You are not logged in.
Since the past few months (probably around upgrade from systemd-204 in September) my system hangs for one, to a few minutes, and the CPU goes to 100% on almost every shutdown.
I've been reading about similar issues on the forum, like this: https://bbs.archlinux.org/viewtopic.php?id=170756&p=3
Things that seem to overlap with my case is that I'm not using any display manager, I log in directly to tty, and do a "startx" from there, and I shutdown from X using: "systemdctl poweroff"
So to see what's going on, I tested to first shutdown my window manager (i.e. go back to tty), and now my CPU goes to 100% (before I issue the shutdown command) and I can tun to top and see the culprit: systemd-journal. This happens every time.
My issue seems to be very similar to what is described here: https://bugs.archlinux.org/task/37007
Only that patch is applied into systemd-208, and I'm still having this issue.
Also I tried:
journalctl -b 0
-- Logs begin at sön 2013-08-25 22:51:23 CEST, end at tis 2013-12-03 07:11:09 CET. --
dec 03 07:11:09 x201 systemd[312]: Failed to open private bus connection: Failed to connect to socket /run/user/1000/dbus/user_bus_socket: No such file or directory
dec 03 07:11:09 x201 systemd[312]: Mounted /sys/kernel/config.
dec 03 07:11:09 x201 systemd[312]: Stopped target Sound Card.
dec 03 07:11:09 x201 systemd[312]: Starting Default.
dec 03 07:11:09 x201 systemd[312]: Reached target Default.
dec 03 07:11:09 x201 systemd[312]: Startup finished in 118ms.Last edited by penguin (2013-12-05 21:43:06)
Offline
That looks like a partition full problem where the persistent journal files are stored.
Offline
I tested to first shutdown my window manager (i.e. go back to tty), and now my CPU goes to 100% (before I issue the shutdown command) and I can tun to top and see the culprit: systemd-journal. This happens every time.
Perhaps a program is crashing or getting stuck in a loop and sending continuous error messages to a log file? I get this 100% CPU systemd-journal problem on X exit with emacsclient due to it spewing out a "back to top level" error message. The fix for that is to install emacs-lucid.
Last edited by rodyaj (2013-12-05 18:56:16)
Offline
@barbea, nope, only two partitions, both with plenty of room, but thank's for the tip.
@rodyaj spot on!
I used to have a self compiled emacs 24.2 with :
./configure --prefix=/usr --sysconfdir=/etc --libexecdir=/usr/lib \
--localstatedir=/var --with-x-toolkit=no --with-xft --without-tiff \
--without-toolkit-scroll-bars --without-gconf --without-gsettings \
--with-gif=nosince I'm not using the tool-bar, or sroll-bar, I skipp lucid as well, and emacsclient will survive the nasty gtk-bug.
I guess pacman replaced my version of emacs with emacs 24.3, with full gtk, thus the gtk bug again.
I recompiled, and it seems to work.
Thanks.
Offline
Glad to help. That bug is a real pain.
Offline
Just out of curiosity, how did you see/realize it was emacsclient that was drowning systemd-journal?
Offline
Sorry for late reply, but I'll post anyway as a reference. When I used a non-systemd OS, emacsclient itself had 100% CPU usage and when switching to another console I could see an infinite message loop of "...back up to top level". I just put two and two together when using Arch that this same bug could be flooding the logs of systemd and making systemd-journald take up all the CPU as reported in top. You could probably tweak /etc/systemd/journald.conf to prevent this sort of flooding from happening again, but I'm not exactly sure how.
Offline