>101101680
>How are you going to add storage?
This thing supports a SATA drive and up to three NVMe drives depending on the exact system configuration. Utilizing USB, NVME-SATA controllers, iSCSI, FCoE or any network file system Anon could expand the storage capacity of his mini-pc to literally multiple petabytes, no problem.
>How are you going to manage this remotely and headless without serial out or BMC, if you're locked out of your computer?
It's a home server. It's in your home, not in a datacenter two hundred miles away. If it doesn't work you connect it to a screen, plug in a keyboard and resolve the issue.
>Do you love your data?
If he does, a mini PC will serve him just fine. He should do backups, just as he should on any other potential platform.
>You will have hard time building reliability or clusters of servers with only one NIC, and no redundancies.
I bet you keep a ton of batteries and a diesel generator in your garage, while still having geo-redundant power feeds connected to two different substations. Of course you have a fire extinguishing system in place for your home server, as well as security and technical engineers on-site 24/7. All your hardware is covered under 24/7/4 service contracts and your core, distribution and access switches are fully meshed to give you that quad-nine uptime so you don't need to live without your beloved chinese cartoons for more than 3.5 minutes a year.
>Your options for hypervisors to run virtual machines are severely limited with a thin-client.
The ThinkCentre is based on a standard AMD64 platform with VT-x support so it should run ESXi, KVM, Qemu and Hyper-V just fine.
>>
>>101105675
PCIe 4.0/5.0 and many lanes, many PCB layers.
>>
noob here
I am going to ask my ISP for a public IP
how do I secure myself over not exposing everything in my local network? or only things in port forwarding are going to be exposed?
anything else I should do to secure myself
>>
>>101106006
>anything else I should do to secure myself
Get diaped up
>>
>>101106006
you already have a public IP, retard.
>>
>>101106006
If you're using IPv4 then a typical NAT router will prevent incoming connections except for port forwards. IPv6 setups can be a bit more varied.
>>
File: 1718331386201944.jpg (105 KB, 853x1024)
105 KB
105 KB JPG
>>101105962
Yeah the ones with pcie3.0 are much cheaper. Does server hardware get a price drop when a new line releases like cosoomer lines do?
>>
>>101106025
>>101106024
>>101106029
i mean like static public ip so im not behind a nat
>>
>>101104920
Now thats where the irritation comes from. I bought 4Tb drives. lsblk calls them 4TB and before I moved to BTRFS they were 4TB.
sda         8:0    0   3,7T  0 disk /media/user/btrfs
sdb 8:16 0 3,7T 0 disk
sdc 8:32 0 3,7T 0 disk
sdd 8:48 0 3,7T 0 disk

Which is why I am asking.
I sort of believe to have figured out
Unallocated:
/dev/sda 1.56TiB
/dev/sdb 1.56TiB
/dev/sdc 1.56TiB
/dev/sdd 1.56TiB

Implies I have 2.08TB + 1.56 TB of real estate per disk and using raid10 I am getting half of that total of storage. And btrfs would start allocating the remaining 3.12 TB once needed. Wrong ?
Regarding raid1c4 do I want to
btrfs balance start -mconvert=raid1c4 /media/user/btrfs/

?
>>
>>101106050
There are different kinds of NAT, I'm guessing you mean you want a fixed IP so that you aren't behind CGNAT where your public IP is shared with others. When i say a typical NAT router will prevent incoming connections I refer to the router you may have at home you shares your connection with the devices on your local network. This may or may not feature NAT depending on if it's IPv4 or IPv6
>>
>>101106050
You are only not behind a nat if you choose to be
>>
>>101106050
even if they take you off CGNAT you may still have a dynamic IP. static IPs cost money.
>>
>>101106006
If you need to ask this question on /hsg/, then you need a firewall.
>>
>>101106075
retard
>>
>>101106079
>static IPs cost money
depends on your ISP
>>
>>101106025
he's probably behind CGNAT, in which case no, he doesn't
or he means a statis IP
>>101106006
either way, your firewall should block everything incomming by default, basically nobody from outside your network can open a connection to you. Only clients inside the network can request content from outside via NAT or UPnP, and the only way to allow an external IP address to facilitate a connection is to port forward.

