Not exactly as funny meme as I would like it to be, but I just found out about that feature after having to hold the power button due to a frozen system countless times, and I had to tell someone.
You must log in or # to comment.
Related and IMO a much better option for Linux desktops:
This is great! Why doesnt distros use this by default???
Just put plasmashell and a few in there and you will have a working oom killer. Finally.
I will install this the first thing tomorrow
Wait… Fedora has this since quite a while, strange.
Because if i’m rendering on blender on my lower end PC with expected freezes but it auto kills the render?
Sounds like niche use cases
Not niche there can be times when you want to run something heavy and it auto kills the exact thing you are trying to run. You have a 1gb ram device and it kills everything? Thats undesirable
Hm… the process itself should not take that much RAM. I dont know if normally the OS should assign the max RAM to the program.
But this should not happen and I wonder how “just letting it freeze” works
It makes system unresponsive, true. But its still running the main things, the render or decompressing or whatever. So it eventually unfreezes when it completes, by giving other programs(including GUI) back the CPU and ram.
Ok so killing is worse than just keeping alive.
This is a fair point.
I dont know a good solution for this, not killing but freezing is likely the best.
I dont know
Your usual pal won’t be running Blender, they’re going to be stumbling their way through LibreOffice and a browser. Massive echo chamber right there.
deleted by creator
We do not break userspace in this household young man.
Why doesnt distros use this by default???
Nohang has some explanations to this.
I.e. kernel devs are ignorant to the issue of oomkiller not working as intended on desktop.
Edit: Lkml is down.
deleted by creator
Is this better than oomd?
Wasn’t oomd the facebook thing for complicated server setups?
edit: yeah, for large data centers. Imho overengineered for single user desktop sessions. Earlyoom is simple and tiny.
Just decrease your swap space.
Unless you have an unusual system, there’s no reason to have several GB of swap.
that won’t solve the system unresponsiveness
Have you tried?
Because it does.
yes, even turning swap off entirely doesn’t solve it. It doesn’t take much to find people reporting a similar experience.
Better enable swap again. Linux expects swap.
or fiddle with the vm/swappiness value
Is hibernate no longer a thing? I thought that needed swap.
Actually, not much.
It always had reliability issues with bad hardware, and computers boot incredibly quickly nowadays. But yeah, it requires swap, and if you want it, there’s a sibling answer here about sawppiness.
ok… I’ll ask… where the heck is the “sysreq” key on my standard keyboard?
Should be the screenshot key
so it is! and I tested it, of course, with alt-sys-b which instantly rebooted my machine. nice.
Prt sc one? Print screen?
Yeah, on my keyboard it’s just an icon so I forgot the actual name lol
How come sysreq + f is not on by default? After discovering and enabling it I haven’t had to hard restart due to hangs or crashes.
Debian has it by default I think. Arch has it disabled because it might be a security risk if someone had physical access to your computer.
those debian daredevils like the thrill of living on the edge
Raising Skinny Elephants Is Utterly Boring … Or so I’ve heard.
That restarts the system. This only attempts to kill the app that uses most memory.
