Okay. Please help me as I ask COMPUTER BABBY QUESTIONS.
I have a Thinkpad T14 Gen 3 (AMD).
It has a 256 GB HD. That's too small. I want to buy a new, bigger one. I have a sense the good hard drives these days are "M.2".
Lenovo's specs page
doesn't say anything about "M.2". It says the hd is "PCIe".
I run "lshw" to see what's on the computer. It says "NVMe".
How do I find out the bestest fastest aftermarket drive Canada Computers carries that my computer will support
I only understand computation as the MANIPULATION OF ABSTRACT PLATONIC FORMS. I do not understand this realm where computers are "physical objects" you manipulate with "screwdrivers". I would prefer to use Math to translate my thoughts directly into action, as if I am casting magic spells
@mcc The spec sheet on that pages says it takes an M.2 2280 PCIe 4.0.
So any M.2 that is PCIe 4.0 compatible and will fit in a 2280 slot will work.
@mcc M.2 is the physical slot interface. That slot can in theory accept either mSATA drives or NVMe drives, physically. The upstream controller chip could accept either protocol, or both.
Your `lshw` shows that your drive is an NVMe drive in the M.2 slot, so your controller chip supports at least that protocol. That's good: mSATA is an older, less well supported protocol these days.
The real question is whether your Thinkpad has BIOS-locked which drives it'll accept. I'll check.
@mcc The T14 have a M.2 2280 NVMe slot. I’ve never had issues with the Samsung Evo 970 models.
@mcc According to https://apac.kioxia.com/en-apac/business/ssd/client-ssd/bg5.html KBG5AZNV256G is a 2280 M.2 NVMe drive, so you should be able to fit any other 2280 M.2 NVMe drive there.
One thing to watch out for: some NVMe drives have chips on both sides of PCB, and some laptops can only fit drives that don't have chips on the bottom side. This can be a problem if you're buying larger (4 TB or more) NVMe drives (figuring out if a NVMe drive you're buying is single or double-sided is a pain – at best the specs tell you is the thickness, but even that's often missing).
Okay thank you all for explaining. I have one more question: Is there actually, like, a difference between drive vendors. Like if I pick WD vs Samsung vs Lexar will it ever make any difference
@mcc oh yes. start by deciding what spec you want, then shop by price, then make a shortlist and read reviews from reputable sources
@dysfun i dont know what a reputable source is that why im ask mastodon :( :(
@mcc it does make a difference.
These days not so much on the performance front, most generations of SSDs are roughly comparable to each other under normal workloads; but reliability differs a lot, and thermal performance does a bit.
(Non-standard workloads are things like sustained max-throughput writes which exhaust DRAM caches, or heavy load when drives are near 100% usage and can't relocate storage blocks quickly enough to meet write load.)
@mcc If you're looking for speed, choose WD SN850X (Samsung used to be good, too, but I've heard that their recent models have some problems). Otherwise just pick any drive that has 5-year warranty.
Note: if you're planning to do a lot of writing, avoid QLC drives – write speed on those drives falls below 30 MB/s once you fill the SLC cache.
@mcc Personally, I stick with drives manufactured by the companies which make their own flash: Intel (Solidigm), Kioxia (Toshiba), Micron (Crucial), Samsung, SK Hynix, and Western Digital (SanDisk).
There’s some difference between vendors, mostly in the controller firmware. This mostly matters in “enterprise” drives rather than consumer drives.
The much bigger difference is whether the drive has DRAM cache or not. In general, drives with DRAM have much higher random I/O performance than drives without it, and random performance equates to perceived “snappiness”.
@mcc Their firmwares have different bugs (I.e. the secure erase function might not work). And of course 1000 vs 1024 for size. From the OS perspective they’re interchangeable. They might have different r/w speed, and MTTF/MTBF, and maybe a different percentage of spare flash cells. I usually add drives I’m interested in to the compare list on https://skinflint.co.uk/?cat=hdssd&xf=4836_7 for easy comparison.
@schrotthaufen wait wait who defines tb as 1000 and who defines it as 1024. is this documented somewhere
@mcc someone in the replies already mentioned the Samsung EVO 970, and I want to echo that. My spouse has a T14 Gen 3 and that's the drive we upgraded it to. If you want to go with another brand, M key and 2280 are the specs to look for so that it will physically fit.
Are you planning to preserve some files? If so you may want to dump them onto some external storage before you remove the existing drive. When I looked a few years ago M key external enclosures were non-existent
@mcc of all the brands on the market, I trust Samsung the most. They have a good history of reliability and they actually make their own flash chips and controllers. All the others are randos buying chips and pcbs off the shelf and slapping 'em together and sticking their label on.
@mcc @schrotthaufen Windows always uses binary sizes (1 kB =1024 bytes, 1 MB = 1024 kB = 1048576 B, 1 GB = 1024 MB = 1073741824 B). Just about everybody else (including drive manufacturers) use decimal sizes ( 1 kB = 1000 B, 1 MB = 1000 kB = 1000000 B …), except for RAM sizes, which seem to be always in binary sizes (also, almost nobody uses Si binary prefixes – KiB, MiB, GiB …)
@mcc you'll probably be fine with the big name brands, samsung, wd. You can get into differences around the controller and all stuff like that, but outside of the difference between slc/mlc/tlc/qlc it'll probably be the same.
@mcc to clarify, sorry, the difference between slc and qlc mainly being speed and longevity, with qlc beeing cheapest, most dense, quickest to die, slc being most expensive, least dense, longest laating
@mcc @schrotthaufen https://en.wikipedia.org/wiki/Binary_prefix#kibi?wprov=sfla1 covers this pretty well.
@mcc as someone who worked in a family-owned computer store last summer, there's some differences. Samsung has definitely fallen off on quality control over the last 6 or so years. WD blue is usually what we stocked and sold, because it was usually the cheapest. We also sold crucial, because their reputation over the decades is a bit better. The speed and reliability differences are generally imperceptible between brands. The only thing I make sure people know is that once a drive starts to have trouble, you need to replace it as soon as possible, because you dont know whether it'll last 6 more months or completely fail and delete all your data later today. Drive failures were by far our biggest hardware issue amongst all categories of customers.
Tldr: dont buy samsung, otherwise it doesn't really matter, just check your drive health and replace as soon as there's any signs of potential failure.
@null_aleph yeah but uhh if they say KB does that actually mean KB or does it mean "KiB but we wrote KB"
@mcc I think that’s in the specs on the vendors site. I know Intel uses 1000. On skinflint they list both TB, and TiB/GiB in the specs. The iB units are the ones that are not misleading.
@mcc Although I guess now this iFixit thing might work? https://www.ifixit.com/products/external-m-2-ssd-nvme-enclosure
@mcc I mean if they put down KB, but meant KiB... That's on them?
@mcc @violator That won’t be a problem on any M.2 SSD. The kind of synthetic testing I’m talking about is filling up the drive and erasing it as quickly as you can, continually, for a year or more.
I write around 150 GB per week to my VM host (rebuilding VMs, mostly) and its drives haven’t spared any pages since the first month I had them. It’s been years.
