Samba dropping connections

[Jerry Clark]

Can anyone help ?

I have installed SAMBA on a local DELL machine - Samba version 2.2.2 running
under RedHat 6.2.

I have configured SAMBA using SWAT and 2 Linux directories are made
accessible to Windows users as NT shares - and they map each share to a
drive letter - the Linux box acts purely as an NT file-server. All
authentication is done locally - each user has an entry in /etc/smbpasswd
and well as in /etc/passwd. The machine sits on a sub-net of a HUGE network
that is primarily a Windows network but has many other UNIX and Linux boxes
attached. My sub-net has a 10Mbit link onto the main network which is mainly
100 and 1000MBit with 20Mbit links to regional offices. I have users in
London, Bristol, Birmingham and Belfast all on various sub-nets of a HUGE
network. In total there are about 50 users connected during the day with a
total of between 2 and 8 files open at any one time - so you can see its not
exactly hammered. Most of the machines on the network are either Windows PCs
(desktop machines) or NT servers - all the servers being managed by a huge
IT department. I needed an NT file share that could be accessed by my users
from PCs and Macs. Sadly corporate IT do not offer a Mac file-serving
service - they consider Macs to be a security risk as they cannot enforce a
login name/password security mechanism on Macs. So I decided to host the
share on a Linux box running SAMBA and Netatalk to "export" Linux
directories to PCs and Macs.....

Its nearly been successful......but......users (PC users) complain of the
SAMBA shares disconnecting from their PCs. Typically they would be working
on, say, an Excel spreadsheet that they have loaded from the server and when
they come to save it says that the server has disconnected. They then have
to save the document on their local disk. They then re-connect ("exploring"
the "Mapped" drive seems to be enough for the PC to re-connect) and then
they can load the spreadsheet from the local file and save it back onto the
server. I've not had this problem myself so I am relying on rather woolly
user feed-back.....

In the file(s) /var/log/samba/log.pc-?????? (all the PCs have names like
pc-132485) I get error messages saying something like:

[2001/12/04 13:04:45, 0] lib/util_sock.c:read_socket_data(478)
  read_socket_data: recv failure for 4. Error = Connection timed out

or:

[2001/12/04 16:28:25, 0] lib/util_sock.c:read_socket_data(478)
   read_socket_data: recv failure for 4. Error = Connection reset by peer

which seem to coincide with that particular PC getting disconnected.

Can anybody help ? I'm getting alot of hassle from senior management about
the ineffectiveness of the system......
All  users have an NT hosted share mapped all the time - they NEVER get the
NT server disconnecting.....

I get other errors reported in /var/log/samba/log.smbd like this:

[2001/12/04 23:12:50, 0] smbd/connection.c:yield_connection(63)
  yield_connection: tdb_delete for name  failed with error Record does not
exist.

but these seem to coincide with my own machine opening the NT share - but
WHY my machine ? its an ordinary W98 desktop like everyone
else's......surely SAMBA doesn't know that it was me who installed/set up
the Linux/SAMBA box ?????? That would be extraordinary......

Any help appreciated......
I can send /etc/smb.conf and any other info if it would help.......

I have noticed a couple of things:

Setting the socket options - the default option seems to be TCP_NODELAY - I
have tried TCP_NODELAY, IPTOS_LOWDELAY and SO_KEEPALIVE in all possible
combinations - when I have SO_KEEPALIVE set I set "keepalive" = 0, otherwise
"keepalive" is set to its default of 300. When SO_KEEPALIVE is set the
problem gets worse - about twice as many users get "disconnected".....

The users with most problems are the "power" users who have several files
open most of the day and the users who are at the "far-end" of the network -
those in Belfast and those who have "RAS"ed in....

Hoping someone can help,
Cheers,
Jerry

> Jerry Clark
> Manager, Design Systems
> BBC MediaArc
> 
> *  	Room LG692 Television Centre, Wood Lane, London, W12 7RJ    
> *	Work:    020 8225 6707  
> *	Mobile:  07710 350190
> *	Fax:        020 8576 8727  
> 
> *	mailto:jerry.clark at bbc.co.uk 
> *	http://www.bbcmediaarc.com 
> 


This e-mail, and any attachment, is confidential. If you have received
it in error, please delete it from your system, do not use or disclose
the information in any way, and notify me immediately. The contents of
this message may contain personal views which are not the views of the
BBC, unless specifically stated.