Runaway init --user consuming CPU and RAM

[Michael Hall]

My Unity 8 froze on my Nexus 4. After connecting it to USB and adb shell, I found that 'init --user' was using 100% CPU and 60% RAM (> 1g). I ran "ubuntu-bug upstart" while it was still running to try and gather any information about what it was doing.

root@ubuntu-phablet:/# system-image-cli -i
current build number: 32
device name: mako
channel: trusty
last update: 2013-11-27 13:22:03
version version: 32
version ubuntu: 20131126.2
version device: 20131126

ProblemType: Bug
DistroRelease: Ubuntu 14.04
Package: upstart 1.11-0ubuntu1
Uname: Linux 3.4.0-3-mako armv7l
ApportVersion: 2.12.7-0ubuntu1
Architecture: armhf
Date: Wed Dec 4 14:08:43 2013
InstallationDate: Installed on 2013-11-26 (8 days ago)
InstallationMedia: Ubuntu Trusty Tahr (development branch) - armhf (20131126.2)
ProcEnviron:
 TERM=linux
 PATH=(custom, no user)
 SHELL=/bin/bash
SourcePackage: upstart
UpgradeStatus: No upgrade log present (probably fresh install)
UpstartBugCategory: Session
UpstartRunningSessionCount: 2
UpstartRunningSystemVersion: init (upstart 1.11)
upstart.tty1.override: manual
upstart.tty2.override: manual
upstart.tty3.override: manual
upstart.tty4.override: manual
upstart.tty5.override: manual
upstart.tty6.override: manual

[James Hunt (jamesodhunt)]

Hi Michael - thanks for raising this bug. However, none of the expected files that apport should have collected are attached. Added to which, I don't have a Nexus 4.

Do you have a crash file or any further details on what might have triggered this? Had the device been running for a long time? Had it just been updated via apt, etc.

Is this problem recreateable? If so, please grab an strace of the Session Init (re-nicing if required) and then maybe kill -ABRT to force a crash that apport will deal with.

The only detail I can glean from this bug which might be relevant is "UpstartRunningSessionCount: 2". I would have expected a count of 1.

[Michael Hall (mhall119)]

No crash report, since the process never actually crashed. This is at least the second time this has happened, so if it happens again I'll try kill -ABRT. The device had been running for maybe a full day, I hadn't done anything with apt since I'm using the system-image.

Can you provide instructions for what to do with strace? I know nothing about that.

[James Hunt (jamesodhunt)]

Hi Michael - if it happens again, please try to ping one of the upstart devs. Otherwise, the following would be very useful:

$ initctl list
$ ps -efwww
$ cat ~/.xsession-errors
$ initctl list-sessions
$ sudo cat /proc/$pid/stack

# let it run for a few seconds then kill it and attach the log.
$ strace -p $pid -fFv -s 1024 -o /tmp/session-init-strace.log

Then try the "kill -ABRT $pid". Thanks!

[Michael Hall (mhall119)]

I don't have strace installed, and I'm using the system-image so I can't apt-get install it. Bug I ran the rest and have attached the output.

[James Hunt (jamesodhunt)]

I haven't so far been able to convince gdb-multiarch to load that core file successfully.

If this is repeatable, please ping an Upstart dev on irc so we can debug this more interactively.

[James Lewis (james-fsck)]

I appear to have this problem, or one very similar on the desktop...

jlewis 4689 0.2 48.6 3133972 2917960 ? Ss Aug15 10:12 init --user

I'm not sure what it's doing, but it's using half my RAM and CPU... it appears to reliably leak memory over time, since I don't have a problem after login, but if I leave it a few days eventually the box grinds to a halt in an almost never ending swap grind.... I end up having to alt-f6 / kill Xorg or hard boot depending on how soon I catch it.

I've attached the output requested above, hope this helps some.... I have a number of 14.04 machines and this is the only one seeing this issue.

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Gareth Woolridge (moon127)]

Also seeing this on the desktop, as james-fsck machine grinds to a halt in swap after a few days.

I was forced to hard reset as I couldn't afford the time to try and get back into the system, if this happens again at a less critical time I'll try and collecty output per above.

[Gareth Woolridge (moon127)]

Disregard above, I only killed compiz and manually started unity to get the existing desktop back long enough to save some work so gathered the requested.

[Gareth Woolridge (moon127)]

Strace log