ADSM-L

Re: TCP/IP connection failure

1996-12-14 17:11:51
Subject: Re: TCP/IP connection failure
From: Frank Ramke <ramke AT VNET.IBM DOT COM>
Date: Sat, 14 Dec 1996 14:11:51 PST
If the ADSM server is disconnecting you, try the following:

Double the  COMMTIMEOUT and IDLETIMEOUT values in your DSMSERV.OPT
file and turn compression off at the client if not running pentium class
machine.

The ADSM server will severe the session if it doesn't hear from
the client ever so often...specifically, the IDLETIMEOUT value
when the session is in IDLE state, and COMMTIMEOUT when the session
is active. You will see a message at the server console that indicates
the session has been severed if one of these thresholds is hit.

You may also be having tcpip or network problems. I've seen
considerable problems with tcpip on netware on ethernet networks.
If you are on ethernet, try setting tcpbuffersize and tcpwindow
size to lower values. These are located in the dsm.opt file on
the client. Also, if you are crossing any bridges/routers, novell has
documented an option for sys:\etc\tcpip.cfg  in the latest
tcpip patch that may help. I'm sorry, but I can't remember the
option name at the moment.

Frank Ramke
ADSM Development
================================================================

At 09:45 AM 12/13/96 -0500, you wrote:
>I have a problem with an older 3.11 Netware server that continues to
>disconnect itself from the ADSM AIX server.  It seems to timeout on
>larger files (in excess of 50MB).  The documentation for the error
>listed below is very nondescript.  In a nutshell, the documentation is
>saying that either the LAN connection went down or the job was
>cancelled.  I have changed the client communication timeout to 300
>seconds in the dsmserv.opt, to no avail.  I have not altered the
>TCPBuffersize of TCPWindowsize in the dsm.opt client file.
>
>Has anyone done this?  If so what are the ramifications of doing such?
> Lastly, am I barking up the wrong tree?
>
>Any comments would be appreciated.
>
>Thanks!
>
>12/12/1996 20:22:51  TcpFlush: Error 60 sending data on Tcp/Ip socket
>78439588.
I have been getting this failure for over 2 months on one volume of a
Netware client.  I have opened an ETR, 0771E with IBM.  I have sent the
output of a MODULES command on the client to IBM.  I am still waiting for a
response.  I agree that the problem appears to be associated with larger
files, sometime only as big as 20M on the one volume.  Other volumes backup
from the same client successfully, every day.  I get the TCP/IP disconnect
every day.

We have the COMMITTIMEOUT at 300 seconds.  If memory serves, the buffer and
window sizes are maxed.  I was having the same problem with another client,
but they moved the machine, reloaded the software and the problem went away.
The client admin loaded the newest version of the TCPIP.NLM, but the problem
still exists.  I am waiting for a time that the client admin can reload
everything and see if that clears it up.  In the meantime, I wait.....
Deb Young
Dow Corning Corp
<Prev in Thread] Current Thread [Next in Thread>