[Samba] XP clients disconnected during trasnfer of larger files tothe samba server

[Derrick MacPherson]

>   wins support = true
>    name resolve order = bcast hosts
> which says run support for wins, but don't use it to resolve netbios
names.  Good > to use it and point the secondary server at it plus all
your windows machines.  
> Right now, you only resolve netbios on the local subnet plus DNS the
same 
> as netbios.

Most of the non-default settings come from extended googling and come
back with possible suggestions to other issues I've seen, such as:

> You also have this
>    use spnego = no
>  From the doco:
> Unless further issues are discovered with our SPNEGO implementation,
there is 
> no reason this should ever be disabled. Have you had an issue?  If
not, 
> delete it.

Ya we were. We were getting about 10-15 seconds of 'delay' when clicking
on a share, and this seem to fix it. Is this a smoking gun of some
problem? I had googled some errors in the logs we were seeing and this
was suggested to fix that problem and it did.

> Another override:
>    oplocks = no
> Use this to avoid specific issues with specific shares or filetypes.

Again from trying to see what could help our problem

> Last, socket endpoint not connected is frequently a hardware issue. I
don't know > the protocol inside out, but it seems the smb protocol is
less forgiving 
> than ftp.  Switch and interface card issues during saturation become
an 
> issue with smb.

Hmm. OK. Well these machines have onboard nics (Tigon3
[partno(BCM95721)), and a quad port Intel nic, though they were not
detected on a duplicate machine by freebsd so we were using the onboard
nics, I'll get that switched to the quad intels, see if that helps.

Thanks Doug, much appreciated.


Regards, Doug


> 
> -----Original Message-----
> From: Derrick MacPherson
> Sent: Friday, August 11, 2006 1:23 AM
> To: Derrick MacPherson; samba at lists.samba.org
> Subject: RE: [Samba] XP clients disconnected during trasnfer of larger
> files tothe samba server
> 
> 
> Any suggestions to look into, or more debug info required?
> 
> The box is running centos 4.3 final, up to date with the latest 
> versions of CentOS updates,, it's running something like 3.0.10 or .11

> - can't recall. 1:22 am. Off to bed...
> 
> Thanks.
> 
> -----Original Message-----
> From: samba-bounces+dmacpherson=mainframe.ca at lists.samba.org
> [mailto:samba-bounces+dmacpherson=mainframe.ca at lists.samba.org] On 
> Behalf Of Derrick MacPherson
> Sent: Thursday, August 10, 2006 5:47 PM
> To: samba at lists.samba.org
> Subject: [Samba] XP clients disconnected during trasnfer of larger 
> files tothe samba server
> 
> 
> i'm having an issue when transferring large files to the samba servers

> from an xp client - files about 1GB or larger. about 70% into the 
> transfer i get a network share no longer exists error and the transfer

> fails. I can pull down from the server fine with no issues. The XP 
> machines are authenticating from a different Samba server though the 
> problem is with that machine as well. Error and config posted below: 
> (FYI - transfering same files and such work fine via FTP)
>         
>         My secondary server config:
>          netbios name = 3Dsrv
>          workgroup = VFX
>          security = user
>          server string = %h server (3D FileServer)
>          password server = 192.168.0.210
>          username map = /etc/samba/smbusers
>          idmap uid = 15000-20000
>          idmap gid = 15000-20000
>         name resolve order =  bcast hosts
>         template primary group = "Domain Users"
>         template shell = /bin/bash
>         winbind separator = +
>         socket options = TCP_NODELAY IPTOS_LOWDELAY SO_SNDBUF=16384 \
>                 SO_RCVBUF=16384
>         oplocks = no
>         
>         smb log - i think this is relevant, though not sure:
>         
>         2006/08/10 11:53:56, 0] lib/util_sock.c:get_peer_addr(1000) 
>         getpeername failed. Error was Transport endpoint is not
>         connected 
>         [2006/08/10 11:53:56, 0] lib/util_sock.c:get_peer_addr(1000) 
>         getpeername failed. Error was Transport endpoint is not
>         connected 
>         [2006/08/10 11:53:56, 0]
lib/:util_sock.c:write_socket_data(430) 
>         write_socket_data: write failure. Error = Connection reset by
>         peer 
>         [2006/08/10 11:53:56, 0] lib/util_sock.c:write_socket(455) 
>         write_socket: Error writing 4 bytes to socket 24: ERRNO =
>         Connection
>         reset by peer
>         
> Primary server config:
> 
> [global]
>         name resolve order = bcast hosts
>         passwd chat debug = yes
>         idmap gid = 15000-20000
>         passwd chat = *Enter\snew\sUNIX\spassword:* %n\n *Retype\snew
> \sUNIX\spassword:* %n\n
>         passwd program = /usr/bin/passwd %u
>         netbios name = 2DSRV
>         printing = CUPS
>         idmap uid = 15000-20000
>         logon script = logon.bat
>         workgroup = VFX
>         os level = 128
>         printcap name = CUPS
>         security = user
>         add machine script = /usr/sbin/useradd -s /bin/false/ -d 
> /var/lib/nobody %u
>         delete user script = /usr/sbin/userdel -r %u
>         log level = 4
>         add group script = /usr/sbin/groupadd %g
>         socket options = TCP_NODELAY IPTOS_LOWDELAY SO_SNDBUF=8192 
> SO_RCVBUF=8192
>         delete group script = /usr/sbin/groupdel %g
>         add user to group script = /usr/sbin/usermod -G %g %u
>         logon drive =
>         domain master = yes
>         username map = /etc/samba/smbusers
>         use spnego = no
>         encrypt passwords = yes
>         passdb backend = tdbsam
>         logon home =
>         wins support = true
>         server string = %h server (Domain Controller)
>         unix password sync = yes
>         logon path =
>         add user script = /usr/sbin/useradd -m %u
>         domain logons = yes
>