Basically your router is a door that only people from inside can open. If you port forward, that's like adding a cat door that only allows cats in and only to the place you put the cat door
>>
>>101106103
Maybe that anon doesn't have a firewall, but only a switch. You're also implying NAT is a security feature or that his router does NAT.
>>
>>101106050
You should still be behind NAT
there are three things you're confusing here
Static IP - an IP address that does not change, typically because you pay to have that IP
Dynamic IP - an IP address that can change via DHCP, much more common than static ones to help with the lack of available v4 addresses
Public IP - an IP address in the public address space accessable directly from the internet
Private IP - an IP address inside one of the reserved private ranges (192.168.0.0/16, 172.16.0.0/12, 10.0.0.0/8) that is not routable on the internet unless via NAT
NAT - when a router converts your local IP address (eg 192.168.0.1) to an address in another range (usually a public one on the internet, eg 210.10.17.38) to allow your machinr to stay securely in an unroutable LAN while still providing one-way internet access. Nearly all routers do this by default
CGNAT - when your ISP has you behind a second NAT layer, meaning that you do not have a public IP address (your "WAN" is just inside their "LAN"). this is done to lower the number of public addresses the ISP needs ro pay for, as you can have thousands of customers sharing one public IP
>>
>>101106133
NAT is not a security feature, HOWEVER
>>
>>101106174
Sir,
fd00::/8
is also a private range (ULA). I need not explain TEST-NETs and such.
>>
>>101106211
>>101106174
More importantly,
100.64.0.0/10
for CGNAT.
>>
>>101106133
>Maybe that anon doesn't have a firewall, but only a switch
a retarded assumption
>You're also implying NAT is a security feature
in conjunction with good firewall rules, it can be
>that his router does NAT.
every consumer router has NAT by default. every single one. you cannot convince me anon has anything other than a consumer router. even the majority of "enterprise" routers do NAT to a default network these days.

You are so out of place here, you keep on making these retarded assumptions and telling people they shouldn't do what they are doing. you're in the wrong thread. You're going to the car meet in an F1 and telling people they shouldn't assume someone elses car runs on petrol because theres a single edge case (you) that runs ethenol
>>
>>101106103
uhh he has a shared public IP. anything that isn't RFC 1918, or APIPA, or a default route, or a broadcast address, or a network address is a public IP.
>>
>>101106133
>You're also implying NAT is a security feature
also, where did I imply this? I explicitly stated the firewall blocks all the connections, and that his local network can only reach the internet via NAT
both of those things are true, and that dies not in any way shape or form imply that NAT is doing the security.
>>
>>101106251
that's not what nearly 100% of people mean when they say "public IP"
if his WAN address is not direcrly routable to the internet because it is behind CGNAT, he does not have a public IP but a CGNAT IP.
>>
>>101106251
>>101106103
also multicast, forgot that one
>>
>>101106251
>anything that isn't RFC 1918, or APIPA, or a default route, or a broadcast address, or a network address is a public IP.
This is not correct.
Addresses in the 192.168.9.0/16, 172.16.0.0/12 and 10.0.0.0/8 ranges are private IP addresses and are not routable via the internet. Most (but not all) other ranges are considered public as they are routable via the internet.

https://community.cisco.com/t5/networking-knowledge-base/networking-basics-private-ip-address/ta-p/3118222#:~:text=Private%20IP%20addresses%20are%20not,to%20another%20unlike%20private%20IP.
>>
>>101106315
Have you read RFC 1918?
>>
>>101106275
that's what it means literally, retard. it also IS routable to the internet in that traffic can transit back and forth to the network. they can not receive unsolicited connections inbound. but, stateful connections are tracked. if it wasn't routable to the internet then udp wouldn't work at all.
>>
>>101106315
are you retarded? that's literally what i said along with other stipulations that you probably don't understand.
>>
>>101106251
>>101106315
Key point being unroutable from the internet.
An address in a CGNAT range is not a public IP address as it is not directly routable from the internet.

