race condition on startup between samba and cups

This bug was fixed in the package samba - 2:3.6.6-3ubuntu2

---------------
samba (2:3.6.6-3ubuntu2) quantal; urgency=low

  * Ensure samba can query cups for printer information on startup
    (LP: #1047262):
    - d/samba.smbd.upstart: Optionally wait for cups to be in state
      'running' if cups is installed.
 -- James Page <email address hidden> Fri, 07 Sep 2012 10:58:08 +0100

The output of:

  sudo testparm -s

Would also be useful - please obfuscate any sensitive information first.

Thanks

Hi James,

        if status cups >/dev/null ; then
                start wait-for-state WAITER=smbd WAIT_FOR=cups WAIT_STATE=running
        fi

This solution has the effect that if the cups job is broken for any reason, smbd will also not be started. I don't think that's desirable.

We've tried hard in the past to ensure samba is *not* dependent on cups being present before samba starts. I think this would be better done as a job in the samba package which is 'start on started cups' and will prompt samba to reload its config.

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Hi Steve

On 07/09/12 16:36, Steve Langasek wrote:
> Hi James,
>
> if status cups >/dev/null ; then start wait-for-state WAITER=smbd
> WAIT_FOR=cups WAIT_STATE=running fi
>
> This solution has the effect that if the cups job is broken for
> any reason, smbd will also not be started. I don't think that's
> desirable.
>
> We've tried hard in the past to ensure samba is *not* dependent on
> cups being present before samba starts. I think this would be
> better done as a job in the samba package which is 'start on
> started cups' and will prompt samba to reload its config.

I'd not considered doing it that way - it would de-couple samba
startup from cups which I agree is preferable - I'll take a look and
do the rework early next week.

Thanks for the feedback.

- --
James Page
Ubuntu Core Developer
Debian Maintainer
<email address hidden>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://www.enigmail.net/
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=233E
-----END PGP SIGNATURE-----

@Steve

I've proposed a branch with a separate job to complete the reload post start of cups.

Your review would be appreciated.

This bug was fixed in the package samba - 2:3.6.6-3ubuntu3

---------------
samba (2:3.6.6-3ubuntu3) quantal; urgency=low

  * Decouple startup of smbd from cups (LP: #1047262):
    - d/samba.smbd.upstart: Revert changes made in 2:3.6.6-3ubuntu2.
    - d/samba.reload-smbd.upstart: Add upstart task which reloads smbd
      once cups has started, ensuring that smbd startup is decoupled
      from cups.
    - d/rules: Install reload-smbd upstart configuration, don't try to
      start on install.
  * Install winbind upstart configuration file with --upstart-only option
    for consistency with samba package.
 -- James Page <email address hidden> Wed, 12 Sep 2012 13:24:16 +0100

>* Install winbind upstart configuration file with --upstart-only option
> for consistency with samba package.

Heh, this part is a wrong change. The upstart-only flag is for upstart jobs which aren't replacing init scripts of the same name.

Are there any plans to fix this issue in Precise now that Quantal has a working patch? Thanks.

I tried the reload-smbd job on Precise but it doesn't work. The job is definitely started after cups but the queues still aren't there. Works fine in Quantal.

Running 'start smbd; start cups; reload smbd' while the system is running does work, and adding a 'sleep 1' to reload.smbd.conf also works, so I guess having this work properly depends on some other change (in cups, possibly? forking too soon?).

I was going to propose an SRU with just the reload-smbd.conf from Quantal but it's looking more complicated than that. Any ideas?

The Precise Pangolin has reached end of life, so this bug will not be fixed for that release