@mcc individual models of drive from all the major brands will sometimes have problems. You probably won’t know for years. You can hedge your bets by buying a mix of brands. Backblaze publishes a report each year. https://www.backblaze.com/cloud-storage/resources/hard-drive-test-data
@mcc Of these, Intel (Solidigm), Kioxia (Toshiba), and SK Hynix make mostly datacenter drives. SK Hynix does make some OEM 2280 drives for laptop manufacturers, but they’re generally low capacity.
I’d go with Crucial (Micron’s consumer brand), Samsung, or Western Digital. Samsung was the brand to beat for a long time, but they’ve fallen lately.
WD’s drives perform well and you pay for it. Their fast drives need a heatsink (which laptops don’t generally have space for) and a lot of airflow.
Crucial’s drives have solidly middle-of-the-pack performance, but they don’t really need extra cooling. They’re also significantly cheaper than WD’s most of the time.
@bob_zim Wait if I'm looking at a WD¹ how do I know if it's one of the ones that needs extra stuff like a heatsink
^ "Black SN770" (??) they also have a SN850x but it might be sold out.
@mcc Mostly look at the box. Looks like the SN850x is offered with or without a heatsink. The SN770 just has a heat spreader label (no extra thickness).
In general, PCIe 4 drives put out an impressive amount of heat. They throttle themselves down when overheating, so they should never be significantly slower than a PCIe 3 drive.
@mcc Looking a bit more into this, I wouldn’t bother with an SN850x in a laptop. That one seems like it’s more a desktop drive. Pretty much all the drives with available heatsinks (and most drives marketed towards gamers) are meant to be fast at the expense of power and heat.
I wish SSD reviews would include actual measured power draw at idle, power consumed to write 1 GB, 10 GB, 100 GB, power consumed to read 1 GB, 10 GB, 100 GB, and similar figures.
@mcc I am not super fond of Canada computers; I have had issues with them here in Ottawa. Also stock issues and hostile web user interface.
I think the last NVMe drive I bought from them was a Corsair? I'll take a look; contrast with their listed stock and go off of the model#
What kind of usage case are we talking about? Are you expecting a LOT of read-write events?
Sorry for the wall of text.
Okay. So I think I have my plans for the hard drive complete. Now here's the shedpainty question:
The old drive has Ubuntu 24.04 on it. I hate it.
Should I trade down to Debian?
Or should I trade up to Pop!_OS?
Will I regret either of these? Will either one, if I just go get a standard usb key installation, cause driver problems with my AMD chipset or secure boot or whatever other junk Lenovo has on board?
@mcc Depends how restless you are. When I was young and restless, I ran #Slackware. Now I am middle aged and sedentary, I run #Debian. For a short interstitial period, I ran #Arch. I couldn't countenance Ub*ntu. They always want to foist some commercial spin on you, web search or snap or whatever, no doubt it will be some AI misery next...
@mcc Given how Ubuntu aggressively pushes Ubuntu Pro, even for packages that they don’t need to back port security patches for, as well as packages that weren’t in the Pro repos when the LTS version released, Debian seems like a good choice. You might have to disable secure boot for the installer to boot, though. The wiki is comprehensive, and I’d advise to read at least the chapter about DKMS https://wiki.debian.org/SecureBoot#DKMS_and_secure_boot
@mcc I've been running Pop!_OS for a year now on a laptop, it's like having an old Ubuntu LTS release but someone already setup the nvidia gunk for me. Kindof annoying from a dev POV because it's using a lot of old packages, probably wouldn't rec over Ubuntu unless someone else handling nvidia gunk is attractive for you
@kbm0 i went Bootleg PPC Redhat -> Bootleg PPC Debian -> Slackware -> Regular Debian -> Gentoo -> Ubuntu -> Kubuntu -> Fake Arch [msys2] -> Ubuntu. Now I am old, and I am just tired of bullshit
@mcc I do think of Debian, Arch and Slackware as the trifecta of honest Linux. Arch is the well meaning youngster, Debian the wise old matriarch and Slackware the idealistic old rebel who would never conform. They are each in their own way true to their principles. ✌️🙂
@mcc I jumped ship to Pop! OS after Ubuntu went to snaps. I absolutely love it.
Pop is a bit opinionated, but I got used to its defaults. (E.g. "super+B" to open the browser, rather than "super+2" to open the second item on my dashbar.)
One problem: Pop ditches grub for systemd boot, which gave me problems with a triple-boot setup. I would expect this to be the biggest possible source of friction. I ended up fixing it in my mobo's BIOS
@lynndotpy hm, I'm not overly fond of grub, it doesn't actually support my screen resolution well
@mcc what you appear to hate about Ubuntu is unity. I have Ubuntu + i3 and it doesn't annoy me at all.
@mcc Also: System76 is working on the next generation of their Cosmic DE, moving from a modded GNOME to a custom Rust-based DE.
This yields two potential sources of regret:
- The latest Pop! OS is based on Ubuntu 22.04, which means it ships with older software out of the gate.
- You might install Pop! OS, get comfy, only for the new COSMIC to drop.
Okay I have more computer build babby questions
I got a hard drive
But I've been warned it's one that runs hot
So I think I want a thermal "strip", which is apparently a heatsink that fits into smol spaces like a laptop
I google
https://www.amazon.ca/Deal4GO-Heatsink-5B40Z68852-Replacement-Thinkpad/dp/B0CDSBKD1X
This looks good! Oh, they're out of stock. Except wait, why doesit say "replacement"?
I watch installation instructions
https://youtu.be/8sm1ScVUHqY?t=108
Is there a hd heatsink strip in my friggin laptop already?? (1/2)
I only want to open up the laptop once. Trying to decide if I should
(a) just open it and assume there's already a heatstrip
(b) I poke around and there's lots of weird blue polymer strips that seem to do the same thing? It wouldn't be that expensive to just buy one and have it around if it turns out there's not one in there already…
(c) set the computer preemptively on fire, so that the hard drive can't be the one to overheat it
(2/2)
@mcc almost all laptops have an inbuilt thermal pad for the M.2.
what temperature it reporting? M.2 drives tend to run pretty hot anyway.
@gsuberland uh… now? the current one?
…how would i get that information in linux?
@mcc no idea on Linux.
I hadn't realised you meant "I'm told the new one runs hot". don't worry about it. the M.2 will work just fine, you don't need to mess with thermals. the onboard controllers are designed to operate up to about 120°C, and the flash ICs actually operate faster at elevated temperatures. people just get worried about it because of the intuition from CPUs running cooler being better.
@gsuberland Ah, OK.
I was told that if the drive starts overheating, it'll just throttle itself until it's cooler again. But I used to have an AMD GPU that would shut off the entire computer when it got too warm so I don't trust AMD and I don't trust Lenovo LOL
I guess if there are problems I'll live with them!!
@whitequark @gsuberland oddly, ubuntu does NOT recommend smartmontools when you type smartctl!
Apparently my current m.2 hd is at 32 celsius when nothing in particular is happening. I guess I'll see how it runs after I install the new one!
