[Aptitude-devel] Strange issue with aptitude causing 100% cpu usage on Sid amd64 DomU

Tue Jul 3 22:47:06 UTC 2012

Hi,

since about a day or such my Sid DomU (Xen virtual machine) gets
extremely sluggish as soon as I start "aptitude -u". It even throws
kernel messages like this several times during the (very slow) package
update:

Message from syslogd at kiva6 at Jul  4 00:02:00 ...
 kernel:[7736331.176014] BUG: soft lockup - CPU#0 stuck for 23s! [ksoftirqd/0:3]

aptitude (0.6.8) consumes 100% cpu according to htop and so do other
processes (irssi, screen, htop itself, ksoftirqd/0 and a few more),
load goes up to 4 to 7 and everything is very laggy. But as soon as
suspend aptitude with Ctrl-Z or stop it with Ctrl-C, everything is
fine again immediately.

If I first run apt-get update and then aptitude without -u, everything
seems fine for a moment, but then all mentioned processes go on 100%
cpu again as soon as the view is generated (maybe a little bit
earlier, too, but not as soon as the blue bars are shown on top and
bottom), funnily except aptitude itself. It stays at 5% to 10%.

So I wonder what's the real cause for that. The machine hasn't
rebooted recently, so it can't be the kernel. Swap usage doesn't
change the game either.

Linux kiva6 3.3.0-trunk-amd64 #1 SMP Thu Mar 22 18:02:10 UTC 2012 x86_64 GNU/Linux
 00:35:57 up 89 days, 13:32, 12 users,  load average: 0.10, 1.89, 2.53

(Ok, I could reboot into the 3.4 kernel which is currently in
experimental. But then again it worked more than 80 days without any
such issue. :-)

"which-pkg-broke aptitude" shows the following potential packages
which could have broken aptitude recently:

libdb5.1:amd64                                         Fri Jun 29 10:47:06 2012
libxapian22                                            Fri Jun 29 10:48:01 2012
libapt-pkg4.12:amd64                                   Fri Jun 29 17:57:34 2012
libc-bin                                               Sat Jun 30 13:08:29 2012
libc6:amd64                                            Sat Jun 30 13:08:42 2012
multiarch-support                                      Sat Jun 30 13:10:06 2012
debconf                                                Tue Jul  3 01:12:57 2012
dpkg                                                   Tue Jul  3 21:12:58 2012

It though IIRC worked fine on the freeze Saturday, so according to
this list, it could only be dpkg or debconf. But both are not running
when aptitude downloads package lists or is just idling around in
interactive mode.

Non-interactively ran it looked like this:

# aptitude upgrade
Resolving dependencies...
open: 1; closed: 0; defer: 0; conflict: 0                                                                                                                  OUnable to resolve dependencies for the upgrade: no solution found.
Unable to safely resolve dependencies, try running with --full-resolver.
# aptitude safe-upgrade
Resolving dependencies...
Unable to resolve dependencies for the upgrade: no solution found.
Unable to safely resolve dependencies, try running with --full-resolver.
# aptitude safe-upgrade --full-resolver
The following NEW packages will be installed:
  aria2{a}
The following packages will be upgraded:
  apt-dater apt-dater-host conkeror debootstrap kiwix
The following partially installed packages will be configured:
  bzr-git
The following packages are RECOMMENDED but will NOT be installed:
  xapian-tools
5 packages upgraded, 1 newly installed, 0 to remove and 2 not upgraded.
Need to get 6,126 kB of archives. After unpacking 4,060 kB will be used.
Do you want to continue? [Y/n/?]
Get: 1 http://ftp.ch.debian.org/debian/ sid/main apt-dater amd64 0.9.0-3 [83.2 kB]
Get: 2 http://ftp.ch.debian.org/debian/ sid/main apt-dater-host all 0.9.0-3 [18.4 kB]
Get: 3 http://noone.org/conkeror-nightly-debs/ sid/main conkeror all 1.0~~pre+git1207032207-~nightly1 [256 kB]
Get: 4 http://ftp.ch.debian.org/debian/ sid/main aria2 amd64 1.15.1-1 [1,453 kB]
Get: 5 http://ftp.ch.debian.org/debian/ sid/main kiwix amd64 0.9~beta6.8-1 [4,254 kB]
Get: 6 http://ftp.ch.debian.org/debian/ sid/main debootstrap all 1.0.42 [59.6 kB]
Fetched 6,126 kB in 0s (17.0 MB/s)
Preconfiguring packages ...
(Reading database ... 350759 files and directories currently installed.)
Preparing to replace apt-dater 0.9.0-2 (using .../apt-dater_0.9.0-3_amd64.deb) ...
Unpacking replacement apt-dater ...
Preparing to replace apt-dater-host 0.9.0-2 (using .../apt-dater-host_0.9.0-3_all.deb) ...
Unpacking replacement apt-dater-host ...
Selecting previously unselected package aria2.
Unpacking aria2 (from .../aria2_1.15.1-1_amd64.deb) ...
Preparing to replace kiwix 0.9~beta6.5-1 (using .../kiwix_0.9~beta6.8-1_amd64.deb) ...
Unpacking replacement kiwix ...
[...]

The system stayed snappy the whole time.

Anyone has an idea where else to look for the cause?
Anyone else experienced such issues?

		Regards, Axel
-- 
 ,''`.  |  Axel Beckert <abe at debian.org>, http://people.debian.org/~abe/
: :' :  |  Debian Developer, ftp.ch.debian.org Admin
`. `'   |  1024D: F067 EA27 26B9 C3FC 1486  202E C09E 1D89 9593 0EDE
  `-    |  4096R: 2517 B724 C5F6 CA99 5329  6E61 2FF9 CD59 6126 16B5