|
Lifroc posted:Haven't noticed dropped frame with Zoom myself. Be aware that there was a major Zoom release literally a couple days ago that improved Wayland support, such as screensharing I hear is working fine now. Worth giving it a try. I don't have weird screenshare or remote desktop requirements, so I've been running Fedora w/ Wayland full time. The laggy and tearing feel of X.org was annoying 5 years ago, I can't stand it now. While I'd like to run Wayland, the flickering in Citrix makes it unusable - it's not just a flicker of existing content, it's bringing in just removed content in a flickering mess over the window. Tough to describe, but very distracting in practice. There's another Zoom / Wayland / NVidia driver issue I forgot to mention: you get blank, transparent windows which show nothing, including the main zoom window. Found a workaround where you set enableAlphaBuffer=true in ~/.config/zoomus.conf - but that still leaves some menu windows showing all black. Fake edit: https://www.reddit.com/r/archlinux/comments/pk1e1y/zoom_display_issue_under_wayland/ describes the same Wayland / Zoom issue and points to a fix PR that still isn't in Zoom v5.11.0. Disappointing for a PR that's been open from Feb '21. Will build it for myself if I can figure out a workaround for Wayland/Citrix.
|
#
?
Jun 27, 2022 18:26
|
|
|
|
| # ? Jun 9, 2024 05:46 |
|
|
Klyith posted:For really easy tryouts, use Ventoy. Instead of having to reformat the usb like rufus, you just drop multiple isos on the stick and it makes an automatic menu for them. It doesn't do persistent storage or other elaborations like a rufus stick, but that doesn't matter in this context.  Where has this been all my life?
|
|
#
?
Jun 27, 2022 18:43
|
|
|
Klyith posted:Run 3 or 4 live boots from a usb with different desktop environments and see which one feels good. Any time you're looking for something in particular, it makes a lot of sense to try before you buy. awesome, a replacement for YUMI, which hasn't been updated in over a decade afaict
|
|
#
?
Jun 27, 2022 20:42
|
|
|
I noticed one could get Red Hat Enterprise Linux for free with a developer account. Is this an option for Desktop if you want a no-frills distro you only have to update once a decade? I mean with snap and flatpack nowadays the apps I care about will be up-to-date regardless of the distro.
|
|
#
?
Jun 28, 2022 08:08
|
|
|
busalover posted:I noticed one could get Red Hat Enterprise Linux for free with a developer account. Is this an option for Desktop if you want a no-frills distro you only have to update once a decade? Sure, but unless you want access to the official support forums, you might as well not bother with an account at all and just install Rocky Linux.
|
|
#
?
Jun 28, 2022 10:09
|
|
|
NihilCredo posted:Sure, but unless you want access to the official support forums, you might as well not bother with an account at all and just install Rocky Linux. Yeah, I've looked at Alma and Rocky, but one reviewer mentioned that it's not sure they'll still be around years from now. vOv
|
|
#
?
Jun 28, 2022 11:25
|
|
|
busalover posted:Is this an option for Desktop if you want a no-frills distro you only have to update once a decade?
|
|
#
?
Jun 28, 2022 14:31
|
|
|
ExcessBLarg! posted:Debian stable? yep, i've been running debian stable on my desktop since my windows install shat itself again in september and i'm never going back there's regular security updates that take about 5 seconds to install and only really require reboots maybe once per month, but no major version changes until debian 12 releases, probably sometime in 2023 i briefly considered centos/rocky but ehhhh, i'm more at home in debian
|
|
#
?
Jun 28, 2022 14:54
|
|
|
busalover posted:Yeah, I've looked at Alma and Rocky, but one reviewer mentioned that it's not sure they'll still be around years from now. vOv Both the Alma and Rocky projects have tried to structure themselves to avoid the Ac&Ex problem that killed Centos. Pablo Bluth fucked around with this message at 15:45 on Jun 28, 2022 |
|
#
?
Jun 28, 2022 15:40
|
|
|
busalover posted:I noticed one could get Red Hat Enterprise Linux for free with a developer account. Is this an option for Desktop if you want a no-frills distro you only have to update once a decade? AFAIK you can install RHEL on 16 machines for free. Of course with no support, but you didn't have support on CentOS either. Alma and Rocky are cool, but being rebuilds you never have 100% confidence they are identical to the real thing. If I need more than 16 licenses for a rock stable distro, it means I'm doing something serious and I'd pay the fee to have a red hat-wearing engineer on call, personally. But 99% of times, any other regular distro is fine. I've worked with multi million dollar businesses running on Ubuntu Server with no support. These days I just need docker/podman to run, the host is not that important. EDIT: to answer your question, flatpak and snap I bet require decently recent kernels to work, by virtue of being new and rapidly evolving technologies. In 9 years I wouldn't be sure flatpaked GNOME 50 will support kernel 5.18. Lifroc fucked around with this message at 17:00 on Jun 28, 2022 |
|
#
?