@mcc @gsuberland ah, good ol’ heat check exception shutdown
@c0dec0dec0de @gsuberland it was one of the dumbest interactions i'd had with a computer in my life, the video card had a fan but it didn't engage it, but i could manually turn the fan up. so i'd keep a heat monitor in the corner of the screen, and when it got too high i'd manually turn the fan high for a few seconds. this was during the semiconductor shortage so i couldn't buy a new video card and it went on for months
@mcc this is how everything works on modern laptops. the chassis is used as a heatsink and there's a thermal limit set based on the safe temperature before it burns your skin. so the CPU will thermal throttle below its theoretical maximum performance most of the time. usually shutting off means there's something super busted with the cooling solution (e.g. thermal paste dried out so the CPU exceeded max temp even with the fan on max)
@mcc Just wondering if you could elaborate on what you hate about Ubuntu 24.04 sometime.
I've been pretty happy with Ubuntu, although I'm running 22.04. But if there's something much better, I'd like to know about it.
@Jvmguy well, actually it's mostly better now. but 24.04 has repeatedly caused problems with one of
- my motherboard-- TWICE there has been a problem specific to amd computers where closing the laptop lid can cause the computer to permanently hang or shut down. both problems lasted over a month. the kernel version with the worst of these bugs never filtered back to 22.04
- all my gnome extensions keep breaking due to being on cutting edge gnome :(
@Jvmguy i assume lts is the stable branch but this has felt a lot more unstable than 23.04 which i was running on a different machine for a while. 24.04 set wayland as the default so maybe that was just so much work they didn't have time for other stability fixes? (also, wayland is pretty broken)
@mcc @whitequark oh that's well within safe bounds. even triple that would be fine under load.
@whitequark @mcc
Nvme works for me (solidgium/Intel drive)
https://www.cyberciti.biz/faq/linux-find-nvme-ssd-temperature-using-command-line/
@mcc while you're shuffling ssds around, you might want to get an enclosure. It'll make moving the data over easier, you can reuse the old drive as a flash stick, and you'll use it again for the next upgrade.
@mcc Are you aware of the need to disable the internal battery prior to opening?
@ieure yes, it's in the video i link above. but thank you anyway as it's very important
@mcc and one other, other consideration would be some blue loctite. You can get it in a bottle (which you have to poke with a needle when it clogs), or the stick, which is annoying to shmear on the tiny laptop screws.
I got sick of finding a random screw on the floor a month after opening my laptop.
@mcc I’ve been enjoying Pop!_OS, in both it’s released form and the new Cosmic Alpha that i’m playing with on an old iMac. Note it is based on Ubuntu, for better or worse.
Another worth a look is ElementaryOS. I had that on the old i!ac for a while and liked it. i changed over to the Cosmic Alpha before Elementary released their new version 8, but reviewers say they have fixed some things and made some nice improvements.
@mcc when you open and close back up a laptop, no matter how much you tighten the screws some of them always find a way to back off again.
Blue loctite glues the screw threads tight enough that you can undo it manually. (Red is way too strong.)
@mcc @rotopenguin it gunks the screw a bit, so you need slightly more force to unscrew it, it can't come loose by itself
good stuff, if you have it around, wouldn't get it just for this. rarely needed needed it, same for a screwdriver with specific torque
other tips all great, any single-sided 2280 ssd will work just fine, never regretted any samsung evo, Pop!_os great choice if you want to move away from Ubuntu, Debian often too limited - for good reason
good
@count @rotopenguin i have been having a lot of problems running the newest ubuntu because i get bleeding edge stuff that has bugs or incompatibilities. maybe if i can get a slightly outdated base OS and install new stuff via FlatPak that will work well for me.
@mcc @rotopenguin yeah my fun with the 24 series was limited as well. if you need stability, any of the LTS editions should be just fine, and the 22 are still supported for a while. you can also always pin apt releases while already listing and optionally installing new packages from newer ones, but usually will need to install a bunch of newer dependencies as well, then.
@count @rotopenguin yes, but also i'm tired of ubuntu for other reasons, mostly the fact that every time i file a bug they're like "you're running the weird ubuntu version— we can't act on this bug because we don't know if ubuntu and/or snap introduced the bug themselves"
@mcc Am I the only person who ever installs Xubuntu??
@gsuberland @mcc Yeah, the most irritating things with M.2 drives (particularly PCIe 4) are heat dissipation and power consumption. Drive reviews focus almost exclusively on desktop use, where both can be ignored entirely. Neither problem is *HUGE*, but they’re annoying if you care about using the laptop as a laptop.
@bob_zim @gsuberland Well, hopefully I didn't make a huge mistake battery life mistake picking this drive, cuz I got the good one :(
@mcc thermal conduction pads/strips are usually grey or blue (sometimes a pinkish colour, rarely yellow). They are semi-tacky and easily show fingerprints. Very very common in laptops on solid state drives and even on top of parts like the CPU and GPU and related components. If you can find out the thickness of the pads already install, it's better but as long as it's not crazy thick, you should be fine. :)
@mcc I never bothered with thermal pads on SN850X in my Framework laptop. Hasn't caused any problems yet.
@kbm0 @hisham_hm @mcc there are also Fedora and SuSE of you want a different brand of quite fresh corporate backed Linux distributions. IIRC fedora also has different, uh, flavours (IIRC they call it spins) that do different things, and Fedora Atomic is an interesting approach.
@gsuberland @mcc @whitequark 642 °C under load would still be fine?
(It’s a joke, because if you’re going to triple a temperature, you have to do it in a unit with a non-arbitrary zero, so I converted 32 °C to Kelvin, triple it, and converted it back, for the laughs.)
@oscherler @gsuberland @whitequark *clearing throat impatiently* He clearly meant M.6
@mcc That was exactly my reaction the last time I wanted to install an ubuntu and it recommended Balena Etcher
I'll stick with good old dd thanks
@nev It sounds extremely fake
Like ChatGPT tried to come up with a name for a computer program
Or if a network television writer had to write a one-off character who is an Artist and somewhat snooty
@mcc hehe, welcome
we have outdated software but there's a lot of it and it mostly works
@whitequark hm
i was hoping that if i ever wanted to run anything not-outdated i could just do it in a flatpak
how hard is it to upgrade stable to testing once i'm in? it's just i've been in ubuntu 24 for a year and it's constantly dropping me new bleeding edge versions of things and s t u f f k e e p s b r e a k i n g
@mcc it depends, sometimes you do want it running natively
upgrading to testing is ~trivial
@whitequark ok. actually i guess here's the question, are they running different kernels
@mcc i run a weird mix of stable+testing which works fine too
@mcc the kernel version is just whatever you install
nothing stops you from using stable kernel on testing, except systemd sometimes (but you should be fine)
Alright one last shedpainting question. Should I install Debian Stable or Debian Testing. Text replies welcome
@mcc my argument for stable is as follows:
- it is very stable, especially if you avoid pinning things or installing too many weird .deb files by hand
- if you need a newer/recent version of software it's fairly likely that it's not in testing either, so either way you'll end up installing it by hand
- any issues you do run across have very likely been seen/documented by someone else
(my read is that you do not enjoy Fucking Around With Linux which colors my recommendation.)
