Bug#514443: upgrade-reports: etch->lenny upgrade: problems with dhcp, user-mode-linux, apcupsd, atftpd, x11-apps, cpufreq-info, nfs-common
Package: upgrade-reports
Severity: normal
>My previous release is:
Etch (including all updates as of today) + volatile + debian-multimedia.org
>I am upgrading to:
Lenny (as of today) + volatile + debian-multimedia.org
>Archive date:
Sat Feb 7 15:00:01 UTC 2009
Using dak v1
Running on host: ries.debian.org
>uname -a before upgrade:
(forgot to grab this, but it was a custom 2.6.28.2 compile)
>uname -a after upgrade:
Linux eternity 2.6.28.4 #1 SMP Sat Feb 7 15:40:47 CET 2009 i686 GNU/Linux
>Method:
* Logged in via ssh using putty from a windows xp laptop:
# dpkg --audit
(no output)
# vim /etc/apt/sources.list /etc/apt/sources.list.d/*
Changed 'etch' to 'lenny' in /etc/apt/sources.list
+ /etc/apt/sources.list.d/*
# apt-get clean
# aptitude update
# apt-get install apt
# aptitude install aptitude
# aptitude search "?false"
# aptitude safe-upgrade
# aptitude dist-upgrade
# (at the end, grabbed linux-2.6.28.4.tar.bz2 from kernel org, compiled,
rebooted)
>Contents of /etc/apt/sources.list:
#
# deb cdrom:[Debian GNU/Linux testing _Etch_ - Official Snapshot i386 Binary-1 (20060307)]/ etch main
#deb http://ftp.no.debian.org/debian/ etch main non-free contrib
#deb-src http://ftp.no.debian.org/debian/ etch main non-free contrib
deb http://ftp.se.debian.org/debian/ lenny main non-free contrib
deb-src http://ftp.se.debian.org/debian/ lenny main non-free contrib
#deb http://ftp.us.debian.org/debian/ etch main non-free contrib
#deb-src http://ftp.us.debian.org/debian/ etch main non-free contrib
deb http://security.debian.org/ lenny/updates main contrib non-free
deb-src http://security.debian.org/ lenny/updates main contrib non-free
deb http://volatile.debian.org/debian-volatile lenny/volatile main contrib non-free
deb http://volatile.debian.org/debian-volatile lenny/volatile-sloppy main contrib non-free
>Contents of /etc/apt/sources.list.d/debian-multimedia.sources.list:
deb http://www.debian-multimedia.org lenny main
>- Were there any non-Debian packages installed before the upgrade? If
> so, what were they?
Sun Java 1.6.0_11 installed into /usr/local/java/jdk1.6.0_11/
(installed from vendor .bin from java.sun.com)
Latencytop 0.4 installed into /usr/local/sbin/latencytop
(compiled from source)
>- Was the system pre-update a pure sarge system? If not, which packages
> were not from sarge?
The system was originally installed via a nightly etch debian-installer
from when etch was still in 'testing', and has tracked etch as
stable since then.
>- Did any packages fail to upgrade?
Initially, yes: apcupsd and apcupsd-doc failed during
"aptitude safe-upgrade" with:
Errors were encountered while processing:
/var/cache/apt/archives/apcupsd-doc_3.14.4-1_all.deb
E: Sub-process /usr/bin/dpkg returned an error code (1)
dpkg: error processing /var/cache/apt/archives/apcupsd-doc_3.14.4-1_all.deb (--unpack):
trying to overwrite `/usr/share/doc/apcupsd/examples/newslave.c.gz', which is also in package apcupsd
Apparently, this was caused by "aptitude safe-upgrade" deciding to hold
"apcupsd" during safe-upgrade, while still trying to upgrade "apcupsd-doc".
>- Were there any problems with the system after upgrading?
Yes, several:
* DHCP server (dhcp3-server):
The system being upgraded acted as a DHCP server for the windows laptop
client used to connect via ssh for the upgrade. The DHCP server was shut
down for an extended period of time during the upgrade (as aptitude
worked its way through the packages). After a while, this caused the
Windows XP DHCP client to drop its lease and switch to a 169.254.* IP
address without prior notice. From the Windows XP event log:
(Several warnings): "Your computer was not able to renew its
address from the network (from the DHCP server)..."
(Final error): "Your computer has lost the lease to its IP address
[redacted] on the Network Card with the network
address [redacted]"
(Warning): "Your computer has automatically configured the IP address for
the Network Card with network address [redacted]. The IP
address being used is 169.254.[redacted]"
Following this, the putty session was immediately disconnected and any
attempt to reconnect resulted in "no route to host". This happened in
the middle of aptitude running "Setting up package...."
To continue, I had to hook up a monitor and keyboard to the server,
and log in as root on the console. "ps ax" showed "aptitude" stuck in
"dpkg --configure -a" (iirc), so I killed the aptitude and dpkg process,
and had to restart aptitude / dpkg configure. Fortunately this seemed
to work out well; I completed the upgrade via the console.
Suggestion: Try to minimize the downtime for dhcpd during a dist upgrade.
(restart it sooner instead of waiting until all other unrelated packages
have been upgraded, if possible?)
* UML guests (user-mode-linux):
The system hosts a user-mode-linux guest (running debian etch).