Jun 28, 2022 16:56
|
|
|
a dingus posted:Thanks all for the GPU help. Im looking through the logs and all I can find that looks suspicious is this snippet. I hooked up a new display which I didnt even think about... hooked it up at the same time as my new GPU. So maybe theres something going on between the two of them. Just following up on this to say thanks, people. I upgrade to fedora 36 and whatever that did, be it upgraded kernel or replaced some corrupted stuff or whatever, the problem seems to be solved.
|
|
#
?
Jun 29, 2022 01:57
|
|
|
a dingus posted:Just following up on this to say thanks, people. I upgrade to fedora 36 and whatever that did, be it upgraded kernel or replaced some corrupted stuff or whatever, the problem seems to be solved. Linux people get really defensive about AMD, but I find that even though their drivers are open source, the driver itself is shoddy and full of bugs. My 5500 XT was unusable until 2 years after launch, my 6800 XT has been pretty great but after spending some time figuring out why some games kept stuttering, and it's because of the automatic power management doing a lovely job at switching frequencies, so even a demanding game can have the GPU frequency rapidly going from max to 350 MHz. The workaround is to tell the driver to just put everything in performance mode. Both issues I had with the cards have been reported upstream, the bugs have been there for years, unresolved. The whole drm/amd repo has 1200+ open bugs, many open for half a decade. People love to share the "NVIDIA, gently caress you!" quip from Torvalds, they're lagging far behind on their user-space part of the driver, but what's there works better than AMD and performance is pretty much identical to Windows. Also, it's paradoxically easier to virtualize a consumer NVIDIA GPU, through firmware hacks/libvfio, than it is to virtualize a consumer AMD GPU. My prediction is either Intel actually delivers with their dedicated GPUs and eats AMD's lunch, or NVIDIA now that has an open source driver eventually will eat AMD's lunch on Linux as well. It's like Firefox, I use them because their product works better on Linux, but I'd like to chuck both in the bin and buy better products if I could.
|
|
#
?
Jun 29, 2022 09:09
|
|
|
It's sucks that both are broken to a point. I don't do any heavy gaming on Linux and my upgrade woes are the worst I've run into for AMD on Linux... I've had many more problems on windows with AMD tbh. Nvidia worked well other than when I tried to run an unsupported kernel. The linux crowd is small so I get it. We're not a huge priority.
|
|
#
?
Jun 29, 2022 14:29
|
|
|
NihilCredo posted:Sure, but unless you want access to the official support forums, you might as well not bother with an account at all and just install Rocky Linux.  Rocky or Stream is probably the best best
|
|
#
?
Jun 29, 2022 15:34
|
|
|
Steam Deck though. I mean, that's also a small thing, but Valve has enough clout to force AMD to fix their poo poo.
|
|
#
?
Jun 29, 2022 15:34
|
|
|
ExcessBLarg! posted:Steam Deck though. I mean, that's also a small thing, but Valve has enough clout to force AMD to fix their poo poo. Valve has enough clout to just pay someone else to fix AMD's poo poo, see the RADV driver being the most used one, and sponsored in part by Valve, while AMD keeps releasing their open source AMDVLK driver which no one uses except enterprise.
|
|
#
?
Jun 29, 2022 16:45
|
|
|
Lifroc posted:Linux people get really defensive about AMD, but I find that even though their drivers are open source, the driver itself is shoddy and full of bugs. I don't think it's defensive, I think it's a different assessment of what bugs are important and which are eh, I can live. Stutters in a game are, in the big picture, not that important. The reason AMD gets universally recommended to people asking "which video card should I get for linux desktop and gaming" is because a person asking that question is probably not an expert. For the inexperienced, the potential problems from nvidia are a crisis: your PC doesn't boot, or only boots the fallback option, or whatever. This really isn't even an nvidia bug! It's a drawback to the whole linux system. But it's a major deal and ruins their day. An expert user doesn't find it insoluble or even super-difficult (usually), but they're probably not asking on a forum which GPU to buy. And if the question was "which should I get for GPU compute on linux" it would be a whole different context. Lifroc posted:My 5500 XT was unusable until 2 years after launch They had problems on the windows side too -- not for 2 years, but late 2019 - early 2020 I could get some pretty consistent bluescreens on my 5700 from using wattman to set power or fan curves. Something about RDNA1 was mildly busted, possibly more in hardware than drivers.
|
|
#
?