@d6 Yeah I uh there is a kind of Fucking Around I want to do and I need a stable base system so I can Fuck Around with the things I want to Fuck Around with instead of fucking around with like, I don't fucking know, my app switcher
@mcc Apocryphal answer incoming:
Both.
Stable is a whole distro, ensure it's installed.
Add testing sources as well.. it's not the whole distro, but what can't be resolved in testing will be resolved by stable.
Yes, once in a blue moon, you'll encounter a weird package issue. Blue moons, however, are very rare.
I'd say add unstable on top (it's the least complete), but that requires nuance, e.g. pin testing as the default, you do NOT want everything to come from unstable by default.
1/2
Pinning Debian to install things from testing by default is the safest way to run stable/testing(/unstable).
As a result, you get rolling-release-Debian. Since testing eventually becomes stable, you'll be ahead of the curve!
I pretty much landed on Debian years ago because of it's solid (or at least manageable) package management system.
I have *never* come across a Debian developer who wasn't horrified by these suggestions. YMMV. For me, It has been great from servers to game rigs.
@mousey How is this different from running testing, I don't think I understand the suggestion
@mcc right i would describe this as "i enjoy planned fucking around" and "i want to avoid unplanned fucking around"
@mcc Been running testing for a decade. I'm too impatient for stable.
Actually I also add unstable and sometimes experimental sources (particularly when there are major ongoing transitions and i'm impatient for everthing to flow from unstable->testing)
1/2
@mcc it's a lot easier to run Debian Stable and pin a couple of notable exceptions than to run Debian Testing full time, but the CI/CD testing has gotten so robust over the last ~7 years that even Testing is pretty stable these days.
@lambdageek Hm. Is it correct that it's easy enough to install stable first and upgrade to testing if I change my mind later?
Also, just checking, is Debian gonna expect me to choose a kernel version myself? The last time I installed it was, uh, *checks* the year 2000, off a floppy disk net installer
@mcc yea, just switch/add sources and `apt full-upgrade`
Not sure what you mean by "choose a kernel version myself". there's a 'linux-image-amd64' package that has a dependency on the latest kernel package. it will upgrade like everything else. Although you may need to manually remove linux-image-n.nn.nn-amd64 packages - I don't think old kernels will automatically uninstall themselves - if your /boot or /boot/EFI starts getting full you might need to remove some.
@mcc Stable tends to have somewhat outdated packages (often surprisingly so; I've seen some more than a year behind with missing backported patches for critical security issues) but there's less day to day hassle with updates breaking things.
Testing is much more up to date but generally works without too much hassle. If you don't keep on top of updates it does become progressively harder to update, but this is generally true for most non-LTS distros anyway.
I'd err towards Testing.
@mcc When I ran Testing over a decade ago, it was because I wanted to be able to get regular updates for software packages, rather than wait for a new Stable release every few years.
‘Testing’ gets a fair bit of testing before packages move into it from Unstable.
And moving from Stable to Testing on Debian meant changing the word ‘stable’ in /etc/apt.sources to ‘testing’. Then running ‘apt dist-upgrade’.
@mcc These modern days? Stable releases seem to come out faster. And I don’t know the state of Debian evolution. But I’d probably still run Testing, just for the continuous release of software updates.
And for all I know, there’s an even easier way to move from Stable to Testing. This is just my experience, not documentation.
@mcc Good luck with the move. Hearing about your problems with your Thinkpad on Ubuntu sounded terrible, so I hope Debian is better.
@beej I think if I really wanted to fix my problems I'd go for KDE rather than GNOME. But one change at a time…
@mcc one thing that does tick me off a bit about Stable is that I've only had a couple of cases where the distro release upgrade actually worked and left me with a functioning OS afterwards. maybe it got better on the desktop release in recent years but the last time I tried on Debian Server was just a year ago and it broke (I had to roll back my snapshot). I've tried it a bunch of times over the years and I'd estimate my success rate at about 15%.
@gsuberland Huh. So in that sense the "rolling" testing distro might actually be nicer because the upgrades come in more smoothly.
@mcc yeah, you run into more things misbehaving but it's rarely a fully system-breaking bug.
@gsuberland
> you run into more things misbehaving but it's rarely a fully system-breaking bug
I was using LMDE for a while, which is based on Debian Testing. Never had any noticeable issues, it was great. Never used Debian Testing directly though.
@gsuberland
> I've only had a couple of cases where the distro release upgrade actually worked and left me with a functioning OS afterwards
True! That's a mission stair I'm so used to, that my habit has been to keep my OS on a separate partition from /home. So I can just do a fresh install of the OS whenever there's a new version.
@strypey @gsuberland Yeah I did this on my last laptop and I *SHOULD* do it on this one but maybe I won't
@mcc @strypey so much stuff to reinstall, there's stuff I've built manually or had to modify or whatever, specific tools I won't remember the name of when I next need it (but can find through checking which I have installed), often config files are in /etc or other locations than /home so there's a bunch of reconfiguration that always needs doing.
@mcc when I used Debian I always chose Stable, for making it basically impossible to break OS with standard apt commands … but used almost exclusively linuxbrew (https://docs.brew.sh/Homebrew-on-Linux) for actual software that I use and was not necessary for the system to boot/run correctly
I've heard kids these days would probably use nix instead of brew :D
(1/2)
@gsuberland
> so much stuff to reinstall
Ae, it's a pain. Like a normal update, A GNU/Linux upgrade ought to be able to reliably carry over personal files *and* configuration, without a reinstall.
(2/2)
It's like we need a concept of an OS interface as a digital hotdesk.
Our desktop, our personal files + configuration (natives apps, browser profiles, customisations, etc) is stored on our own machine. Currently we only login to it from our own machine.
But what if we could plug a boot USB into any computer, anywhere in the world, boot into GNU/Linux, and login to our own desktop. Digital hotdesking!
GNU GUIX is working on a distro that would facilitate this, with only Free Code.
@gsuberland @mcc @strypey yeah I did it once then stopped for the same reason. It's not necessarily worse than not doing it, but it's not seamless
@catch56 @mcc @strypey my average lifetime for a Windows install is pushing a decade. prior to Win11 it was smooth sailing on major release upgrades too. once you're used to that level of low friction in UX it's sorta hard to accept how Linux treats this process* by comparison. although Win11 is just a total mess; the Win10->11 Workstation SKU upgrade path just doesn't work at all.
(*I am well aware of the tradeoffs please don't turn this into an exercise in proselytising for the gnulords)
@catch56 @mcc @strypey I am personally quite a fan of how FreeBSD does things because it's extremely upfront about the fact that if you don't keep up with updates, eventually the necessary repos for maintaining / upgrading your version will cease to exist and you're on your own. It feels very clean and simple. Of course the tradeoff is that nobody remembers FreeBSD exists so everything is written with Linux in mind and you just kinda have to hope it works for you.
