Server upgrade from 7.04 to 7.10 fails

[Andreas Gustafsson]

I just tried to upgrade a Ubuntu Server 7.04 installation to 7.10
using the procedure documented in "Network upgrade for Ubuntu
servers (recommended)" at
<https://help.ubuntu.com/community/GutsyUpgrades>.

That is, I ran

  sudo aptitude install update-manager-core
  sudo do-release-upgrade

At the package configuration stage, there was a large number of
error messages which I am unable to reproduce in their entirety as
I ran the upgrade from the console, and don't know any way of
saving the contents of the console scrollback buffer short of manually
transcribing them. The last few messages, manually transcribed,
were:

dpkg: dependency problems prevent configuration of grub:
 grub depends on volumeid; however:
  Package volumeid is not configured yet.
dpkg: error processing grub (--configure):
 dependency problems - leaving unconfigured
dpkg: too many errors, stopping
dpkg: ../../src/packages.c:251: process_queue: Assertion `!queuelen' failed.
#

Before those, there were similar messages about "dependency
problems" in other packages.

I now appear to be left with a half upgraded system and no obvious way
to recover.

A similar incident was reported by another user in
<http://ubuntuforums.org/archive/index.php/t-651880.html>.

[Iulian Udrea (iulian)]

Thank you for your bug report. This bug did not have a package associated with it, I have classified this bug as a bug in update-manager. Could you please add the log files from /var/log/dist-upgrade/ to this bug report as attachments?

Thank you.

[Andreas Gustafsson (gson)]

> Could you please add the log files from /var/log/dist-upgrade/ to this bug report as attachments?

I will as soon as I get the machine to boot again.

[Andreas Gustafsson (gson)]

Here are the log files from /var/log/dist-upgrade as requested. I have omitted
the file apt.log which is of zero length.

I should also note that the installation I attempted to upgrade was running
under Xen as the dom0.

[Michael Vogt (mvo)]

Thanks for the logfiles.

The logs say that something in the packages postinst script fails with a segmentation fault. If you still can access the system, could you please try running dpkg -i on the modules-init-tools package in /var/cache/apt/archives ? Does that work? Or does it fail in the same way?

[Andreas Gustafsson (gson)]

Michale Vogt wrote:
> The logs say that something in the packages postinst script fails with a segmentation fault. If you still can access the system, could you please try running dpkg -i on the modules-init-tools package in /var/cache/apt/archives ?

As a result of the failed upgrade, the system is not in a bootable state. I can mount the disk on a working system and examine its contents that way (which is how I obtained the log files), but I am unable to run commands on the system itself.

[Michael Vogt (mvo)]

Thanks for this update. So e.g. chroot /mnt/the_failed_upgrade_root does not work and bash can not be run? What is the error message if you attempt it? Does that segfault too?

[Andreas Gustafsson (gson)]

> So e.g. chroot /mnt/the_failed_upgrade_root does not work and bash
> can not be run?

Though the system doesn't boot, running bash chrooted to it worked.
Using that method, I tried what you suggested in an earlier message:

> could you please try running dpkg -i on the modules-init-tools
> package in /var/cache/apt/archives

I assume you mean module-init-tools. There were two such packages:

  -r--r--r-- 1 root root 86570 2007-08-20 16:21 module-init-tools_3.3-pre3-1ubuntu7_i386.deb
  -rw-r--r-- 1 root root 87414 2007-10-05 20:04 module-init-tools_3.3-pre4-2ubuntu4_i386.deb

I chose the latter one. This is what happened:

  root@guesthouse:/var/cache/apt/archives# dpkg -i module-init-tools_3.3-pre4-2ubuntu4_i386.deb
  (Reading database ... 36359 files and directories currently installed.)
  Preparing to replace module-init-tools 3.3-pre4-2ubuntu4 (using module-init-tools_3.3-pre4-2ubuntu4_i386.deb) ...
  Unpacking replacement module-init-tools ...
  Setting up module-init-tools (3.3-pre4-2ubuntu4) ...
  Segmentation fault (core dumped)
  dpkg: error processing module-init-tools (--install):
   subprocess post-installation script returned error exit status 139
  Errors were encountered while processing:
   module-init-tools

Rerunning dpkg under "strace -f" revealed that the core dump is from
/usr/sbin/update-rc.d. I located the core file in /etc/rc0.d/core.

Running "/usr/sbin/update-rc.d module-init-tools start 15 S ." by hand
also causes a core dump.

I will attach the part of the strace -f output pertaining to the
/usr/sbin/update-rc.d process.

Since /usr/sbin/update-rc.d is a perl script, I note that the
installed version of perl is 5.8.8-7ubuntu3.1 according to
dpkg-query --show.

[Giulio Fidente (gfidente)]

having the same identical segmentation fault here ... upgrading 7.04 server to 7.10

[Nathaniel Eliot (temujin9)]

Identical segmentation fault: this is probably a xen-related bug. Workaround is to move /lib/tls to /lib/tls~, solution is probably to install libc6-xen (or move it up in the upgrade process to before module-init-tools.

Solution better documented here:
http://strugglers.net/~andy/blog/2007/09/02/if-your-ubuntu-gutsy-chrootdomu-segfaults-weirdly-under-xen/#comment-53417

[Rolf Leggewie (r0lf)]

This ticket is tagged feisty2gutsy, an upgrade path that is no longer supported. Is there still anything left that needs to be done in a later release?

[Andreas Gustafsson (gson)]

> Is there still anything left that needs to be done in a later release?

I don't know; I haven't attempted any further upgrades, and I'm no longer running Ubuntu under Xen.

[Rolf Leggewie (r0lf)]

Thank you for the quick response. This looks like a bug in either Xen or module-init-tools. Reassigning to those packages for further triage.

[Rolf Leggewie (r0lf)]

resetting to unconfirmed state since it is not clear which of the two packages is/was the affected one.

[Scott James Remnant (Canonical) (canonical-scott)]

Neither 7.04 nor 7.10 are supported anymore