SystemD shutdown timer changes

General system Community contributions
1

Recently Fedora merged a change that reduces the shutdown timer for quicker shutdowns.

https://fedoraproject.org/wiki/Changes/Shorter_Shutdown_Timer

This is a pretty great change IMO but I wouldn’t think carrying patches against systemd to be sustainable for EOS.

I’m looking for some folks more familiar with lower levels of systemd and the possibility of integration into EOS through config files instead. It seems possible to do from my current research but I’m looking for opinions and maybe more knowledge on the matter.

1 Like
2

There’s a ticket for it on the Arch bugtracker.

1 Like
3

My shutdown on KDE is 4 secs. Doesn’t get any better unless you pull the plug. :smile:

1 Like
4

Nice, hopefully that gets merged

Edit: looks like they’re not interested in doing this change unless its upstream

5

Yeah but depending on your setup some services can end up needlessly delaying shutdown for 2min which considering the speed of modern storage is likely much longer than necessary for most cases before killing something that’s misbehaving

6

Yes i understand a lot of people have this issue it seems. It definitely would be extremely annoying. My shutdown and boot times are pretty fast. I also notice my dmesg is very clean on my Ryzen hardware. I guess i have very good UEFI firmware.

7

Arch Linux is looking into patching systemd instead of using config file. If they make a change then EnOS wouldn’t need to do anything. IMO EnOS should wait for a change in Arch Linux.

8

It doesnt look like they want to, they want it to get merged upstream. Upstream is dragging on this change and seems like so Fedora just did it on their own. The Pull Request for this change is over a year old so not likely upstream any time soon.

9

Looks like it’s easy, just need to create /etc/systemd/system.conf.d/override.conf and /etc/systemd/user.conf.d/override.conf containing :

[Manager]
DefaultTimeoutStopSec=20s
TimeoutStopFailureMode=abort

TimeoutStopFailureMode is in case a unit is faulty, you will find its coredump in the journal for debugging.

The discussion on Fedora mailing list is interesting :
https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org/thread/N6TW6EBPG5Q6D6SEV4FFGQH3E6HUTVE2/

1 Like