Debian Bug report logs - #439763
debconf: hangs on puppet installs on preseed install

Reply or subscribe to this bug.

Display info messages

Message #5 received at [email protected] (full text, mbox, reply):

From: [email protected] (Adrian Bridgett)

To: [email protected]

Subject: debconf: hangs on puppet installs on preseed install

Date: Mon, 27 Aug 2007 10:42:41 +0100

[Message part 1 (text/plain, inline)]

Package: debconf
Version: 1.5.14

(Mostly cloned from #425397 at the request of Joey Hess.  I've been
running puppet in the d-i/late-command part of the install to get these
problems, if it helps I'll try and trim it down (I suspect a preseeded
postfix install might be sufficient))

I'm using puppet at the end of a preseeded install and some packages
keep hanging (sysstat and postfix namely).  I suspect it's the
preseeding that's causing the problem.  

What I see is dpkg-preconfigure running /tmp/postfix.config.12345
(sometimes with a zombie'd process in the middle).  dpkg-preconfigure
is reading from FD8 (which is FD1 from the postfix.config),  the
postfix-config in turn is reading from FD0 (which is FD7 from the
dpkg-preconfigure).  Deadlock ensues :-(

I've been trying to do some debugging but it's a bit tricky as I'm not
sure exactly which frontends etc end up being run.  puppet is calling
apt which is set to run dpkg-preconfigure, overall therefore, the
config file thus gets run once by dpkg-preconfigure and once normally.

When postfix.config runs /usr/share/debconf/confmodules, at line 45 it
complains that FD3 is a bad file descriptor.  Listing /proc/$$/fd
shows that FD3 doesn't exist.  FD2 points to the overall stderr, FD1
is a pipe back to FD8 of dpkg-preconfigure.  FD0 is FD7 from
dpkg-preconfigure.

FD3 exists and DEBCONF_REDIR is set when ConfModule::startup() is
called, however it looks like FD3 does not persist into the open2()
call.

It appears
that when kicked off through a preseed install, DEBCONF_REDIR and file
descriptor 3 (hereafter FD3) are set.

These are set all the way down the stack upto and including apt-get.
However when apt-get runs dpkg-preconfigure, FD3 is no where to be
seen.  I put a "sleep" right at the top of dpkg-preconfigure and I
still couldn't see it - so I'm fairly sure that apt-get is not passing
FD3 though.

(It looks like apt-get is closing the FDs in ExecFork which does look
at an APT::Keep-Fds option but I've not looked any further.)

As a gross hack workaround (which might break god knows what else)
I've edited /usr/share/debconf/confmodule to unset DEBCONF_REDIR if
/proc/$$/fd/3 does not exist.

I now no longer get the confmodule line 42: 3: Bad file descriptor
error messages :-)

However, I still get some hangs (one I've seen before). In this case,
dpkg-preconfigure is sat in a waitpid(pidof /tmp/postfix.config...)
whilst postfix.config is sat reading STDIN (which is FD#8 from the
dpkg-preconfigure.  There is a dpkg-preconfigure zombie process
between these two.

Well after several days of pouring over "ls -l /proc/$$/fd" output etc
(and with very little actually understanding of how the
DEBIAN_FRONTEND, DEBIAN_HAS_FRONTEND, DEBCONF_REDIR is supposed to
be being used!), I have some workarounds which let me use puppet with
a debian preseed to make a fully automated install.

I've attached the three patches I'm using (on top of your debconf
patch) which allow this to work.  Mostly they involve checking if file
descriptor #3 is actually open - if it's not then unset DEBCONF_REDIR
(and possibly DEBIAN_HAS_FRONTEND - I only just saw that there is also
DEBIAN_FRONTEND :-( ).    BTW ClientConfModule.pm is the
"Client/ConfModule.pm" file.