(Before starting the upgrade on the host system, the guest was shut down
gracefully manually).
After upgrading to lenny's user-mode-linux package, the guest failed to
boot properly. This seems to be caused by a change in how "hostfs" mounts
are handled: Inside the guest, /lib/modules is configured to be mounted
via hosts from the host's /usr/lib/uml/modules. The old fstab entry
on the guest, which was
hostfs /lib/modules hostfs defaults,ro,/usr/lib/uml/modules 0 0
failed to run, yielding an error "mount: special device hostfs does not
exist". Correcting the fstab entry in the guest to:
none /lib/modules hostfs defaults 0 0
fixed the issue. (On the host, the uml process is started with:
screen -d -m linux.uml ubd0=/path/to/guest/root.img ubd1=/path/to/guest/swap.img mem=256M umid=guestname eth0=tuntap,tap0,,redacted hostfs=/usr/lib/uml/modules con=pty con0=fd:0,fd:1
)
Also, for some reason, I can no longer connect to the console of the uml
guest. Previously, I could use both "screen /dev/ttyp0" or "screen -r"
to connect. Now, using "screen /dev/ttyp0" gives a blank terminal, and
"screen -r" (as the user running linux.uml process) gives: "Cannot open
your terminal '/dev/pts/6' - please check."
Apparently there's been a change in how pseudo-terminals work? /dev/pts/6
appears to be owned by the uid owning the 'xterm'?
* cpufrequtils:
"/usr/bin/cpufreq-info" no longer prints a linebreak at the end of the
last line of each CPU's statistics, causing the text "analyzing CPU 1"
to appear at the same line as the last statusline for CPU 0. This broke
a simple script that used to parse the output via "cpufreq-info | grep 'CPU '"
* x11-apps:
xconsole uses a very weird, ugly, large font after upgrading (running
inside 'vncserver' / 'Xrealvnc'). I had to add this to ~/.Xdefaults to
get back the usual font:
XConsole*text*font: -*-fixed-bold-*-*-*-*-*-*-*-*-*-*-*
and add this to the middle of ~/.Xsession:
xrdb -merge "$HOME/.Xdefaults".
* atftpd:
During upgrade, this package threw a weird warning about an unrecognized
entry in /etc/inetd.conf:
"tftp dgram udp wait nobody /usr/sbin/tcpd /usr/sbin/in.tftpd --tftpd-timeout 300 --retry-timeout 5 --mcast-port 1758 --mcast-addr 239.239.239.0-255 --mcast-ttl 1 --maxthread 100 --verbose=5 /tftpboot"
while trying to add:
"tftp dgram udp4 wait nobody /usr/sbin/tcpd /usr/sbin/in.tftpd --tftpd-timeout 300 --retry-timeout 5 --mcast-port 1758 --mcast-addr 239.239.239.0-255 --mcast-ttl 1 --maxthread 100 --verbose=5 /tftpboot"
Had to manually edit this file and replace 'udp' with 'udp4'
* nfs-common:
This is actually from doing an etch->lenny upgrade inside the previously
mentioned UML guest: Trying to run "aptitude safe-upgrade" failed with:
Preparing to replace mount 2.12r-19etch1 (using .../mount_2.13.1.1-1_i386.deb) ...
You have NFS mount points currently mounted, and this version of mount
requires that nfs-common be upgraded before NFS mounts will work.
Aborting install.
dpkg: error processing /var/cache/apt/archives/mount_2.13.1.1-1_i386.deb (--unpack):
subprocess pre-installation script returned error exit status 1
Errors were encountered while processing:
/var/cache/apt/archives/mount_2.13.1.1-1_i386.deb
E: Sub-process /usr/bin/dpkg returned an error code (1)
A package failed to install. Trying to recover:
To work around this, I had to manually "apt-get install nfs-common" before
proceeding with "aptitude safe-upgrade"
* samba:
Windows XP client mappings stopped working. Previously, smb.conf used
"security = share" and the client drive mapping was configured with
the server's IP address (as in "\\1.2.3.4\sharename"). After a lot of
trial and failure, it would appear that the new samba package no longer
responded correctly to mounts based on the IP address, (yielding
'resource not found' type errors), so the mapping had to be
changed to "\\servername\sharename" on the client). (I also took the
opportunity to change over to "security = user", but I don't
think that's related)
> Further Comments/Problems:
Trying to run "aptitude upgrade" as specified in the release notes at
http://www.debian.org/releases/lenny/i386/release-notes/ch-upgrading.en.html#minimal-upgrade
resulted in a warning:
W: The "upgrade" command is deprecated; use "safe-upgrade" instead.
I used "safe-upgrade" throughout the process, perhaps the release notes
should be updated to prevent confusion.
Also, clamav was stuck for a while during upgrading as it tried to run
freshclam to download new signatures, this failed naturally as the
bind9 daemon was shutdown by aptitude. It'd be useful for both
named and dhcpd to become available earlier during the aptitude upgade run.
-- System Information:
Debian Release: 5.0
APT prefers testing
APT policy: (500, 'testing')
Architecture: i386 (i686)
Kernel: Linux 2.6.28.4 (SMP w/2 CPU cores)
Locale: LANG=en_US, LC_CTYPE=nb_NO.iso88591 (charmap=ISO-8859-1)
Shell: /bin/sh linked to /bin/bash
Reply to: