Update to 1.8-0ubuntu1.2 in raring causes system hang at configure

[Stephan Frank]

Hi,

the recent upgrade to upstart 1.9.1 seems to have introduced a bug that causes a hang during the package configuration phase. The package has been extracted by the system updater and once configuration starts the systems freezes completely (mouse pointer disappears, no remote login possible, system does not seem to respond in any way).

Dpkg is now in a state where I am demanded to execute 'dpkg --configure -a', however since this freezes the system immediately again, I have no idea how or where to analyze the problem or how to proceed.

[Stephan Frank (sfrank)]

I have deleted the file 0000 in /var/lib/dpkg/updates/ and can at least use apt-get again. What this means for the state of the upstart package I do not know.

[Stephan Frank (sfrank)]

OK, one more clarification (sorry about that, but this thing is a bit aggravating since I have to reboot after every failed step due to the system freeze):

- the system is current raring
- failing upstart is 1.8-0ubuntu1.2 (not 1.9.1 as in the subject line!)

[Dimitri John Ledkov (xnox)]

Bugs against packages in Ubuntu, should be filed against Ubuntu distribution and correct package (upstart in this case). I refiled this report.

[Dimitri John Ledkov (xnox)]

Can you please attach:
/var/log/dpkg.log
/var/log/apt/term.log
/var/log/apt/history.log

such that we can investigate what has happened?

[Stephan Frank (sfrank)]

On 07/08/2013 12:38 PM, Dmitrijs Ledkovs wrote:
> Can you please attach:
> /var/log/dpkg.log
> /var/log/apt/term.log
> /var/log/apt/history.log
>
> such that we can investigate what has happened?
>

Please find the files attached. What you see there is that I tried two
things:

1. I tried to manually set upstart from half-configured to installed
manually by editing the status file to avoid any trouble when new
updates roll in. This didn't work and I reverted it.

2. I tried to downgrade to the last working version and this froze as well

Regs,
Stephan

[Steve Langasek (vorlon)]

Unfortunately, if the system is locking up completely, the logs afterwards are not likely to be very useful (and these don't seem to be).

To debug this, please try the following:

- run 'sudo telinit u' from the commandline, and let us know whether this triggers the failure. (This is the most likely cause of the failure.)
- if it does:
  - please send us the output of 'initctl version'
  - switch out to a text console (Ctrl+Alt+F1), log in, and run 'sudo telinit u' again. If there is any output on the screen, take a picture and attach it to this bug.

[Stephan Frank (sfrank)]

On 07/08/2013 09:46 PM, Steve Langasek wrote:
> Unfortunately, if the system is locking up completely, the logs
> afterwards are not likely to be very useful (and these don't seem to
> be).
>
> To debug this, please try the following:
>
> - run 'sudo telinit u' from the commandline, and let us know whether this triggers the failure. (This is the most likely cause of the failure.)
Does not freeze, but does not produce any output

> - if it does:
> - please send us the output of 'initctl version'
init (upstart 1.8)
> - switch out to a text console (Ctrl+Alt+F1), log in, and run 'sudo telinit u' again. If there is any output on the screen, take a picture and attach it to this bug.
Does not switch to the console (a problem with the FGLRX driver? I would
have sworn that worked recently).

Altogether no very helpful, my apologies.

Regs,
Stephan

[Steve Langasek (vorlon)]

oh, that's actually a very interesting result, because 'telinit u' is pretty much the only substantial thing done by the upstart maintainer script on upgrade. So if calling 'sudo telinit u' directly does *not* trigger the problem, then it seems we need to look deeper.

BTW, it appears that your logs not only don't capture the bits that we need, but are actually corrupted. What filesystem are you using?

The other thing I notice in the logs is that the corruption always happens immediately after the message about 'Processing triggers for hicolor-icon-theme ...'. So it's possible the upstart package itself is not what's causing this problem.

Can you reproduce the hang by running 'sudo dpkg --configure upstart' from the commandline?

[Stephan Frank (sfrank)]

On 07/08/2013 10:38 PM, Steve Langasek wrote:
> oh, that's actually a very interesting result, because 'telinit u' is
> pretty much the only substantial thing done by the upstart maintainer
> script on upgrade. So if calling 'sudo telinit u' directly does *not*
> trigger the problem, then it seems we need to look deeper.
>
> BTW, it appears that your logs not only don't capture the bits that we
> need, but are actually corrupted. What filesystem are you using?

file system is ext4

> The other thing I notice in the logs is that the corruption always
> happens immediately after the message about 'Processing triggers for
> hicolor-icon-theme ...'. So it's possible the upstart package itself is
> not what's causing this problem.
>
> Can you reproduce the hang by running 'sudo dpkg --configure upstart'
> from the commandline?
>

That went through; it was the pending downgrade from my previous tries.
After that the upgrade to the new version 1.8-0ubuntu1.2 went
successfully as well. So unfortunately, I can't reproduce the problem
any more.

Regs,
Stephan

[Steve Langasek (vorlon)]

Ok, thanks for following up. I'm closing this report as invalid, then, since it's unreproducible (for you or for us) and we have no explanation for the behavior.