@d6 @mcc e.g. if you buy a new laptop you likely want to be running the latest or near the latest kernel just so it actually like, works. There's usually some number of small functionality enablement that needs to happen. Then if you want to run new software (like e.g. steam games) you need graphics drivers that aren't from the middle ages. And any applications you want to use, you likely want the latest version of those too. (Debian stable has blender 3.4, for example, current upstream is 4.3)
@dotstdy @d6 I have a 1 year old laptop. The Ubuntu 24.04 kernel has had some bugs over the last year actually but now seems to be working, so if I can get this kernel in stable I'll take it. I have no intent of installing blender from .deb, that would be a Flatpak. The only piece of .deb software I want up to date is Firefox (but I'd be happy with Firefox ESR, honestly)
@mcc I run stable on my desktop, with the testing and unstable sources added in case I need them. I didn't need them in the last 2 years....
Flatpaks for every GUI app that's not in the Debian repo's.
What makes my system even more resilient is setting it up with btrfs snapshots and timeshift. Adding grub-btrfs and timeshift-autosnap-apt creates a system that creates a snapshot every time I install software.
@d6 @mcc flatpack sandboxes apps, so you trade these things kinda working on ye old distro for random stuff being annoying or broken inside the sandbox. /shrug but ime any notion of "stable" distros actually being stable for daily desktop use isn't really how it ends up working out unless you have very specific desires.
@mcc @d6 indeed that sounds like a snap problem that you'll likely also have in flatpack. I'd just be prepared for a different set of issues for flatpack v.s. snap :P (ultimately though, it'll all be fine enough either way, I'd just generally recommend the complete opposite, rolling distros being more stable for desktop use than stable distros are ime)
@mcc There is definitely some work at the time of setup: moving the existing btrfs root, creating a subvolume for your home folder, creating a few folders for mountpoints and editing fstab. After the Debian setup is done, you have to clone 2 git repo's and do a make install.
If you need a guide, I can put mine online later today (it's 8am here and life is happening).
@iceqbe I don't think I'm doing it this install but if you write that I'll read it!
@mcc Debian is towards the end of a release cycle (likely first half of next year) so if you’re thinking of running stable it might be worth considering installing testing now and keeping that release as it becomes stable - that’d save you an upgrade when the release comes out and sidesteps some of the issues with current stable possibly not having heard of newer hardware.
@mcc Do you want new features, bugs, and issues? Testing. Older versions and stability, stable!!
@jaj I've heard Fedora is really good now, yeah. I think I'm avoiding IBM rn but maybe I'm hurting myself by doing so.
@broonie *rubbing eyes* okay so… maybe i don't understand. if I install Testing am I Testing forever? Or if I install Testing do I get debian-13 and I stay on debian-13 until I run dist-upgrade?
@directhex @broonie okay, fantastic. and if i'm installing a new version of debian, i do this by installing testing and then editing my sources.list after install?
@mcc particularly this year, you don't want testing/unstable since they're in the middle of a very disruptive transition. Stable is pretty nice.
@eqe hm, but doesn't that mean i'll just get the disruption when it upgrades next year?
Okay so I just got the YouTube video up, aligned all the pieces I need to move my SSD into an enclosure and a new blank enclosure into the laptop, convinced myself I easily have time to do the entire operation and even get started on a Debian install before going out for NYE, and then immediately realized that it would be better to put this off until Thursday so I'm not thinking about setting up Debian all evening and then pulling out my laptop to tinker with Debian when I'm w/friends tomorrow.
@mcc I HAVE IMPORTANT INFORMATION REGARDING YOUR ENCLOSURE!
It uses a USB-C connector!
If you run a speed test on it and it seems way too slow:
FLIP THE USB-C PLUG OVER.
Putting on a big fluffy sweater and an anti-static band and then realizing these two probably cancel each other out and placing the big fluffy sweater a minimum safe distance away
OK I actually have a kinda dumb question. I've set up a nice clean space on a desk to do my minor laptop surgery. And I've put on my anti-static strip. And I don't know what to attach the strip *to*. My computer is a ways away and this table has metal feet.
My UPS, in reach, has an convenient exposed screw on the back labeled "TVSS GRND". Is this a good thing to ground or a thing that will specifically electrocute me if I electrically couple it to my body
As a piece of utterly unhelpful context, the power in our building is EXTREMELY dirty and although the wall socket has a hole for a ground pin, I'm not certain it's actually connected to anything
@mcc Plumbing is an option, too. Do you have a radiator nearby?
@darkling It's close enough that the band will reach but far enough it'll be a stretch :(
@waterbear I mean I could just go for it and see if i destroy my laptop and/or body
@mcc well rest assured you would only risk electrocution either way if you actually had a short in your ups. the question is whether it will ground you at all.
@gsuberland @mcc it's not a country with a sensible electrical system, there will be no ground in the mains connector by default unless they've upgraded to the grounded plug.
@gsuberland @dysfun @mcc It probably does have 3 pins, but that assumes that the socket on the wall is actually grounded. I've lived in places where the ground terminals in the socket simply didn't connect to anything.
@tojiro @dysfun @mcc the US electrical system is so cursed. I remember helping someone on here (pre-migration; probably back on the bad place) who lived in an SF apartment where the metal breaker box was floating live. they didn't have a multimeter so we did the "wire to a socket ground, resistor on a stick, poke the breaker panel" thing and sure enough the resistor blew up instantly.
In every type of modern electronics repair there is a step that's like "Yeah, just force it" and it's like I AM AFRAID TO JUST FORCE IT
@mcc It's always such a delicate balance of pressure, like which snaps first the part or my nerves?
OMG. I can't believe that YouTube kicked up this lovely Dutch fellow addressing the same, weirdly specific issue.
https://www.youtube.com/watch?v=-r9vo0lfpoU
Upshot: your radiator plan sounds like a plan that aligns with his thinking.
May the rest of your day be zap free.
The back of the laptop will not come off :( :( :( no matter what I do :(
There is a diagram on the Lenovo site. They didn't seem to think this was important to include int he video, the video was just like "use caution". It's also baffling. There are "latches" that have to be "pried up". How do you "pry" a "latch". What does that mean. Does it mean apply force. Latch 3D simply will not unlatch and I can see new little-but-distinct creases forming in the aluminum back of the unit
@mcc is there an ifixit guide for it? (that is, assuming that's not what the video screencap was)
@mcc In my 27 years of tinkering with computer insides, I never used an antistatic strip. If there was an outlet near, I'd touch the ground pin there (Schuko has ground pins exposed), but otherwise I just work on the insides as-is. Never had a problem.
@mcc Try pushing the bottom either towards the front, or the back – sometimes the latches will undo this way. Otherwise you just have to use more force – it'll unlatch this way.
@mcc Oh, and if you're just starting with prying the bottom off – just use force, no other way around it (the push advice is if you have 3 sides already unlatches, and one side doesn't seem to want to come off). If the laptop's a few years old, you probably will break a latch or two, but it won't cause a problem when you put the laptop back together, speaking from experience.
