Re: [Qemu-devel] Apparent udev hang during reboot of guest which has been up for some time.

[Stefan Hajnoczi]

On Wed, Jan 23, 2013 at 11:08:59AM -0500, Oisin Feeley wrote:
> We have experienced symptoms very similar to those described[1] by
> Kenton Cabiness in 2011:  guests which have been up for some time
> (we're working on trying to find out exactly how long, but anecdotally
> it's about a week or more) will take up to several hours in the udev
> phase when rebooting.  However, freshly created instances of the exact
> same image on the exact same hardware will not display this hang.
[...]
> Guest:
> Speculatively: CentOS-5.8-x86_64
> Definitely based on some version of RHEL

Sounds like udev is the best starting point.  Here is a link on
increasing the udev log verbosity level:

http://www.novell.com/support/kb/doc.php?id=7000506

> Does anyone have suggestions about how we should go about bisecting
> this problem?

Bisecting is useful if the bug is a regression (it used to work but now
fails in a new version).  Do you have a known-good configuration where
this problem doesn't happen?

> dependent on how long the guest has been running and then stracing kvm
> for another instance which crashes.

Stracing kvm is unlikely to show anything useful since the problem is
inside the guest.  The strace output will probably be too low-level to
figure out what is happening inside the guest.

Stracing udev inside the guest might be interesting.  Not sure how easy
it is to do that in the CentOS 5.8 boot process.

Good luck,
Stefan