Jun 29, 2022 17:53
|
|
|
I've noticed that Linux is simultaneously more aware of hardware level bug, but also more forgiving about being able to work around them
|
|
#
?
Jun 29, 2022 18:20
|
|
|
Since around 2008 my main machines have all been AMD CPUs with Nvidia GPUs, so I might just be used to it at this point.
|
|
#
?
Jun 29, 2022 19:08
|
|
|
Goons, I'll never be able to run Linux as my main desktop OS. Not only do I not know poo poo about really using it, I also have to consider the needs of my wife and stepson who need the computer working as expected. They need to boot straight into Windows and just have the familiar login screen and everything in its right place, such as Office. I can't dick around with everything when that's the case. So I'm relegated to faffing about with Linux distros in VMs, which I'm OK with. Zorin OS seems pretty nice but I've noticed its kernel is a year old (and behind the Ubuntu 22.04 release upon which it's based). Just wondering - is there a way to update the kernel? I know there's no point in this use case but it's just to play about with stuff without ruining my household's life.
|
|
#
?
Jun 29, 2022 21:36
|
|
|
WattsvilleBlues posted:So I'm relegated to faffing about with Linux distros in VMs, which I'm OK with. Zorin OS seems pretty nice but I've noticed its kernel is a year old (and behind the Ubuntu 22.04 release upon which it's based). Just wondering - is there a way to update the kernel? I know there's no point in this use case but it's just to play about with stuff without ruining my household's life. I run ZorinOS 16.1 on my “faffing about” 2011-era laptop. I’ve had zero troubles with it: main use is bed-gaming (old games like Contra collection/Castlevania collection from Steam or Baldur’s Gate/Planescape stuff from GoG) and nighttime web browsing (w/Chrome!). I have set up LAMPS stuff in case I want to attempt to make it a home server, but I haven’t quite put the  in to learn it correctly yet. I think the new kernel was in 16.1, but could be mistaken ofc.
|
|
#
?
Jun 29, 2022 23:12
|
|
|
Klyith posted:I don't think it's defensive, I think it's a different assessment of what bugs are important and which are eh, I can live. Stutters in a game are, in the big picture, not that important. I'm a moron and bought a bleeding edge AMD SOC to run linux on and now this is my life periodically: Jun 29 18:57:22 fedora kernel: [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx_0.0.0 timeout, signaled seq=45856, emitted seq=45858 Jun 29 18:57:22 fedora kernel: [drm:amdgpu_dm_atomic_commit_tail [amdgpu]] *ERROR* Waiting for fences timed out! Went to Fedora KDE since kernel >= 5.17 enables GPU recovery for this GPU but as it turns out, recovery fails all the time!
|
|
#
?
Jun 30, 2022 00:19
|
|
|
yeahhhh laptops pretty often crash on linux right out of the gate because either "brand new SOC" or (usually more likely unless it's a "linux compatible" laptop) "OEM does some hosed up proprietary poo poo" if it's the former it gets fixed quick, if it's the latter better hope someone who knows their way around the kernel has seen your model around or you might have to wait a couple weeks lol
|
|
#
?
Jun 30, 2022 00:27
|
|
|
Still can't get my pg_dump process to use more than 72 GB of RAM (128 GB total). My boss was worried that we pay $texas for the RAM we can't use  Tried: - oom_score_adj - tinkering with postgresql.conf memory values - numactl --interleave=all Trying to find information how to get Linux to use the whole memory for process is just pain in the butt (cloud) as everyone's just being worried that caches eat all their RAMs. Time to bug our DBAs if they can come up with another solution. Kivi fucked around with this message at 12:51 on Jun 30, 2022 |
|
#
?
Jun 30, 2022 12:46
|
|
|
WattsvilleBlues posted:Goons, I'll never be able to run Linux as my main desktop OS. Not only do I not know poo poo about really using it, I also have to consider the needs of my wife and stepson who need the computer working as expected. They need to boot straight into Windows and just have the familiar login screen and everything in its right place, such as Office. I can't dick around with everything when that's the case. Buy more than one computer
|
|
#
?
Jun 30, 2022 12:48
|
|
|
Kivi posted:Still can't get my pg_dump process to use more than 72 GB of RAM (128 GB total). My boss was worried that we pay $texas for the RAM we can't use Show us some logs of the OOM killer in action.
|
|
#
?