EVENTUALLY IT MADE A SNAPPY SOUND AND A TINY PIECE OF PLASTIC FLEW OFF AND IT OPENED. FUCK YOU LENOVO, I LITERALLY ONLY CHOSE YOU BECAUSE YOU WERE SUPPOSED TO BE THE BRAND GOOD FOR THIS KIND OF THING. Hopefully the back will go on again later! :(
@mcc and I feel that Thinkpads are some of the more serviceable notebook computers out there. The screws have retainers, most things snap, clip, or plug. But yes the hidden plastic case snappy clips are awful. (Start at the corner and go slowly with a nylon/plastic tool).
Makes me yearn for my undergrad days in the student lab -smoldering soldering iron, classic rock CDs playing in the background, green pci boards all over, a beer in one hand and a SIMM in the other. Wrist wrap? What's that?
@mcc Yup, that happens about half of the time when the laptop's a few years old. Don't worry, it'll still hold together, and you won't notice the part missing.
@mcc one hint when screwing the back back on (after snapping the case together):
On each screw to avoid crossing threads, slowly unscrew it with slight downward pressure up to a full turn until you feel/hear the click, then, without releasing downward pressure screw it forward to hand-tight. do not over tighten.
*Opening the M.2 packaging* Ooh, that's so nice of them! They included a free stick of gum!
"So this here is called a 'Zero Insertion Force' socket" "Okay, how do I use it?" "Apply moderate force"
@mcc Don't forget to remove the plastic if you'll use the thermal pad.
I can't :( it won't :( it won't go in :( a critical part of this plan was transferring my old M.2 to a USB enclosure. The instructions say only to insert it at 30 degrees and push it down. I can't figure out how it's supposed to go in. I can lightly lay it on top, but then it doesn't expose the screw hole to lock it in, and then I don't know if I did it wrong or if my drive is too tall somehow
@mcc Definitely not a dumb question. When working on a laptop, you should connect to the ground of the laptop. It’s fine if the laptop isn’t connected to anything, you’re just trying to be at the same electrical potential as the laptop itself.
@mcc Once it’s in the socket fully you should be able to see the pins, at least not to the extent that you can currently. Try giving it a little shove parallel to the socket
@h0m54r I feel like I can't tell if i'm putting it into the right hole, sort of. like the pins seem to be at an angle and i can't tell if i was supposed to lay them atop the pins or fit into the pins somehow
@mcc I was following a video describing how to remove a piece of trim from my car to get to a light bulb underneath it, and the professional demonstrating how to remove it made the exact same “This feels like I’m going to break it” cringe face I make when pulling trim. I took that as a good sign.
@h0m54r :O :O :O wait it snapped. it did a little snap and then it worked. thank you for giving me Confidence
@mcc You know how with a SO-DIMM, you insert it into the socket at an angle, then snap it down into the brackets to hold it? M.2 sockets work similarly.
Wait HECK I just realized. There wasn't a heatsink on the original M.2 like there was in the Lenovo video! And I already used my one "gum" heat strip on the USB enclosure!!!! :(
There were "replacement" Tinkpad HD heatsink strips on Amazon, I almost bought one. I guess I should have!! Crap what do I do now. I don't think Canada Computers had the "chewing gum" heatstrips.
The new M.2 has like, a long sticker on it. I think the stickers are supposed to be thermally active. Maybe good enough? :(
@mcc Getting the brain out is the easy part. The hard part is getting the brain *out*!
I guess what happens is my new HD doesn't spread out heat quite as well as it could have and either it gets hot inside my laptop or my HD doesn't run as fast as it could because it notices it's getting warm and autothrottles. Or I order the heatstrip after and go back in later to add one? Crap. I don't want to have to open this thing up again, like, am I gonna lose another little plastic bracket every time I open it? :(
Everyone I talked to was ambivalent about whether I needed the heatsink…
…you know, now that I've got it open… I wonder what's supposed to go here.
@mcc Don't worry about the cooling pad – as I mentioned earlier, I have 4TB WD SN850X in my laptop without any heatsinks at all. The drive might slow down a bit when writing a lot of data, but it really doesn't impact regular use.
@mcc Worst case, you can run the new drive with no thermal pad and it will throttle. It won’t break anything.
If possible, I would try to get the pad out of the external enclosure. Just go slowly and you should be able to move it to go under the new drive and conduct heat into the laptop’s logic board. Then you can get a second thermal pad and use it in the enclosure later (if you find you need one).
@mcc A LTE module most likely (might take an M.2 2242 SSD, too, depending on the slot key).
@mcc Generally a cell modem goes there. Slots like that typically have PCI-ID restrictions in the system firmware to only let the system boot with an approved cell modem installed or the slot empty. Without such restrictions, you might be able to install a second SSD there (looks like a 3042 slot to me), but they generally only run one PCIe lane to the cell modem slot.
@mcc M.2 4G/5G module. they're wider and shorter than a typical NVMe SSD hence the footprint.
@mcc It's a "zero insertion" Force Socket. You have to force it for zero insertion.
@mcc here's one of the fancy new ones that does 5G on the mmWave bands.
@mcc A number of things could go there, most likely a wifi card or a cellular modem.
The 850x will probably be fine for day-to-day use without a heatsink. You can probably induce throttling with extended speed tests, but I wouldn't worry about it.
Nothing on the inside is going to be sensitive information, it's all just uniform barcodes for IDing the type of SSD/motherboard/processor/et cetera. The only thing that's unique is your laptop serial number.
@doephin @mcc I don’t know. SO-DIMM retention clips are awful. Removing a stick feels scratchy and wrong, inserting a stick feels like it takes too much force.
The M.2 system is weird the first time you use it, but takes no real force at any point. Much nicer feel to use. Now if only we could get laptop makers to make their cases similarly smooth.
Of course, either is nicer to use than vertical DIMM slots. Those always make me feel like I’m going to wreck the motherboard.
Booted without an OS. As expected the Thinkpad firmware was just like "lol wat?". It can see the drive which is a good sign.
…What happens if I select "Lenovo Cloud"?
@mcc you'll be fine. NVMe SSDs aren't that picky about heat.
@mcc and yeah, sticker is metal foil and acts as a small heatspreader.
Debian installer just looks so charmingly goofy. All trying to be graphical but it never bothered asking the monitor its aspect ratio. "2025 is the year of Linux on this laptop"!!
Debian disk setup is asking if I want "LVM". I have never heard of that in my life. Will "LVM" make it harder or easier later to resize the partitions and maybe slap Windows 10 at the end?
@mcc Logical Volume Manager can make it easier to resize partitions and such after the fact iirc. The main reason I use it is so I just mount one big volume for all my shit, `/vg_root` or whatever, instead of having smaller separate volumes for /boot, /tmp, etc.
And… the Debian installer is successfully engaging with my new hard drive! That's success! I think! FUCK, this is literally the easiest type of user servicing you can do on this laptop and it was a huge stressful deal! Took me a week to psych myself up to do it and I think I have to go back for another pass later. The physical universe SUCKS. I want to go back to building imaginary castles in Rust.