You can be an autist all you like but it's just going to make comunicating with everyone else a frustrating experience as you say words that everyone else has a different meaning for. There is merit in being understood over being "correct".
>>
>>101106353
the CGNAT address that hosts multiple customers is a public address. it's that simple. if it wasn't public, it would be private, or reserved, or multicast.
>>
>>101106341
so being behind CGNAT is exactly the same in all possible ways as being on a traditional, non CGNAT public IP address and there is no need to ever make a distinction? we're never allowed to refer to the two as CGNAT and Public for easy differentiation? awesome I hated making comunication clear and easy for people that haven't done CCNAs anyway
>>
>>101106395
sure, but anon's router's WAN IP address is NOT THAT ONE
ergo HE does not have the public IP address
what's so hard for you to understand here
>>
>>101106395
>the CGNAT address that hosts multiple customers is a public address
never once denied this or argued contrary
addresses INSIDE the CGNAT range are not public, and any customer behind CGNAT does not have their own public IP address.
>>
>>101106409
you will still have an address that is in the public address space even if it is behind CGNAT.
>>
>>101106424
they are not advertised but they are public, delete your account
>>
>>101106469
>delete your account
kek
>>
should i just bite the bullet and buy a domain or can i get away with a free subdomain from freedns or smthn if i want to run a mail server?
also how does domain reputation work wrt subdomains? do they inherit the base domain’s reputation?
>>
>>101106567
free domains won't be taken seriously for email, you'll also want to check the reputation of the IP you use.
it'll take a little while to build reputation but i haven't had any issues since setting it all up with DKIM, SPF & DMARC.
Obviously advice varies if you're actually going to send bulk amounts of mail.
>>
>>101101052
i think the thing about nas drives is that if you run a bunch of drives they will start ressonate with eachother eventually and might damage the drives. nas drives has a sensor that detects this and changes the drive speed or something so it wont ressonate with other drives

dont know if this is sales fud or real though
>>
>>101106103
my router has basically no settings when it comes to firewall, only firewall spi checkbox
>>101106174
my ISP advertises it as "optional public ip" for 5$ a month
>>
>>101106930
Despite what confident sounding anons say
- not all consumer routers will use NAT, e.g if IPv6
- not all consumer routers will feature a firewall, e.g if they are IPv4
If the public ip that you can pay for is IPv4 then you will have a consumer router with NAT that creates a set of internal IPs for your own devices that map to your one public IP. In this scenario, no incoming connections will be allowed unless port forwards are added. NAT acts "like" a firewall in that it prevents unsolicited connections but it isn't a firewall. Personally i have a public IPv4 address, many devices on my internal network and while I do have a router with firewall, NAT would be "good enough" for simple use cases.
>>
>>101106059
drunk anon here

keep an eye out for "m" and "d" when configing btrfs, as this indicates meta or data
>>
what router should I buy if i want to get a nas in the future and want to use openwrt???
>>
>>101107089
XY problem. You shouldn't be using OpenWRT for this, and could benefit from a firewall acting as your router. I doubt you need BGP, OSPF and alike.
>>
>>101107005
yeah, trying not to be that goofy.
So I decided it's a worthwhile experiment, because regardless wether my suspicion or >>101104920 is correct I had the space for -mconvert=raid1c4 either way.
Turns out I might have been correcto and it would just allocate more space. This is what it looks now:
sudo btrfs fi usage /media/user/btrfs
Overall:
Device size: 14.55TiB
Device allocated: 8.35TiB
Device unallocated: 6.21TiB
Device missing: 0.00B
Used: 8.28TiB
Free (estimated): 3.13TiB (min: 1.58TiB)
Data ratio: 2.00
Metadata ratio: 4.00
Global reserve: 512.00MiB (used: 0.00B)

Data,RAID10: Size:4.15TiB, Used:4.12TiB (99.26%)
/dev/sda 2.08TiB
/dev/sdb 2.08TiB
/dev/sdc 2.08TiB
/dev/sdd 2.08TiB

Metadata,RAID1C4: Size:11.00GiB, Used:8.32GiB (75.60%)
/dev/sda 11.00GiB
/dev/sdb 11.00GiB
/dev/sdc 11.00GiB
/dev/sdd 11.00GiB

System,RAID1C4: Size:64.00MiB, Used:464.00KiB (0.71%)
/dev/sda 64.00MiB
/dev/sdb 64.00MiB
/dev/sdc 64.00MiB
/dev/sdd 64.00MiB