Adrian

[ClientConfModule.pm.patch (text/x-diff, attachment)]

[confmodule.patch (text/x-diff, attachment)]

[dpkg-preconfigure.patch (text/x-diff, attachment)]

Message #10 received at [email protected] (full text, mbox, reply):

From: Colin Watson <[email protected]>

To: Adrian Bridgett <[email protected]>, [email protected]

Subject: Re: [Debconf-devel] Bug#439763: debconf: hangs on puppet installs on preseed install

Date: Tue, 18 Dec 2007 17:16:03 +0000

[Message part 1 (text/plain, inline)]

[Sorry for the delay in replying to this.]

On Mon, Aug 27, 2007 at 10:42:41AM +0100, Adrian Bridgett wrote:
> (Mostly cloned from #425397 at the request of Joey Hess.  I've been
> running puppet in the d-i/late-command part of the install to get these
> problems, if it helps I'll try and trim it down (I suspect a preseeded
> postfix install might be sufficient))
> 
> I'm using puppet at the end of a preseeded install and some packages
> keep hanging (sysstat and postfix namely).  I suspect it's the
> preseeding that's causing the problem.  
> 
> What I see is dpkg-preconfigure running /tmp/postfix.config.12345
> (sometimes with a zombie'd process in the middle).  dpkg-preconfigure
> is reading from FD8 (which is FD1 from the postfix.config),  the
> postfix-config in turn is reading from FD0 (which is FD7 from the
> dpkg-preconfigure).  Deadlock ensues :-(

I'm not entirely sure that this is the same thing (I'd probably need to
peruse a full strace -f -s 4096 or so to be sure), but I wonder if this
might be related to
https://bugs.launchpad.net/ubuntu/+source/tasksel/+bug/141601, which
Søren Hansen tracked down to a debconf-apt-progress bug recently. Could
you possibly try r2249 from debconf's Subversion repository? I've
attached the patch for your convenience.

postfix spawns a daemon from its postinst but I don't think sysstat does
(however, it does do other weird stuff from its postinst). Thus, it
would be particularly interesting if this patch fixed one but not the
other.

> When postfix.config runs /usr/share/debconf/confmodules, at line 45 it
> complains that FD3 is a bad file descriptor.  Listing /proc/$$/fd
> shows that FD3 doesn't exist.  FD2 points to the overall stderr, FD1
> is a pipe back to FD8 of dpkg-preconfigure.  FD0 is FD7 from
> dpkg-preconfigure.
> 
> FD3 exists and DEBCONF_REDIR is set when ConfModule::startup() is
> called, however it looks like FD3 does not persist into the open2()
> call.

On the other hand, it might be something completely different. I note
that both sysstat and postfix call db_stop. I don't think this should
actually break the cdebconf instance way down at the bottom of the
stack, but it's possible that it will confuse the debconf running
debconf-apt-progress. If you could run the whole install under
DEBCONF_DEBUG=developer (you did something like this earlier), then this
might shed some light on exactly which debconf frontend is seeing what
commands.

> As a gross hack workaround (which might break god knows what else)
> I've edited /usr/share/debconf/confmodule to unset DEBCONF_REDIR if
> /proc/$$/fd/3 does not exist.

I'm, er, not exactly sure what that will do. :-) I think I'd rather
figure out the underlying problem. It may be that we need to ignore
db_stop in some circumstances.

Cheers,

-- 
Colin Watson                                       [[email protected]]

[debconf.2249.patch (text/x-diff, attachment)]

Message #15 received at [email protected] (full text, mbox, reply):

From: Matthew Palmer <[email protected]>

To: [email protected]

Subject: preseeding hang still present in Lenny

Date: Tue, 5 May 2009 13:09:50 +1000

I've just hit this bug myself, in much the same way as Adrian Bridgett did
-- trying to install and run Puppet inside the late_command of a preseeded
installation.  I'm running Lenny d-i, as built Right Now out of
debian-installer-20090123lenny1 (I'm "slipstreaming" a couple of local udebs
that provide firmware and such into the initrd, hence the local build, but
I'm quite confident that they're not the source of the problem).

I'm trying to track down the problem now.  Any hints for how to get lots of
useful diagnostic info would be appreciated.

- Matt

Message #20 received at [email protected] (full text, mbox, reply):

From: Matthew Palmer <[email protected]>

To: [email protected]

Subject: More info on the hanging debconf problem

Date: Tue, 5 May 2009 14:44:51 +1000

OK, so it's not actually Puppet installs that do it for me... it's
openssh-server (which I was installing alongside it).  Nothing like an
assumption to ruin your day...

At any rate, the symptoms are still substantively the same -- preseeded
install, a hang at "Running preseed", the error
"/usr/share/debconf/confmodule: line 42: 3: Bad file descriptor", and a
zombied dpkg-preconfigure.

Happily, though, it would appear that pkgsel/include (which I've only just
found) works around the problem nicely, and as such it could be argued that
this bug, as currently manifest, isn't all that important.  Whether it
causes (or has the potential to cause) hassle in other circumstances I
couldn't say, though.

- Matt

Message #25 received at [email protected] (full text, mbox, reply):

From: Matthew Palmer <[email protected]>

To: [email protected]

Subject: Workaround insufficient (Debconf hangs on packages installed with late_command)

Date: Thu, 14 May 2009 12:25:09 +1000

Turns out I was wrong to say that using pkgsel/include was a sufficient
workaround -- there's too much stuff that Puppet does to make installing all
those packages in pkgsel/include practical.

I have, however, gotten a puppetless reproduction method.  Just set
preseed/late_command to something like:

apt-get install -y postfix

And the hang will magically present itself.  The symptoms I'm getting are
that the config script is trying to read from fd 0, which is a pipe to an fd
in dpkg-preconfigure, but dpkg-preconfigure is in wait4() on the config
script.

One interesting data point is that the problem seems to be restricted to
packages that use Perl for their config scripts.  Installing less or
ssl-certs, for instance, doesn't hang, but postfix does.  It might not
actually be a perl vs shell problem (I'm sure Postfix's config script is
more complicated than less') but since it happens at dpkg-preconfigure time,
I doubt it can be something like attempting to start a daemon or something
like that.

- Matt

Message #30 received at [email protected] (full text, mbox, reply):

From: Adrian Bridgett <[email protected]>

To: [email protected]

Subject: Re: debconf: hangs on puppet installs on preseed install

Date: Tue, 15 Dec 2009 09:06:35 +0000

Sorry it's taken so long to update the status of this - it's been a
while since I've been in a position to retest it.

It's still breaking on a fresh lenny build for me (debconf 1.5.24) -
exactly the same symptoms.  I wonder if #50595 and #294043 are related
too.  Looks like patches should still apply fine.

Adrian
-- 
bitcube.co.uk - Expert Linux infrastructure consultancy
Puppet, Debian, Red Hat, Ubuntu, CentOS

Message #35 received at [email protected] (full text, mbox, reply):

From: Jacob Luna Lundberg <[email protected]>

To: [email protected]

Subject: Re: debconf: hangs on puppet installs on preseed install

Date: Mon, 28 Feb 2011 18:07:04 -0800

This is still happening with squeeze.  It's something in the environment 
left over from the installer.  This bit of bash is working for me in my 
post-install script:


# Environment
unset -v $(set | sed -e 's/=.*//g' | egrep -v '^(BASH*|HO*|IFS|PATH*|PWD|SHELL*|TERM*)')
. /etc/profile
export LANG=en_US.UTF-8


Before that, the shell variables are:
-------------------------------------

BASH=/bin/bash
BASHOPTS=cmdhist:extquote:force_fignore:hostcomplete:interactive_comments:progcomp:promptvars:sourcepath
BASH_ALIASES=()
BASH_ARGC=()
BASH_ARGV=()
BASH_CMDS=()
BASH_LINENO=()
BASH_SOURCE=()
BASH_VERSINFO=([0]="4" [1]="1" [2]="5" [3]="1" [4]="release" [5]="i486-pc-linux-gnu")
BASH_VERSION='4.1.5(1)-release'
BOOT_IMAGE=debian6.0/i386/linux
DEBCONF_OLD_FD_BASE=4
DEBCONF_REDIR=1
DEBIAN_FRONTEND=newt
DEBIAN_HAS_FRONTEND=1
DIRSTACK=()
EUID=0
GROUPS=()
HOME=/
HOSTNAME=swiss
HOSTTYPE=i486
IFS=$' \t\n'
LANG=C.UTF-8
MACHTYPE=i486-pc-linux-gnu
MENU=/usr/bin/main-menu
OPTERR=1
OPTIND=1
OSTYPE=linux-gnu
PATH=/sbin:/usr/sbin:/bin:/usr/bin
PIPESTATUS=([0]="0")
PPID=6436
PS4='+ '
PWD=/
SHELL=/bin/sh
SHELLOPTS=braceexpand:hashall:interactive-comments
SHLVL=1
TERM=bterm
TERM_FRAMEBUFFER=yes
TERM_TYPE=virtual
UDPKG_QUIET=y
UID=0
USER=root


And after it, they are:
-----------------------

BASH=/bin/bash
BASHOPTS=cmdhist:extquote:force_fignore:hostcomplete:interactive_comments:progcomp:promptvars:sourcepath
BASH_ALIASES=()
BASH_ARGC=()
BASH_ARGV=()
BASH_CMDS=()
BASH_LINENO=()
BASH_SOURCE=()
BASH_VERSINFO=([0]="4" [1]="1" [2]="5" [3]="1" [4]="release" 
[5]="i486-pc-linux-gnu")
BASH_VERSION='4.1.5(1)-release'
EUID=0
HOME=/
HOSTNAME=swiss
HOSTTYPE=i486
IFS=$' \t\n'
LANG=en_US.UTF-8
PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
PIPESTATUS=([0]="0")
PPID=6436
PWD=/
SHELL=/bin/sh
SHELLOPTS=braceexpand:hashall:interactive-comments
TERM=bterm
TERM_FRAMEBUFFER=yes
TERM_TYPE=virtual
UID=0


-Jacob

Message #40 received at [email protected] (full text, mbox, reply):

From: Vincent Caron <[email protected]>

To: [email protected]

Subject: Re: debconf: hangs on puppet installs on preseed install

Date: Tue, 08 Mar 2011 23:27:13 +0100

Hello,

  I'm also using some unattended installation script from
preseed/late_command hook with Squeeze and had the same problem.

  It looks like a debconf frontend mismatch. If I want to use the
current d-i frontend while chrooted in /target I got:

  debconf: unable to initialize frontend: Newt

  And if I rely on the usual trick setting
DEBIAN_FRONTEND=noninteractive, I got the famous "3: bad file
descriptor".

  I'd love to make my preseed script communicate with d-i's frontend but
it looks like it's not possible right now. My current workaround is to
start my preseed script with:

  export DEBIAN_FRONTEND=noninteractive
  unset  DEBIAN_HAS_FRONTEND
  unset  DEBCONF_REDIR
  unset  DEBCONF_OLD_FD_BASE

  Then this works reliably when run from preseed/late_command:

  chroot /target apt-get install -qq -y package ...

Message #45 received at [email protected] (full text, mbox, reply):

From: [email protected] (Marco d'Itri)

To: Matthew Palmer <[email protected]>, [email protected]

Cc: [email protected]

Subject: Re: Bug#439763: More info on the hanging debconf problem

Date: Wed, 4 Jan 2012 17:16:27 +0100

[Message part 1 (text/plain, inline)]

On May 05, Matthew Palmer <[email protected]> wrote:

> OK, so it's not actually Puppet installs that do it for me... it's
> openssh-server (which I was installing alongside it).  Nothing like an
> assumption to ruin your day...
FWIW, I can still reproduce this with Ubuntu 11.10 (but not with 10.04?)
and openssh-server.

-- 
ciao,
Marco

[signature.asc (application/pgp-signature, inline)]

Message #50 received at [email protected] (full text, mbox, reply):

From: John Morrissey <[email protected]>

To: Debconf Developers <[email protected]>

Cc: Colin Watson <[email protected]>, Adrian Bridgett <[email protected]>, [email protected]

Subject: Re: Bug#439763: debconf: hangs on puppet installs on preseed install

Date: Mon, 20 Aug 2012 10:07:18 -0700

On Tue, Dec 18, 2007 at 05:16:03PM +0000, Colin Watson wrote:
> On Mon, Aug 27, 2007 at 10:42:41AM +0100, Adrian Bridgett wrote:
> > When postfix.config runs /usr/share/debconf/confmodules, at line 45 it
> > complains that FD3 is a bad file descriptor.  Listing /proc/$$/fd
> > shows that FD3 doesn't exist.  FD2 points to the overall stderr, FD1
> > is a pipe back to FD8 of dpkg-preconfigure.  FD0 is FD7 from
> > dpkg-preconfigure.
> > 
> > FD3 exists and DEBCONF_REDIR is set when ConfModule::startup() is
> > called, however it looks like FD3 does not persist into the open2()
> > call.
> 
> On the other hand, it might be something completely different. I note
> that both sysstat and postfix call db_stop. I don't think this should
> actually break the cdebconf instance way down at the bottom of the
> stack, but it's possible that it will confuse the debconf running
> debconf-apt-progress. If you could run the whole install under
> DEBCONF_DEBUG=developer (you did something like this earlier), then this
> might shed some light on exactly which debconf frontend is seeing what
> commands.
[snip]
> I think I'd rather figure out the underlying problem. It may be that we
> need to ignore db_stop in some circumstances.

I came across this with d-i wheezy beta1. I have a late_command that runs
Puppet in /target, and most package installations were failing due to EBADF
on fd3, mostly when handling man-db triggers. man-db.postinst fails when the
first db_* tries to write to fd3. Adding an lsof to man-db.postinst shows
that fd3 isn't present.

I can reproduce with something as simple as 'chroot /target apt-get -y
install php5-cli' in late_command.

I've posted the d-i syslog with DEBCONF_DEBUG=developer, late_command, and
env/lsof/strace output at http://horde.net/~jwm/debian/bug439763/ .

Could someone more knowledgable about debconf internals take a look at this,
please? I'm happy to help debug further, but I'm at a loss on how to
continue.

john
-- 
John Morrissey          _o            /\         ----  __o
[email protected]        _-< \_          /  \       ----  <  \,
www.horde.net/    __(_)/_(_)________/    \_______(_) /_(_)__

Message #55 received at [email protected] (full text, mbox, reply):

From: Christian PERRIER <[email protected]>

To: John Morrissey <[email protected]>, [email protected]

Cc: Debconf Developers <[email protected]>, Colin Watson <[email protected]>, Adrian Bridgett <[email protected]>

Subject: Re: Bug#439763: debconf: hangs on puppet installs on preseed install

Date: Tue, 21 Aug 2012 05:56:31 +0200

[Message part 1 (text/plain, inline)]

Quoting John Morrissey ([email protected]):

> I've posted the d-i syslog with DEBCONF_DEBUG=developer, late_command, and
> env/lsof/strace output at http://horde.net/~jwm/debian/bug439763/ .


May I suggest you to create a tarball, compress it with xz, and send
it to this bug? That would guarantee that the data is available over
time if someone comes over this bug report in the future.

The size is not a big problem.

[signature.asc (application/pgp-signature, inline)]

Message #60 received at [email protected] (full text, mbox, reply):

From: John Morrissey <[email protected]>

To: [email protected]

Subject: Re: Bug#439763: debconf: hangs on puppet installs on preseed install

Date: Tue, 21 Aug 2012 07:17:04 -0700

[Message part 1 (text/plain, inline)]

On Tue, Aug 21, 2012 at 05:56:31AM +0200, Christian PERRIER wrote:
> Quoting John Morrissey ([email protected]):
> > I've posted the d-i syslog with DEBCONF_DEBUG=developer, late_command,
> > and env/lsof/strace output at http://horde.net/~jwm/debian/bug439763/ .
> 
> May I suggest you to create a tarball, compress it with xz, and send
> it to this bug? That would guarantee that the data is available over
> time if someone comes over this bug report in the future.
> 
> The size is not a big problem.