Bikeshed color suggestions once again invited. Debian offered to let me set up with a separate /home partition, which yes please actually.
What do you think of these default sizes (see previous post)? Is 500MB enough for the EFI? Is 30GB *really* enough for /? How in tarnation is 1GB an adequate swap? Is that even a swap, at that point?
I'm inclined to go up to like, 128GB for / at least, but any insight on EFI or swap helpful.
Can modern Debian change these ratios later without data loss?
@mcc I have broken off soooo many of the clips over the years when upgrading company laptops… The screws are enough to keep it nicely closed, though.
@mcc do yourself a favor and don't touch x86 booting then
@multisn8 :O I was thinking about making a version of "Breakout" that runs in EFI…
@mcc 1GB swap seems excessively low, 500MB seems enough for the EFI, 30GB for /
seems low but it depends on what you want to do i guess
@Claire Well, what I want to do is install a bunch of gigantic FPGA IDEs in ~/usr, but also I guess I want to leave myself open to change my mind later about what it is I want to do…?
@mcc EFI: bump it up to a gig. It doesn't need much but it's annoying to expand later.
Swap: 8-16GB is my go-to range on modern systems, although you can go a bit higher on Linux because OOMkiller is indiscriminate in its carnage during heavy memory pressure. ignore anyone who advises disabling swap; it may stand between you losing work and not, and it also means that data which is required to be present in virtual memory but which is used once (or never) can be freed up for more file cache.
@gsuberland I've got 16GB of RAM in. I had swap disabled on Ubuntu and it went badly. I kind of don't want to be hitting swap and I don't mind OOMkilling but Ubuntu would do this thing where if memory ran out it would just lock up hard for 90 seconds.
Does 16GB of swap mean "you have 16gb of physical memory, then 16gb virtual-only memory on top of that?" This always confuses me.
@mcc harder.
It’s useful if you have multiple drives, or partitions spread all over, and wish you didn’t
But “shrink my main partition to make space on the drive” is particularly troublesome
@mcc let’s say you’re on a cloud server, and need more storage. You add a drive to the VM, you add the drive to your LVM volume, and you extend your partition onto it. Rinse and repeat. Easier than trying to spread your data around, or migrate onto larger drives
@directhex Hmmm actually this is a problem I have very specifically had on my cloud server (they upgrade my storage for free every few years…)
@mcc @gsuberland it means there is 16GB of disk the OS can swap memory too, which is almost the same thing.
@dysfun @gsuberland "Almost the same" as the second thing, you mean?
I know how to write a virtual memory system which doesn't help because it means i can imagine more than one way for the thing to work :(
@mcc I’d use at least 50GB for /, if you want to use docker at least 100GB.
You can resize partitions, and w/o cryptsetup, and LVM that’s not even too much of a hassle. Best to boot grml, and use gparted for that, though.
I wouldn’t create a swap partition at all. Using swap files is more flexible. https://www.digitalocean.com/community/tutorials/how-to-add-swap-space-on-debian-11
@mcc @gsuberland You should have a userspace oom daemon ("oomd"), which most distros preinstall today. This will take care of killing any thrashing programs long before your computer freezes.
The main purpose of swap in a modern system is to optimise memory by enabling unloading of rarely used data in favor of more frequently used data and for that you don't need that much, perhaps a few gigs at most. Some distros use memory compression for this purpose ("swap on zram"), idk if this one does
@mcc haven't seen the whole thread, this just popped up on my feed, but to expand on others' comments;
- 500MB is enough for EFI, but i'd suggest 1GB, especially if you want to install additional kernels
- 30GB for root is technically enough, but if you need anything slightly more than the base install, it's too low, so go for something like 100GB at least
- 1GB swap is probably a mistake. 8GB should be enough for most, but better to use the same amount as RAM if you want swap with hibernate
@mcc @gsuberland One trick is to keep a ‘nice -20’ console, serial, or ssh session as root always open. If the system starts swapping to death, it may still be available and responsive.
@mcc 30GB for / was enough for me in the pre-flatpak era, but strained after even my extremely limited usage of flatpak started becoming unavoidable. a year ago i had an excuse to shift some partitions and sized up to 50GB and that has been adequate.
@mcc essentially, yes.
simplifying a bit here: there's a lookup table that the kernel uses to map between virtual memory pages (the things you program sees as memory addresses) and physical pages (locations in system RAM, mostly). what the system can do is purposefully not put an entry in that page table for some virtual addresses. when the program tries to access such a virtual address, a page fault is raised, and a fault handler can manually implement loading/storing elsewhere (e.g. swap).
@mcc 500M EFI is more than enough (and trimming it down will make you more potential trouble than 200M is worth)
your "real" rootfs will probably not exceed 2G. maybe 4G if you install most of the archive. throw /var in there and 30G is enough for the next two decades of server-style use. I can't think of a reason to go to 128G, you're just wasting space
that said, I personally don't see how a laptop, of all things, would benefit from a split-/home at all
if you're hibernating you want at least your RAM size in swap. if not then swap is basically perfunctory
i would /not/ use a swap partition at all. if you use a swapfile, you'll be able to re-apportion the space as desired (this is the default configuration on ubuntu, and it works indistinguishably from a partition after setup). so if you decide that you do need bigger swap (smaller swap), you can just do it without repartitioning at all. this is the configuration I'm using on all my systems (that aren't running ZFS)
@mcc at any rate, you can, of course, always resize this without losing data. admittedly some configurations will be harder; in the photo you posted you can basically trade swap size for rootfs/homedir size at will, but growing the rootfs will be annoying (admittedly, there'll be no reason to do it because 31G is more than enough for anyone)
if you're looking for additional hot tips, you can enable noatime and/or nodiratime in the configuration for the ext4s in the partitioner; unless you're a known atime enthusiast this will net you lowered IOPS on your flash for no discernable difference
by default debian trims all filesystems once a week. unless you're really (like, inordinately) unlucky with your flash this is good, but in the case you /are/ unlucky and your filesystem stops fscking cleanly, this will be fstrim.service/fstrim.timer
@nabijaczleweli "but growing the rootfs will be annoying" wait though couldn't i just zap away the swap partition and grow rootfs to cover?
@mcc @jplebreton when you first tell flatpak about "the only repo you really need", throw in --user.
Oh, changing any of these defaults is actually a real pain. The Debian disk partitioner is not nearly as good as the Ubuntu one :(
I've got a gparted boot disk here but I don't want to back all the way out of the installer. Like 27 years of using Linux now and the partitioning is still a pain…
@mcc good question, i'm not sure. if there is i'd guess it'd be one of those Nonstandard Configurations that brings no end of trouble, but maybe an expert can politely chime in.
@mcc Can you ctrl+alt+F2 or whatever to switch to a virtual console, do the gparted, then go back fast enough in the installer to tell it to reread the partition table?
Me: Using GiB when setting up partitions like a good little girl
Debian installer: 17.2 GB
@mcc the sticker is thermally conductive.
Worst case, under heavy load, the drive will thermally throttle its own performance.
