[Samba] Serious Slowness Issues with Printing

[Kazsuk, Tim]

After some time looking into this problem, I was able to track the
slowness problem down to the smbpasswd file or what seems to be a
problem with the smbpasswd file and having the "security" option set to
"SHARE".

When we upgraded to 3.0.11, I moved over the previous smbpasswd file
from 2.0.7, which has all the entries for the 3,192 associates we have.
During the slowness I tried resetting some of the entries for those
associates having the worst problem (smbpasswd <username> <passwd>),
this did not work.  I tried moving the entries for those associates to
the top of the smbpasswd file, this did not work.  It wasn't until I
removed the smbpasswd file that I saw a difference.  I then went in and
removed the entry "smb passwd file=" via SWAT and things dramatically
sped up with access times on the Windows clients.  

Can anyone shed some light on why this may be?  Is this the way it's
suppose to behave?  Are the others having issues with printing slowness
setup with the "security" option set to "SHARE" and a smbpasswd file
specified?

Regards,

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Tim Kazsuk 
Senior Systems Engineer 
The Home Depot Supply Division 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~


-----Original Message-----
From: Jeremy Allison [mailto:jra at samba.org] 
Sent: Friday, June 17, 2005 9:40 AM
To: Kazsuk, Tim
Cc: William Enestvedt; samba at lists.samba.org
Subject: Re: [Samba] Serious Slowness Issues with Printing

On Fri, Jun 17, 2005 at 09:21:37AM -0400, Kazsuk, Tim wrote:
> In our case, the jobs being sent to the printers are fairly small
across
> the board, with little or no graphics.  The biggest problem is the
time
> it takes Office to open up with a Samba printer set as default on the
> PC.  If a fake local printer is installed and set as default, Office
> opens up quickly, but when the end user has to print, going in to
select
> the printer can take some time to select on and then print to it.
This
> is becoming a big problem.
> 
> I did compile the version I'm using, and for the most part took the
> defaults on the options.
> 
> I've looked at the underlying disks (iostat is showing no problems),
but
> not yet at the network settings on the server.  We're running Gigabit
> from the server to the switch, all desktops are 100 Full Duplex.  Our
> current settings are auto-auto on the Cisco switch side and auto-auto
on
> the desktop side for Speed-Duplex.  Leaving them as auto-auto or
locking
> them down to 100 Full Duplex does not seem to make a difference.

The main thing to look at is the network traffic between client and
Samba server - ie. does the wait occur on the client when preparing
the job or in the network traffic between client and server or at
the server end once the job has been received at the server ?

This is the first step you need to take in debugging a problem like
this. The Samba server submits the job to the backend once it receives
the EndJob RPC call.

Jeremy.