For sure, attached.

john
-- 
John Morrissey          _o            /\         ----  __o
[email protected]        _-< \_          /  \       ----  <  \,
www.horde.net/    __(_)/_(_)________/    \_______(_) /_(_)__

[bug439763.tar.xz (application/octet-stream, attachment)]

Message #65 received at [email protected] (full text, mbox, reply):

From: "Juan Carlos Moreno [ackstorm]" <[email protected]>

To: [email protected]

Subject: Re: debconf: hangs on puppet installs on preseed install

Date: Thu, 23 Aug 2012 11:14:34 +0200

Hi there,

same problem here, using puppet and installing mysql-server, solved (for
me) redirecting puppet log to syslog: puppet apply --logdest syslog
XXXXX

Hope that helps someone
Regards

-- 

Un saludo.
Juan Carlos Moreno
[email protected]
Project Manager
 
-----------------------------------------------------------------

Message #70 received at [email protected] (full text, mbox, reply):

From: "Juan Carlos Moreno [ackstorm]" <[email protected]>

To: [email protected]

Subject: Re: debconf: hangs on puppet installs on preseed install

Date: Thu, 23 Aug 2012 11:42:44 +0200

Oppsss, sorry! I have made a lot of changes...

Redirecting puppet logdest is not the solution but setting
'DEBCONF_REDIR' to empty, it works!

This is how I call puppet, and it works for me (now):

[..]
env = environ.copy()
env['DEBCONF_DEBUG']='developer'
env['DEBCONF_REDIR']=''
p = Popen(['puppet', 'apply', '--detailed-exitcodes',
           '--debug','--apply', data_pson],
            cwd=cwd, stdin=None, env=env)
stdout, stderr = p.communicate()
[..]


On Thu, 2012-08-23 at 11:14 +0200, Juan Carlos Moreno [ackstorm] wrote:
> Hi there,
> 
> same problem here, using puppet and installing mysql-server, solved (for
> me) redirecting puppet log to syslog: puppet apply --logdest syslog
> XXXXX
> 
> Hope that helps someone
> Regards
> 

-- 

Un saludo.
Juan Carlos Moreno
[email protected]
Project Manager
 
Oficina central: 902 888 345
 
ACK STORM, S.L.
ISO 9001:2008 (Cert.nº. 536932)
http://ackstorm.es

Este mensaje electrónico contiene información de ACK STORM, S.L. que es
privada y confidencial, siendo para el uso exclusivo de la persona(s) o
entidades arriba mencionadas. Si usted no es el destinatario señalado,
le informamos que cualquier divulgación, copia, distribución o uso de los
contenidos está prohibida. Si usted ha recibido este mensaje por error,
por favor borre su contenido y comuníquenoslo en la dirección
[email protected]
-----------------------------------------------------------------

Message #75 received at [email protected] (full text, mbox, reply):

From: John Morrissey <[email protected]>

To: [email protected], [email protected]

Subject: Re: [Debconf-devel] Bug#439763: debconf: hangs on puppet installs on preseed install

Date: Thu, 23 Aug 2012 10:33:02 -0700

On Thu, Aug 23, 2012 at 11:42:44AM +0200, Juan Carlos Moreno [ackstorm] wrote:
> Oppsss, sorry! I have made a lot of changes...
> 
> Redirecting puppet logdest is not the solution but setting
> 'DEBCONF_REDIR' to empty, it works!
> 
> This is how I call puppet, and it works for me (now):
> 
> [..]
> env = environ.copy()
> env['DEBCONF_DEBUG']='developer'
> env['DEBCONF_REDIR']=''
> p = Popen(['puppet', 'apply', '--detailed-exitcodes',
>            '--debug','--apply', data_pson],
>             cwd=cwd, stdin=None, env=env)
> stdout, stderr = p.communicate()
> [..]