@directhex @jernej__s Yeah it just uh… it feels very silly, to have paid more for the fast hard drive, but the hard drive is too fast, so it gets hot, so it slows down. Like why did I do any of that.
Do you think if I just get one of these it will just fit? Like it says alienware but 2280 is 2280 as long as it's not thick right
https://www.amazon.ca/Fleshy-Leaf-Heatsink-Replacement-Alienware/dp/B0B66G3ZV6
Anyway I think I *AM* gonna buy another heat spreader thingy, it has a heat spreader sticker but it only covers part of the main chip, do you think this would be a sensible thing to get? Would I be better getting one of the blue gum spreaders? Is there a meaningful difference? Does literally anything matter?
https://www.amazon.ca/Fleshy-Leaf-Heatsink-Replacement-Alienware/dp/B0B66G3ZV6
@mcc honestly it'll be fine without. they sell the drives as-is with the thermal sticker and they'll perform fine like that.
@noiob @mcc yes. the flash actually performs slightly better at higher temperatures, within its overall rated temperature range, and generally those flash ICs won't self-heat to a level that matters. the controller IC can get a bit toasty (hence why they use the thermal spreader sticker on it) but they'll operate up to about 90°C before they start to throttle back.
@mcc
"Apply slowly increasing force until you hear a loud crack. Your unit is now open, or possibly your screen is sparkly confetti..."
@gsuberland @mcc I'm interested that you've explicitly proscribed disabling swap: I'm not trying to change your mind, but one reason I started doing it a couple years back is because I figured it was a whole bunch of write-wearing on my SSD that I really didn't need to have happen at all, with sufficient RAM. I've definitely occasionally noticed its absence (esp. since w/o it, I can't hibernate); but my solution so far has been "MOAR RAM" (next machine's coming w/ 64GB). Do you think there's some point (maybe 1TB 😛) at which there's enough RAM to abandon swap entirely?
@eigen @mcc the wear amount is low. keep in mind that most flash is rated for at least 0.5DWPD and more commonly at least 1DWPD longevity, which means many terabytes per day for the warranty period. the swap also writes in full blocks so it has ideal patterns for optimal wear levelling and avoiding write amplification.
@mcc don't use LVM for your purposes in this case, it'll cost you complexity and a tiny bit of performance but won't grant you any meaningful advantages.
@mcc I suspect the old "hit any key to continue" option might work here. 😄
@mcc you'll be able to resize /, /swap, and a separate /home if you want to do that. Resizing EFI is much harder. Give it a gig if you still can, but 500MB is plenty. (The extra space only matters if you want to keep several kernel versions bootable with large initrd images, possibly stuffed with lots of firmware blobs. Only hooligans do this.)
@mcc the blue gum strips are thermal interfacing, not spreaders themselves. Think of them as less goopy thermal paste, intended to transfer heat across a void from heat sources to an aluminum dissipation plate. Won't hurt anything to add a gum strip, but unless it the other side of the strip physically touches the laptop case lid it's not doing much good.
@mcc I wouldn't bother, unless you're planning to do a lot of writing (constant writing for 10+ minutes at maximum speed the SSD supports). Not worth bothering otherwise.
@mcc The colour of my bike shed is
everything on the disk is one partition. Swap is swap files on that partition.
@mcc @dysfun @gsuberland The first VMM I wrote, once I had it almost working, crashed during the first stress test. it load a workload page over its own code. I'd forgotten to mark the pages it occupied as frozen in physical memory.
@mcc 2 things; 1) you'll need LVM if you want full disk enc and B) I recommend installing windows before *nix OR on a separate drive altogether. It will trash your grub if installed on the same drive and you'll have to repair it afterwards.
@mothershrugger yeah, i've done the grub / win bootloader dance before *_* but thanks
@mcc Grub is... I really don't understand why everyone still uses Grub...
Honestly, I'm still sad that the Syslinux family never took off as a better alternative for all of this. I especially liked how I could have a dedicated partition for it and just edit the configuration on said partition instead of doing complex mounting and chroot just to repair a broken Grub install.
Not sure about the new Systemd bootloader yet.
@nazokiyoubinbou @mcc Same - I used to use syslinux for everything. It's just easy to use. Why anybody bothers with grub I still don't understand; maybe I'm just being parochial, but a boot loader that requires a full desktop app to configure seems... It doesn't make any sense to me.
As for systemfail's UEFI boot loader, I use it on Leandra and Windbringer and it's been solid. I have a hook for mkinitcpio to keep the active files updated and if I didn't have it in my wiki I'd forget about it. Might be the only good thing about systemfail.
@drwho @mcc I've had to repair Grub installations multiple times and I literally have to search it each time because there is no way I can remember the weird mount commands you have to do before chrooting to actually be able to reinstall Grub.
I don't even know how to repair a Systemd bootloader yet. Would definitely have to google it.
Sad thing is, syslinux is still technically relevant. It even has an EFI version. But you'd have to manually update the kernel and initramfs every single update I guess. Or someone more knowledgeable than myself would have to make a script to do it, but that's beyond me. I really wish Syslinux had taken off more. It just feels more advanced and easier in general.
@nazokiyoubinbou @mcc It depends on how you configure it. Windbringer and Leandra both use Arch-style boot image naming so for syslinux I just had to re-run the boot loader and that was it.
As for systemfail's EFI boot loader it's also set-and-forget, because that's how UEFI is supposed to work.
@drwho @mcc Hmm. You may be right. Some like Mint give a different filename with each update, but now that I look in /boot on my Arch install I see just plain filenames. In fact, they aren't even duplicates. It just straight up doesn't bother with the numbers. The old one is just "-fallback" it seems.
I'm not sure if I have the energy to deal with manually switching at this time, but... This seems kind of viable...
@nazokiyoubinbou @mcc I have writeups about using both, if you're interested. It's pretty straightforward.
@nazokiyoubinbou
Well if you want to use modern hardware enclaves / tpm or you know a hardware integrated standard it's what you're going to have to deal with irrespective at this point. Adding anything more on top of it is just layering more junk
@mcc
@jwp @mcc A thing isn't junk just because it's on top of something like that.
I recognize that we have to deal with UEFI. I didn't say we don't. But these things don't really replace it. They all work with it to provide more than what it does. Even Grub does more. (And yes, Grub can inherit the system resolution. That was just a bad default that Mcc ran into.)
@nazokiyoubinbou
If your uefi device tree equivalent doesn't explicitly include the panel attached to the output. Then adding another layer that needs to guess sane defaults is just adding potential failure domains. Syslinux/grub/winbootmgr/refind etc all need to make educated guess as to a sane set of Lowest common denominators. Unikernel approach makes sense to me rather than reimplementing another os effectively in a loader which adds more stuff that needs maintaining
@mcc
@jwp @mcc It's not reinventing a wheel if the wheel has been there for ages. It's using the wheel that works better instead of the one that is basically just a barebones emergency one.
We're not going to agree on this point. I prefer to use the better thing. I don't see this conversation going anywhere from here.
If your wheel is Syslinux then, it's a bicycle bike tyre you're attempting to fit to a car