Unallocated:
/dev/sda 1.55TiB
/dev/sdb 1.55TiB
/dev/sdc 1.55TiB
/dev/sdd 1.55TiB

So from each disk another 10 GB or so were allocated dynamically to accomodate the increase in metadata redundancy. Hence 1.55TiB unallocated now.
Sweet. I guess I have another 3 TB for Data, as anticipated. BTRFS is just reporting in a confusing fashion.
>>
>>101107089
What does this have to do with servers?
>>>/wsr/
>>
File: 20240622_073344.jpg (277 KB, 1480x1515)
277 KB
277 KB JPG
Pulled the trigger on parts finally. Can't wait to play with my lego's.
>>
>>101106567
>i want to run a mail server?
Don't bother, you will spend 5 hours a week emailing people asking to get your domains whitelisted only for them to get blacklisted again a week later and nobody with a google email account will ever get mail from you delivered to them.
It's really just not possible to run your own email server anymore unless you want a new partime job as an email monkey.
>>
>>101107343
if you configure something incorrectly, have a domain or ip with bad history, this can be true.
However, if the above is not the case then it's really not that difficult. I've run my own email domain for years and regularly correspond with google apps / gmail / outlook etc. I spend exactly 0 time maintaining reputation since setting it up years ago.
>>
>>101107343
This is only true if you have no idea what you're doing and fuck something up (which is the case for the vast majority of people who try)
>>
>>101107460
Ive read multiple blog posts and harcker news posts from people who have been doing it for 20-30 years who say it's just not worth the effort anymore. You will get blacklisted randomly, they will ignore your emails for weeks or months at a time.
>>
Anyone here self host a journal? Whatcha use? I'm using Joplin but thinking about consolidating to just git
>>
>>101107743
syncthing + journal.txt (.md if you feeling fancy)
>>
>>101107590
I've seen this "common wisdom" as well but it just doesn't reflect the reality i live in.
>>
>>101107743
I've tried jrnl (https://jrnl.sh/) in the past, but it being Python its poor performance became a pain point after several hundreds of journal entries.
>>
>>101107743
I used Joplin for a while, currently using Logseq. It still feels a bit clunky some times but it's the best I've found to date that meets my 'requirements'
>>
>>101107743
I was thinking of trying out Joplin, Loqseq or Obsidian next.
>>
File: IMG_3340.png (5 KB, 170x170)
5 KB
5 KB PNG
xmpp bros do any of you use biboumi for accessing irc? how reliable is it?
>>
>>101107768
Share your setup then so we can try and see for ourselves.
But you won't :)
>>
Diving into setting up a VPN on my Truenas server. Never done it before.

Wireguard, Zerotier or Tailscale. I trust you guys more than my family so I don't need any real explanation.

Basically just using it to access my server apps, and files remotely.

Thanks frens.
>>
>>101109128
wireguard
>>
>>101109128
wireguard
>>
>>101109145
>>101109153
Thanks guys. You guys are awesome. WAGMI
>>
What options should be enabled for the N100 iGPU for a jellyfin server? Why are the low-power encoders disabled by default? Why is HEVC decoding disabled by default?
>>
>>101109128
No IPsec as an option? Fine.
My gripe with using WireGuard would be scalability and management for more than 1-3 devices. Please add a nice UI on top.
>>
File: Capture.png (84 KB, 2276x773)
84 KB
84 KB PNG
>>101109158
I'm a noob. These are my choices from both Truenas and Truecharts.
>>
>>101109128
It would probably be better to setup the VPN server on your firewall instead.
>>
>>101109157
>Why are the low-power encoders disabled by default?
Looks like it's a compatibility thing. The docs explain it better
https://jellyfin.org/docs/general/administration/hardware-acceleration/intel/#low-power-encoding
You might just want to read the whole article on hardware acceleration since it has some other configuration suggestions
https://jellyfin.org/docs/general/administration/hardware-acceleration/#enable-hardware-acceleration
>>
File: Capture.png (40 KB, 781x664)
40 KB
40 KB PNG
>>101109268
I think you are right. I installed wireguard and my router did NOT like it. Once I opened the port to my VPN, my router went crazy and kept booting me from the WebUI. Had to hard reboot. I think I am not qualified for this; probably lock myself out of my router for good. Attached is the basic functionality of the VPN on the router; seems pretty unforgiving for changing IP address allowance remotely. Can't see how I will know what ip addres my laptop will be when I travel aforehand.
>>
>>101109361
You're correct, site-to-site VPNs aren't made for your use case but a roadwarrior setup would be.
Is this EdgeOS? You may need to use the CLI to configure
set vpn ipsec remote-access
options and such (EdgeOS 2.0.9 or later).
https://www.vanwerkhoven.org/blog/2022/home-network-configuration/
https://help.ui.com/hc/en-us/articles/204950294-EdgeRouter-L2TP-IPsec-VPN-Server
If you had a better firewall appliance (read: enterprise grade), it'd be easier with click-ops (despite EdgeRouter still being better than UniFi).
>>
>>101109268
you should prefer a client VPN endpoint behind the firewall, rather than on it, so that it can be easily deactivated, backed up, checkpointed, and/or scanned for vulnerabilities periodically without impacting operations. if a bad vulnerability ever gets released for wireguard (one that can be scripted and mass deployed) you can still depend on infrastructure upstream to reduce attack surface area.