Thanks for the info, Juan.

At least for me with d-i wheezy beta1, if I only unset DEBCONF_REDIR,
scripts in the chroot still try to communicate with the d-i debconf frontend
(outside the chroot), and block because something seems amiss with the file
descriptor manipulation. I suspect this is because DEBCONF_HAS_FRONTEND is
still set, so confmodule doesn't spawn a different frontend.

As a result, man-db.postinst's call to db_version emits the debconf command
to stdout, where it gets caught by log-output (if I chroot my script with
in-target, the same blocked postinst happens if I simply chroot my script):

    Aug 23 16:55:01 log-output: VERSION 2.0

And the postinst blocks on reading the unchrooted debconf frontend's
response, since the frontend never got the request:

root     12444  0.1  5.2  62140 26556 tty1     S+   16:54   0:00
   \_ apt-get -y install php5-cli
root     12461  0.0  2.1  29580 10760 tty1     S+   16:55   0:00
     \_ /usr/bin/dpkg --status-fd 22 --unpack --auto-deconfigure /var/cache/apt/archives/php5-common_5.4.4-4_amd64.deb /var/cache/apt/archives/php5-cli_5.4.4-4_amd64.deb
root     12483  0.0  0.1   4164   628 tty1     S+   16:55   0:00
       \_ /bin/sh -e /var/lib/dpkg/info/man-db.postinst triggered /usr/share/man

root@debian:/# strace -p 12483
Process 12483 attached - interrupt to quit
read(0, ^C <unfinished ...>
root@debian:/# lsof -p 12483 | grep 0[rwu]
man-db.po 12483 root    0r  FIFO    0,8      0t0   3215 pipe

root       235  1.1  3.9  50268 20180 tty1     S+   16:39   0:14 debconf -o d-i /usr/bin/main-menu

root@debian:/# lsof -p 235 | grep 3125
debconf 235 root    4w  FIFO                0,8      0t0 3215 pipe


I also tried unsetting DEBIAN_HAS_FRONTEND, so a separate debconf frontend
gets spawned in the chroot. Unfortunately, the chrooted frontend apparently
can't interact with the console, since its 0 and 1 FDs are still pipes to
the unchrooted debconf frontend.

If I unset DEBCONF_REDIR and DEBIAN_HAS_FRONTEND, then use 'debconf-disconnect
chroot /target myscript', Puppet package installations still get EBADF when
trying to interact with file descriptor 3.

john
-- 
John Morrissey          _o            /\         ----  __o
[email protected]        _-< \_          /  \       ----  <  \,
www.horde.net/    __(_)/_(_)________/    \_______(_) /_(_)__

Message #80 received at [email protected] (full text, mbox, reply):

From: Michael Prokop <[email protected]>

To: John Morrissey <[email protected]>, [email protected]

Subject: Re: Bug#439763: [Debconf-devel] Bug#439763: debconf: hangs on puppet installs on preseed install

Date: Thu, 24 Apr 2014 14:19:22 +0200

[Message part 1 (text/plain, inline)]

* John Morrissey [Thu Aug 23, 2012 at 10:33:02AM -0700]:

[...]
> As a result, man-db.postinst's call to db_version emits the debconf command
> to stdout, where it gets caught by log-output (if I chroot my script with
> in-target, the same blocked postinst happens if I simply chroot my script):
[...]

Sadly I'm still seeing this on Debian/wheezy systems
(puppet, screen + man-db all being from wheezy - package
installation running via puppet agent):

,---- [ rewrapped c/p from puppet dashboard log ]
| Unpacking screen (from .../screen_4.1.0~20120320gitdb59704-7_amd64.deb) ...
| Processing triggers for install-info ...
| Processing triggers for man-db ...
| /var/lib/dpkg/info/man-db.postinst: 44: /var/lib/dpkg/info/man-db.postinst: 3: Bad file descriptor
| dpkg: error processing man-db (--unpack): subprocess installed post-installation script returned error exit status 2
| Errors were encountered while processing: man-db E: Sub-process /usr/bin/dpkg returned an error code (1)
`----

