open-iscsi shutdown failure due to missing dir

I have the same issue. It seems that the removal of /etc/init.d/mountkernfs.sh introduced this problem.

Same problem with
Ubuntu 10.04
open-iscsi 2.0.871-0ubuntu4

I can confirm that the patch in Bug 567143 (mkdir -p /lib/init/rw/sendsigs.omit.d in /etc/init.d/open-iscsi) works around this issue.

But open-iscsi is not the only init script that relies on /lib/init/rw/sendsigs.omit.d, so it is not open-iscsi's job to create it. This is clearly a bug in the initscripts package (source package sysvinit), and I have opened a bug against that package:

https://bugs.launchpad.net/ubuntu/+source/sysvinit/+bug/603760

I certainly don't think we should be fixing this in every package that uses sendsigs.omit.d.

The convention in Ubuntu has been to use /var/run rather than /lib/init/rw for these things.
I can easily add /var/run/sendsigs.omit.d/ support to sendsigs, and I've done so in my local tree, but I don't really feel like playing whack-a-mole to move all the users over to that directory, and having /var/run/sendsigs.omit on the one hand and /lib/init/rw/sendsigs.omit.d/ on the other seems wilfully confusing.

I've asked Scott if he'd object to a new mounted-libinitrw.conf job in mountall that just does 'exec mkdir -p /lib/init/rw/sendsigs.omit.d'. I think this would be the easiest fix. (Note that, since /lib/init/rw is a mountpoint, we can't just ship the directory in the initscripts package.)

This bug was fixed in the package sysvinit - 2.87dsf-4ubuntu18

---------------
sysvinit (2.87dsf-4ubuntu18) maverick; urgency=low

  * Allocate pidof/killall5 omitpid buffers dynamically. 16 is too small
    for killall5 now that all Upstart jobs are omitted.
  * Create /lib/init/rw as a symlink to /var/run on new installations, and
    fix it up in /etc/init.d/umountroot on upgrade, as it's difficult to do
    this at any other time; this saves us chasing around all the individual
    packages that use one or the other for sendsigs.omit.d (LP: #541512).
  * Handle /var/run/sendsigs.omit.d explicitly, just in case.
 -- Colin Watson <email address hidden> Fri, 24 Sep 2010 10:48:28 +0100

This bug was fixed in the package mountall - 2.18

---------------
mountall (2.18) maverick; urgency=low

  [ Dustin Kirkland ]
  * conf/mounted-varrun.conf: seed /var/run/motd on boot, to ensure that
    initial logins are quick, LP: #587858

  [ Steve Langasek ]
  * Don't check /etc/environment for locale settings; this is an obsolete
    usage we transitioned away from years ago.
  * Add LC_ALL to the list of exported locale variables.

  [ Colin Watson ]
  * Flush standard output and error before daemonising, so that we don't get
    three copies of our early debug output.
  * If a mountpoint is a symlink to another mountpoint, add a dependency on
    the link target rather than trying to mount it again and failing
    (LP: #541512).
 -- Colin Watson <email address hidden> Fri, 24 Sep 2010 10:52:20 +0100

Could we please have this in lucid-updates?
It breaks nbd-client too.

any word on if we can have this in lucid-updates?

agree, need this on LTS please!

Agree! we need this on LTS ... please......

Since this has not been fixed in lucid I personally believe that the status cannot be listed as fix released. maverick is not LTS and is therefore not suitable in many server environments where iSCSI would be used.

This is still a problem on Ubuntu 10.04.3.

Just ran into this issue too, going to try and get this into lucid-updates

I can confirm that problem replicates in our own Lucid 10.04.3. I also agree that a bug that impacts an current LTS should be fixed in the updates for the current LTS.

FYI: the portmap package in Lucid creates this directory on boot, so that a symlink to the portmap.pid can be placed there. Installing it is a way of ensuring that this directory is created properly without creating a custom hack. I believe that later versions of this included in packages after Lucid may change this behavior.

I discovered this bug on a fresh 10.04.4 Server install today. I'm surprised it hasn't been fixed in Lucid. Ubuntu folks, is it possible that this could still be considered for an SRU? Is there something I could do (e.g. prepare a debdiff) to move the process forward?

Freshly installed LTS, apparently portmap creates the needed directory, still the iscsi shutdown hangs...

I can confirm that installing portmap fixes the shutdown issue on 10.04.4. The mkdir workaround in the init script does not work correctly any longer. This should be addressed in SRU as it is a serious loss of stability when working with remote servers.