Jun 30, 2022 13:00
|
|
|
WattsvilleBlues posted:Goons, I'll never be able to run Linux as my main desktop OS.
These days Chromebooks have native support for running Linux applications. This is done through a custom VM/container setup called Crostini. From a user perspective when you enable Linux support, you get a Terminal application that opens to a Debian command line, and from there you can install whatever applications, run them, and they just show up as if they were native. They really worked on getting the plumbing right here so it feels better than running a separate Linux desktop in a VM, even though it is using a VM for isolation under the hood. You can also enable Developer Mode on a Chromebook which basically takes the gloves off and lets you mess around with the OS however you want, though you could (not permanently) break things. I still use Developer Mode since occasionally I need to do things like run dumpcap to take a packet capture which you can't do in Crostini. Still, pretty flexible and if you don't poke around too much then auto-updates still work as expected. I'm not really sure how this compares to a modern Windows experience with WSL. My wife has a Windows 8(ish) laptop and her biggest complaint is that forced updates take forever to install (unlike Chrome OS which does it behind the scenes until a final 15 second reboot). Never used Windows 10/11.
|
|
#
?
Jun 30, 2022 16:48
|
|
|
Kivi posted:Still can't get my pg_dump process to use more than 72 GB of RAM (128 GB total). My boss was worried that we pay $texas for the RAM we can't use You’ve lost me. Why are you wanting this? Trust the Linux memory manager. Your disk is probably slower than your ram so that’s all it wants and needs to write it to a file. Check your io. jaegerx fucked around with this message at 00:48 on Jul 1, 2022 |
|
#
?
Jul 1, 2022 00:44
|
|
|
jaegerx posted:Trust the Linux memory manager. To reinforce this and also for a discussion on why using all of your memory for the process can be a bad idea, see this preso: https://www.youtube.com/watch?v=QZZWAsBI_zY One of the points he makes is that those caches are very much a part of the working set of your process and in some cases are more important than your in-process pages. So trust the memory manager and let it use it for cache. Came across that preso after reading this paper by him: In defence of swap: common misconceptions, And that paper in turn is linked at the bottom of the systemd-oom killer man page. The really cool point he makes in there is not merely that swap isn't to be considered emergency memory but rather it is what provides backing store for anonymous pages (like data pages in a db for eg) and allows the kernel to treat all memory as reclaimable not just fs cache pages that are backed by disk. So then all anon pages aren't locked into memory forever and the kernel can swap out anon pages simply because the cache pages are more accessed right now and better retained in memory. It's a good read and watch.
|
|
#
?
Jul 1, 2022 01:28
|
|
|
smarter people way above us are writing the linux memory manager. do not fight it.
|
|
#
?
Jul 1, 2022 03:03
|
|
|
Lifroc posted:Show us some logs of the OOM killer in action. jaegerx posted:smarter people way above us are writing the linux memory manager. do not fight it. The swap doesn't get touched. I could try increasing it, but why add swap when there's perfectly good memory lying as "caches". I have 128 gigs of RAM, 56 cores and several TBs of disk. Having a single process use all of it once in the box's lifetime is not big ask. Kivi fucked around with this message at 06:30 on Jul 1, 2022 |
|
#
?
Jul 1, 2022 06:24
|
|
|
Kivi posted:That's the thing. I can't. Nothing in the logs. Just pg_dump dying of "out of memory" as last line. thats not oomkiller killing it, thats it deciding its out of memory and dying all by itself. it typically gives you a HINT what can fix it
|
|
#
?
Jul 1, 2022 06:29
|
|
|
Kivi posted:That's the thing. I can't. Nothing in the logs. Just pg_dump dying of "out of memory" as its last line. That makes it sound like it is an internal limitation of pg_dump. You probably want to ask on their specialized mailinglist/discord/forum/bugtracker.
|
|
#
?
Jul 1, 2022 07:25
|
|
|
RFC2324 posted:I've noticed that Linux is simultaneously more aware of hardware level bug, but also more forgiving about being able to work around them This has been my favourite thing about Linux since migrating earlier this year. I'm still a huge Linux noob, and I only understand 10% of what is discussed in this thread. But I can easily use my computer just like when I was running windows, except it no longer randomly breaks on its own due to secret updates, and absolutely any problem I run into can be fixed. If I want to I can just rip out and replace huge chunks of the operating system. As if that wasn't enough, it's all free!
|
|
#
?
Jul 1, 2022 09:18
|
|
|
One really handy thing that was added to FreeBSD not too long ago is distinct log entries for why a process gets killed by the OOM killer.
|
|
#
?
Jul 1, 2022 09:33
|
|
|
Kivi posted:I'm not sure just having 60 GB of RAM in available and caches is smart. Sure, it's not free per se, but it should be available for use by just dumping it. firstly, on my postgresql boxes ram usage never goes beyond about 4 GB, and even those are big outliers that'd probably be better served by some other db, and i'm just being stubborn about consistency. most stay below 1-2GB and have the vast majority of ram filled by linux cache, with only 2-4gb actually free ram that stays free because it's regularly used by complex queries or other temporary processes postgres works as well as it does primarily by efficiently abusing the linux cache system. this is reinforced in postgresql docs, telling you to use fairly low numbers for most memory size settings and increase the effective cache size value outside very specific workloads, to let the postgresql query planner decide what's what and lean on system cache as much as possible as for your actual issue, i'm 99% sure you're hitting the pg_dump hard memory limit of 1gb for data somewhere in your dump, which can cause problems with deceptively small fields that have binary data in them. pg_dump tries to convert it into text (escaped hex iirc, it's been a while) for the dump file, hitting you with an out of memory error real quick because ascii text is much less space efficient than binary data, i think a 200mb bytea field is already too large you can temporarily work around this by using pg_dumpbinary, though i'm not sure if that has its own pitfalls somewhere, as the docs insist you only use it if pg_dump fails. i haven't had issues with it personally, but i've also complained to our devs when pg_dump failed until they fixed the fields in the db, so i've only ever used it very temporarily incidentally, how have you been backing up your database so far?
|
|
#
?
Jul 1, 2022 12:31
|
|
|
Resident set sizes can range from 1GB to 10TB, so comparing is complete nonsense when you're not going into workload specific details. If the OOM killer is killing processes just for using memory, the resident set size is either bigger than the amount of memory physically in the machine, or there's something wrong with the OOM killer.
|
|
#
?
Jul 1, 2022 12:37
|
|
|
BlankSystemDaemon posted:One really handy thing that was added to FreeBSD not too long ago is distinct log entries for why a process gets killed by the OOM killer. Linux oomkiller absolutely logs what it killed and why. The problem is the why part is 2-3 pages of extremely arcane and difficult to read text. I'd find an example, but
|
|
#
?
Jul 1, 2022 16:29
|
|
|
So, now that I'm pretty satisfied that I can use desktop linux as my main OS, I am converting my storage & data over to linux-native from ntfs. Question time: 1) How do people generally like to mount supplemental drives or volumes? The file manager auto-mounts stuff in /run/media/username, but if I'm putting stuff in fstab should that be in /mnt or what? I didn't go with LVM for my first install, so I'm just doing manual storage management. I dunno how useful LVM would be for me anyways: I have one each NVMe, SSD, and HDD (plus a smaller SSD for windows) and so spanning across 3 very different drive speeds seems questionable. 2) Usernames, groups, and IDs: are they as simple as they look? Particularly when it comes to user & group ownership of files. Say my username is Sexcopter & UID is 1001. Sometime in the future I move a HDD to a new machine which also has Sexcopter:x:1001 in passwd. Will the folders and files owned by Sexcopter still be owned by the new Sexcopter on the new install? (I ask because this was a thing that was super-annoying in Windows, your true ID was a huge GUID so if you reinstalled or whatever you had to reset ownership everywhere.) 3) Are there any other common gotchas or noob mistakes with user/group ownership? I've read "don't chmod +777 stuff" which I can get behind, but is there for example a problem with assigning ownership of a folder to nobody?
|
|
#
?
Jul 2, 2022 03:30
|
|
|
|
| # ? Jun 9, 2024 05:46 |
|
|
There's a bunch of standards/suggestions out there for how to mount disks but it really doesn't matter, just do whatever makes sense to you. I guess the only suggestion that might have some technical merit is don't mount remote disks on /, make it a subdirectory instead (like /mnt/butts instead of /butts). My toplevel is usually /srv for extra disks that will be serving data (either via nfs or a daemon) and /mnt/whatever for general storage. Yes, UID's and GID's are that simple. You won't be able chown files/folders to another UID as an unprivileged account. If you want to limit who can access a file/folder it's totally fine to do that (but it's a root activity to set it up). There's nothing noteworthy about the nobody account, it's just a UID/GID set up with no shell and no home directory. It's a common ownership to assign when no one but root should be able to read it, but you don't want it actually owned by root (which can give it elevated privileges).
|
|
#
?
Jul 2, 2022 04:35
|
|