Has anyone made any further progress on this issue?

regards,
-mika-

[signature.asc (application/pgp-signature, inline)]

Message #85 received at [email protected] (full text, mbox, reply):

From: Will Aoki <[email protected]>

To: [email protected]

Subject: Re: Bug #439763 [Debconf-devel] Bug#439763: debconf: hangs on puppet installs on preseed install

Date: Mon, 28 Apr 2014 09:33:38 -0600

This started happening on April 8th on most of my wheezy systems. I'm
running apt-get from cfengine, not puppet, but symptoms are otherwise
the same, e.g.:

cfengine:xxxxx:/bin/sh -c 'DEB: (Reading database ... 326575 files and directories currently installed.)
cfengine:xxxxx:/bin/sh -c 'DEB: Preparing to replace openssl 1.0.1e-2+deb7u5 (using .../openssl_1.0.1e-2+deb7u6_amd64.deb) ...
cfengine:xxxxx:/bin/sh -c 'DEB: Unpacking replacement openssl ...
cfengine:xxxxx:/bin/sh -c 'DEB: Processing triggers for man-db ...
cfengine:xxxxx:/bin/sh -c 'DEB: /var/lib/dpkg/info/man-db.postinst: 44: /var/lib/dpkg/info/man-db.postinst: 3: Bad file descriptor
cfengine:xxxxx:/bin/sh -c 'DEB: dpkg: error processing man-db (--unpack):
cfengine:xxxxx:/bin/sh -c 'DEB:  subprocess installed post-installation script returned error exit status 2
cfengine:xxxxx:/bin/sh -c 'DEB: Errors were encountered while processing:
cfengine:xxxxx:/bin/sh -c 'DEB:  man-db
cfengine:xxxxx:/bin/sh -c 'DEB: E: Sub-process /usr/bin/dpkg returned an error code (1)

This hadn't been happening before that date. I haven't identified why it
works on some but not on others -- there isn't supposed to be any
configuration difference between the two groups.

Message #90 received at [email protected] (full text, mbox, reply):

From: Will Aoki <[email protected]>

To: [email protected]

Subject: Re: Bug #439763 [Debconf-devel] Bug#439763: debconf: hangs on puppet installs on preseed install

Date: Mon, 5 May 2014 11:21:22 -0600

On Mon, Apr 28, 2014 at 09:33:38AM -0600, Will Aoki wrote:
> This hadn't been happening before that date. I haven't identified why it
> works on some but not on others -- there isn't supposed to be any
> configuration difference between the two groups.

All the computers I've had that experience this problem were using dash
for /bin/sh. Using bash for /bin/sh has, so far, made the problem go
away; however, not all computers with dash as /bin/sh have been
experiencing this problem.

Message #95 received at [email protected] (full text, mbox, reply):

From: Tilman Blumenbach <[email protected]>

To: [email protected]

Subject: Similar problem with Puppet 3.7.1

Date: Mon, 24 Nov 2014 18:18:50 +0100

We are having a similar problem with Puppet 3.7.1 (running as a puppet-apply provisioner under Vagrant 1.6.3).

When Puppet installs packages using apt-get, the package installation sometimes hangs when running the mandb postinst trigger (it does not look like it is always the same package which causes the hang); closer investigation reveals that it is in fact the mandb process which hangs because apparently a file descriptor it tries to access has become inaccessible.

Setting DEBCONF_REDIR="" when running puppet-apply seems to help.

Message #100 received at [email protected] (full text, mbox, reply):

From: Carl Karsten <[email protected]>

To: [email protected]

Subject: in-target seems to fix

Date: Wed, 19 Apr 2017 01:48:41 -0500

[Message part 1 (text/plain, inline)]

I had 'this problem' I guess.


I was able to repo it with something like
 late_command ... chroot target apt install unattended-updates

I changed chroot target to in-target and the problem went away.


-- 
Carl K

[Message part 2 (text/html, inline)]

Send a report that this bug log contains spam.