in the case of IPSec (intersite) VPN tunnels you should terminate on CE.
>>
>>101109361
Speaking of EdgeOS, there's WireGuard for Ubiquiti: https://github.com/WireGuard/wireguard-vyatta-ubnt
>>
File: GQaYcgtaEAApzvQ.jpg (454 KB, 3508x2376)
454 KB
454 KB JPG
>>101102684
I somehow missed that this enterprise hardware argument was started by a response to me. I ended up buying basically what was in my first post, >>101095229 but with 4x32 3200 memory instead. Reasons for needing this?

1. I needed more pcie lanes to do specifically what I wanted to do.

2. I wanted it.

3. It's cool.

4. I felt like it.

5. I'm too lazy to do something more useful with that money.

My current server is a ryzen 5 5500gt, asus rog strix x470-f gaming (old pc motherboard), an nvidia tesla m40 mutilated to fit a cpu aio for ai tinkering and a 2.5gbe nic. It also has 44TB of raw storage. The plan is to use it as a backup and eventually move it out of the house to have a proper backup.
>>
who still runnin openvpn here? and why not wireguard?
>>
>>101094230
>>101094506
There are videos on the subject, but you'll have to do some disassembly first anyway to find out what fan they shoved in it and what sort of connector it has, likely a two-pin.

>>101097107
You would have to get freebsd (assuming they even support the chip) and figure out compiling opnsense from the source.
>Cross-building for other architecures
>Currently available device are: BANANAPI and RPI2
You would be better off investigating their forums. Or just getting something with x86

>>101104368
Feels good doesn't it?

>>101109520
You're allowed to purpose build systems but when you say "file server" without all that other shit you mentioned then anons are going to point out you're crazy.

>>101109874
I use both to keep purposes/networks separate.
I still have to figure out ipsec for a roadwarrior situation.
>>
sonarr wasnt hardlink my fucking files because i forgot to add
>[Service]
above umask to the drop in file
i fucking hate niggers
>>
>>101113062
Its niggers' fault you cant config for shit?
>>
Anyone have experience with direct play add ons like jellycon (with or without kodi) my server is right next to my stereo receiver so its quite stupid to go server-phone-bluetooth-receiver and i want to just connect it with audio cable
>>
>>101089602
I want to get an enterprise-grade server with a fast single-thread clockspeed for under $2000. Any suggestions?
>>
Does frigate only work with IP cameras?
>>
I just updated my Jellyfin client and server to 10.9.6 and new/updated media is no longer getting recognized. Using apt version on debian.
Restarting, removing and readding the media library does not fix this.
What do?
>>
>>101113612
A big four-letter OEM name, Xeon E series CPUs, 1U form factor.
>>
>>101114875
>E series CPUs
>fast single-thread clockspeed
lol
>>
>>101091522
>We should have adopted eSATA when we had the chance
It's called a SAS HBA with external connectors.
>>
>>101115377
Yes, Xeon E series have a higher single-thread clockspeed than contemporary Xeon Scalable processors. Problem?
>>
>>101113612
https://www.cpubenchmark.net/singleThread.html
intel consumer CPUs top the single threaded charts and are affordable, but there's some xeons in there too
>>
>>101115451
Oh I thought you meant the E5s. E-2488 looks great but what's with the low max RAM and PCIe lane count? Is this just a renamed consumer CPU? lol
>>101115533
Im looking for server hardware
>>
>>101115590
Renamed and validated against hypervisors and other server use cases unlike Intel Core processors, but not exact clockspeeds (usually a few bins lower). E-2400 series also dropped the iGP, as the Intel Flex discrete GPUs were introduced around that time (year 2022).
Xeon W series (workstation CPUs) were discontinued; Xeon W-1390P was 1:1 specifications with a Core i9-11900K. Core processors became ECC memory enabled with workstation boards (such as W680). Xeon w (lowercase w) series then became competitors for AMD Threadripper and Threadripper PRO processors.
Xeon E3/E has always been for "entry level" servers, higher clockspeeds but less RAM and PCIe lanes.
>>
File: gowron.jpg (425 KB, 975x741)
425 KB
425 KB JPG
The consensus seems to be that used SFF business PCs are an excellent choice but I can't for the life of me find one that would support even two 3.5 HDDs.
What's the go-to used SFF PC that has room for two or more 3.5s without getting creative with drills and pliers?
>Plenty support to if you replace the optical drive
From what I've seen these optical drives are connected to SATA II ports and more often than not it would free up room for a 2.5, not a 3.5.
>>
>>101116910
hp elitedesk 800G4 sff has 2 3.5 bays and 1 2.5 bay, couple of nvme slots as well iirc
you can always get a SATA pcie extender for external hdd bays
>>
>>101116910
What does this have to do with servers?
>>>/wsr/
>>
sup /hsg/
I need extension of my storage and looking at picrel to get around the fact that most external hdd shit is usb
would the 3.5 cages be enough for shock absorption or would I break the drives by just having this stand somewhere?
>>
>>101117770
This is an incorrect approach to the problem. Buy a server chassis with more storage space (front trays), or a storage appliance.
>>
>>101117810
I don't have money or space for that
>>
>>101117770
you fine, wondering how you gonna connect this tho. I look forward to adding this to my tower that has 4x 5.25 bays
>>
What am I giving up by going with one big HDD over several smaller HDDs totaling the same capacity under the following assumptions?
1. They cost more or less the same
2. It will be used as a Jellyfin server
3. The vast majority of data would be pirated media which would kinda suck to lose all at once but if I really wanted to recover that data it would just be a matter of downloading them again, a hassle to be sure but nothing more.
4. What little important data would be stored there is already backed up.

