[systemd-devel] Randomly on shutdown, stop timeout for user at .service (repeated report, different user)

[Lennart Poettering]

On Mon, 30.09.13 18:55, Toms Seisums (toms.seisums at gmail.com) wrote:

> It appears that history repeats itself.
> 
> Once (July 2013) a similar issue was reported here, by different user:
> http://lists.freedesktop.org/archives/systemd-devel/2013-July/012283.html
> 
> I'm also using Arch Linux, systemd-207, linux-3.11.2.
> 
> On shutdown and reboot, user at 0.service occasionally freezes my system until
> it times out and gets killed.
> 
> Steps to reproduce:
> 1) Boot
> 2) Login
> 3) Execute either systemctl reboot or reboot or shutdown -h now or whatever.
> 
> I've went through the steps here, to get a proper shutdown log:
> http://freedesktop.org/wiki/Software/systemd/Debugging/#shutdowncompleteseventually
> 
> And, the shutdown log can be found here: http://pastebin.com/wbr04AQw
> 
> The actual lines related to issue:
> 
> #2223 [   74.387130] systemd[1]: Got D-Bus request:
> org.freedesktop.DBus.Local.Disconnected() on /org/freedesktop/DBus/Local
> #2224 [  163.814925] systemd[1]: user at 0.service stopping timed out. Killing.
> #2225 [  163.815447] systemd[1]: user at 0.service changed stop-sigterm ->
> stop-sigkill
> #2226 [  163.816063] systemd[1]: Received SIGCHLD from PID 326 (systemd).
> #2227 [  163.816093] systemd[1]: Got SIGCHLD for process 326 (systemd)
> #2228 [  163.816153] systemd[1]: Child 326 died (code=killed, status=9/KILL)
> #2229 [  163.816158] systemd[1]: Child 326 belongs to user at 0.service
> #2230 [  163.816171] systemd[1]: user at 0.service: main process exited,
> code=killed, status=9/KILL
> #2231 [  163.816183] systemd[1]: user at 0.service changed stop-sigkill ->
> failed
> 
> I currently haven't found out a pattern for this. I have attempted to quit
> some processes before initiating the shutdown/reboot, but the result is
> random.
> 
> About 95% of shutdowns/reboots it will be a long one.
> 3% it will be noticeably longer than the quickest one, but shorter than the
> above one.
> Remaining 2%, the shutdown will be very quick.

Hmm, this is the systemd user instance for root which fails to shut
down. Can you reproduce this if you start/stop "user at 0.service" quickly
in a loop?

Lennart

-- 
Lennart Poettering - Red Hat, Inc.