[PLUG] Help with large file samba file transfer

[Ryan E. Helfter]

How are your routing tables set up?  `netstat -rn`

is there any collisions on the hub?  it is possible, however very unlikely
that your tcp stack is somewhat hosed...  if memory serves me correctly,
the tcp stack is handled by the libc libraries...  (not quite sure, but
pretty sure)

ry

On Mon, 16 Oct 2000, Eric Pratt wrote:

> Date: Mon, 16 Oct 2000 19:42:16 -0500 (EST)
> From: Eric Pratt <pratte@ecn.purdue.edu>
> To: plug@csociety.purdue.edu
> Subject: [PLUG] Help with large file samba file transfer
> 
> 
> I have a RedHat 6.2 machine running samba.  Until recently, I have had no
> problems with it.  Now, I cannot copy large files from the Linux machine
> to a PC on the same local LAN (10.0.2.0 network via the eth1
> interface.  eth0 is the gateway device to the internet.)
> 
> To be more exact, it happens with large data transfers via eth1.  If
> I try to transfer a file the size of an mp3, it will take about one minute
> to do the transfer.  I only have a 10Mbps hub, but it should still only
> take about 3 or 4 seconds to transfer a song (it used to.)  Watching the
> hub, I see that during file transfer, the activity lights will go crazy
> for about a second, and then will go dead for about 15 - 30 seconds, and
> then become active for another second and go dead again.  It will do this
> until the file is transferred.  But, I can transfer small files (in the KB
> range) with no problem and I can stream mp3s all day.  I cannot however
> stream a 3Mbps mpeg video.  I also experience this problem when
> starting and XDM session with the Linux machine.  The Linux machine is
> acting as a  firewall.  When this is going on, all access to the internet
> is stopped as well including telneting into the linux machine.  All
> services come back during the 1 second burst of activity, but then go away
> again.  I have checked samba logs and syslogd logs, I have checked to make
> sure that all services are still running using ps, and I cannot find
> anything wrong.
> 
> I have a feeling it has something to do with the way eth1 is setup but
> cannot figure out what else to look at.
> 
> Any ideas?
> 
> Eric
> 
> 
> ____________________________________________________
> The Purdue Linux Users' Group (PLUG) mailing list.
> plug mailing list  -  plug@csociety.purdue.edu
> http://csociety.ecn.purdue.edu/mailman/listinfo/plug
> 

-- 
Ryan E. Helfter
reh@blacksoul.net
Black Soul Networks, LLC
--