network knnower people, am i gonna have to kill this instance if i want to switch from just running everything no container managing dependencies with apt to having things in contaimers.
@prettygood im currently just running akkoma and jellyfin on a debian 13 server. ive finally managed to get some extra storage, which means i have to do system stuff anyway, and so I'd like to make things a little more manageable for running things side by side and whatnot
@prettygood@haematophage you can export db to a container just fine, it’d be a lot of custom work but you could do it just about as easily as moving an instance from one server to the next
@haematophage@prettygood you’re going to put in more effort than you’ll ever save moving it to a container unless you want to just containerize the entire current OS and then what’s the point
@sapphire@haematophage yeah I kinda agree with this, if you've got it running already on bare metal, and the software is managed sanely (via a package manager, or a git clone, etc.) then you aren't really saving a lot of effort by containerizing it. I'm actually slowly moving all of my containerized stuff *back* to bare metal because the containers cause me network consternation and I want to trim the overhead, since my VPS is quite burdened.
@sapphire@prettygood the point is that i want to potentially do more things on the server and holy fuck i have no clue how apt works i should not have used debian. i tried installing matrix and now i have no clue where packages are coming from cause i fucked up the repos.
@prettygood@haematophage I like prototyping in containers cuz I can just docker purge and nothing leaves weird 3GB files laying around but yeah once it’s stable just bare metal
@prettygood@sapphire the sogtware is NOOOT running sanely. i dont know what im doing at all wrt. dependencies and i have no clue how to use aptitude (plus i dont have a down arrow key on my keyboard cause it fell off and the squishy thing came out and for some reason the arrow keys on my numpad dont work over ssh while in an x session)
@haematophage@sapphire I don't think so, its what I have used for a public and internal server for years. If its what you know then there's no good reason not to use it. It will have some pain doing builds on slim hardware (like a VPS) but it isn't your electricity at that point.
But this is all with the caveat that you already know gentoo; a server is not a good platform to *learn* gentoo.
play around until you know what you're not doing either, since pleroma fixed the key fetch issue you can make as many mistakes as you want and not blow the domain on an uncachable rsa key in someone elses database
@haematophage@sapphire if you can do that then yes running a server on gentoo is reasonable. I don't recommend doing your own kernel on a VPS though; hardware changes can screw you over.
@prettygood@sapphire im not intimately familiar but on my laptop i've configured and compiled my own kernel and written a couple ebuilds and that sorta stuff (plus im using it daily so it's not like i have to switch "modes" to modify stuff) so relative to debian i know things
@haematophage@sapphire@prettygood worst case you can hop on an alt and ask for help, contrary to the commonly held myth there's a ton of dweebs itching to help other people run the userspace bits
@prettygood@haematophage just remember to back up your entire Postgres folder and /opt/akkoma or wherever it installs, since you’re local the size won’t matter as much and moving it whole is easier than Postgres’ backup tools imo
@prettygood@sapphire yeah it's an old laptop too so like. built in terminal access :p (older, anyway, it's got a 10th gen i7 but i dropped it a few times and the bottom of it is held together by duct tape so bringing it to lectures became infeasible, had to get a new one).
while i have u kind people here uh. any filesystem recommendations? i have three unevenly sized disks (one 6tb hdd, one 500gb ssd, both sata in an external usbc enclosure, and one 500gb (but a different 500gb than the sata? lmao one is 476 and one is 465 in lsblk idfk.) nvme inside the laptop. so raid isnt an option, til i get another 6tb drive anyway. so i dunno if going through the hassle of getting zfs set up is worth it or whatever.
idk. im just rambling at this point i dont have tech friends that want to hear about this
@sapphire@haematophage yes, silly. Filesystems are a spot where I absolutely refuse to fuck around in the slightest: ext4 only. I make and test backups of critical files, but if anything *ANYTHING* I'm doing is I/O bound then I've fucked up somewhere and no amount of filesystem gymnastics are going to save me.
@haematophage@prettygood zfs is great for exactly what it’s designed for: being a better raid. Technically you could use the two 500GB as read/write cache for the 6tb but that is very silly
@prettygood@haematophage if you have a NAS or other warm/cold storage with multiple drives ZFS is incredibly well designed for that purpose. You don’t have to understand the inner workings for it to put in work for you either
@haematophage@prettygood if you get more drives without getting a new server first I’m coming over to personally whack you. No environment deserves USB storage
@sapphire@prettygood yeah i mean i feel like the only reason to do zfs would be futureproofing for if i ever *do* have hard drives that are the same size (granted, i have zfs on root as a built-in kernel module on my single disk gentoo setup* because i thought it'd be fun to do, so it's not like id be treading new ground).
*zfs owns the entire disk that is the gentoo filesystem but there are two ssds in the laptop and the ESP is on the one that I also have arch on
@prettygood@haematophage@sapphire I like to live dangerously so I use whatever seem nice and fastest. Currently my laptop runs XFS, I'd run XFS on a server as well. I used to like BTRFS but some stuff triggered paranoia in me.
@haematophage@sapphire@prettygood ZFS isn't great for futureproofing. It's the set-and-forget file storage. Moving away from ZFS, or even resizing ZFS, usually means transferring off of that raid, destroying it, and then recreating/remigrating into it.
I've never lost a server to BTRFingers, I've never zfsck'd myself, never fumbled a postgres database... there was the time I had glibc build with specific compiler flags and then Linode changed the hardware my VPS was on so nothing worked, I'll take an L on that one.
@haematophage@prettygood speed matters a lot less when it’s on its own ghettoized pcie lane and the shitty controller in the enclosure only has 1 bit of cache shared with the write active LED
@sapphire@prettygood it's usb-c! it's 5Gbps! my download/upload speeds are barely 1Gbps anyway!!! but like. yeah my goal is ultimately to get an actual box that i can take apart without holding my breath or something but i have no job, am 11 months from a Gift Receiving Holiday, and also need to replace (at the very least, who knows what else is broken) the psu in my desktop
@AGARTHA_NOBLE@haematophage@prettygood if I had to guess it’d be somewhere around “however long it takes to write <your dataset> / <number of disks>” tard
@sapphire@haematophage@prettygood Wait "for how long" for that data to be restriped tho? Ceph degrades, but so long as you're good on replicas, you can keep truckin.
@AGARTHA_NOBLE@haematophage@prettygood even here it looks like the only caveat is you’ve gotta be an adult and actually wait for the new larger drive to populate before you move on to the next one, literally just 8x zpool offline disk, zpool replace, wait
@AGARTHA_NOBLE@haematophage@prettygood maybe if you’re adding 12x new drives at a larger capacity but why add 8 10TB drives to your 8x8tb array when you can just add two 8tb disks
@AGARTHA_NOBLE@haematophage@prettygood your inability to maintain or scale a system doesn’t make it unreliable or unscalable, it just makes you bad at computer
@sapphire@haematophage@prettygood Depends on the tool you wanna use. If you just need a few dozen terabytes on one node, and you need it to be relatively fast, zraid has you covered. If you want amazing scaling and reliability, Ceph.
@AGARTHA_NOBLE@haematophage@prettygood I’m not giving LTT even potential ad revenue but let me guess: no redundancy, disabled checksums to “get more space”
@sapphire@haematophage@prettygood And there you would be wrong. For every command line chad, there is a jungle-asian-fucking dweeb, who will manage to pull this off.