Only thing that comes to mind is that I'd lose on the performance improvement of using some sort of RAIDZ configuration, anything else?
>>
>>101117770
I have a 8x 2.5 version of this for my fractal design 7, it's fantastic.
>>
File: ddddd.jpg (11 KB, 225x225)
11 KB
11 KB JPG
>>101118067
>wondering how you gonna connect this tho.
>>
File: HELP.jpg (31 KB, 456x628)
31 KB
31 KB JPG
Hello /g/, I need some assistance. I've got an old(ish) gaming PC that I've thrown Ubuntu 22.04LTS on, and use for plex/game servers among friends. I've been using the old case with 2 14TB HDD's from WD to store everything, and booting from a 500gb m.2 drive, but I'm looking to expand. I've gotten gifted a meshify XL case, and I want to add more drives.
Conundrum is this: I want to buy another 5 drives, and I want to move to a RAID 6 configuration and keep a drive on hand to hotswap. I also want to leave room for further drive expansion. That said, I've never set up RAID arrays before, and I don't know how to add all of these drives and set them up. Throwing a wrench in things is that I've only got 6 SATA ports available (mobo: ASRock Z270 Killer SLI/ac)
What'd be the non-retarded way of expanding the drive capacity of my motherboard and getting this RAID array set up? Perhaps a SAS SATA Expansion Card w/ a HW RAID controller? I figure I can somehow get the other drives set up in an array, locally transfer existing data, and integrate my existing drives to the array afterward.
If I'm entirely off-base, let me know, because I'm pretty uncertain about all this. While I'm fairly IT-knowledgeable, this is unexplored territory for me.
>>
>>101109472
>>101109449
Thanks. Let me look into this. I have PTSD trying to configure anything application wise in UNIX from just a shell command interface. but no guts no glory.
>>
>101117645
>101117810
This is enterpriseschizo, do not reply to him
>>
>>101118125
>>101118081
>>101118079
>>101118067
>>101117810
>>101118308

new thread you guys:

>>101118353

>>101118353

>>101118353
>>
>>101118125
these are the ultimate chink tier devices, I strongly suggest branded external storage
>>
>>101116910
HP Elite 800 SFF, G3 and newer. G4 has a plat psu.
HP Z2 SFF, you can even find some workstation variants with xeons but they're more expensive.
These variants come with more pcie slots too.

Any more than two 3.5" bays you'll have to transplant into a real case or just diy something instead. Though one anon uses the pcie slot for an hba and feeds the cable out the front to a 3.5" cage. He seemingly had some issues with power however and then died.

>>101118136
You generally can't expand a raid6 configuration.
An hba card flashed for IT mode would suffice for connecting more drives.
>>
>>101117770
I have multiple of these, and they're great.



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.