Ubuntu Developers Have An Idea For Handling The Over-Eager Systemd OOMD App Killing

[timrichardson]

systemd oom has one job really: avoid the kernel oom killer waking up. If the apps causing out of memory are isolated from systemd killing, doesn't this just return us to the original problem, the invocation of the ponderous kernel oom killer?

[oleid]

Just buy more RAM. It is cheap. A modern 2022 desktop should have 32GBs of RAM at the minimum.

Well, my development machine from 2016 has 32 GiB and occasionally hits the limit because CLion likes eating RAM for breakfast and if you open multiple projects with large codebases oomd hits eventually.

And this is a good thing. Otherwise the system would end up in a crawl and the machine would become unusable for minutes without reboot. That's why I configured the daemon in the first place.

[oleid]

oomd killing apps without actually solving the problem with underlying heuristics/algorithm.

If there is actually a problem with the algorithm or heuristics. Looks logical to me. Kill the process that keep requesting more memory = browser.

[openminded]

It works fine on 4gb ram on fedora. May be the Ubuntu is doing some funny things in background?

"Ah, snap!" (c) Ubuntu devs, probably

[arun54321]

It works fine on 4gb ram on fedora. May be the Ubuntu is doing some funny things in background?

[Jabberwocky]

I'm glad I didn't install a distro using this new "feature". I typically run Slack, Brave, Firefox, Chromium, Spotify, Discord and others on my host.

I can't wait for Facebook's new memory swapping daemon to be activated on desktop systems by some crazy distro. /s

Seems to me a solution to a non-problem. I mean memory is cheap and plentiful. If you experience out-of-memory problems, your machine isn't configured properly for the use case. Don't get it .

My 2700X has 2 memory controllers. I loose "precious XMP/AMP timing" or system stability when if I go 4 DIMMs route. The idea is to stay as close to Infinity Fabric speeds as possible (3733Mhz for me) and drop latency as low as you can (CL14 for me). Low latency 2x32GB modules are pricey if you can find stock. The other option is Threadripper... well... that's considerably more $$$. I would love to go zen3 just for the improved UMC but I need two+ CCXs for my VMs. That means I need a 5900X or better and again that's expensive.

I'm running multiple VMs including Windows gaming. Yesterday games used 8GB RAM, today some are using 12GB, and tomorrow games are using 16GB. Things change overtime.

[hax0r]

so solution they are after is to insert "avoid" directive into userspace systemd daemons files to workaround (or should we say, 'hack around') oomd killing apps without actually solving the problem with underlying heuristics/algorithm.

[AndyChow]

Pfff, n00bs. Just get a faster internet connection and download more RAM.

[middy]

Seems to me a solution to a non-problem. I mean memory is cheap and plentiful. If you experience out-of-memory problems, your machine isn't configured properly for the use case. Don't get it .

ah, the let them eat tesla argument of 2022. "just buy more ram bro, what's the problem? what are you, poor? LOL"
if it was so easy for people to obtain ram, we all would have ungodly amounts of ram and wouldn't need such nonsensical things like swap. what's cheap for you doesn't mean is cheap for someone else. that's also ignoring the fact not everyone's machine can even be upgraded to high levels of ram too. and no, telling them to buy newer machines isn't going to help. if they could afford it, they would.

[TemplarGR]

Just buy more RAM. It is cheap. A modern 2022 desktop should have 32GBs of RAM at the minimum.