ADSM-L

MVS TCP/IP failures - possible resolution

1998-02-25 17:30:24
Subject: MVS TCP/IP failures - possible resolution
From: Daniel Thompson <thompsod AT USAA DOT COM>
Date: Wed, 25 Feb 1998 16:30:24 -0600
Folks,

There has been a lot of traffic recently on ANS4017E messages and other
symptoms of TCP/IP session failures when connecting to a MVS server.  We
found the following problem, obviously this may or may not be everyones
problem.

Our network team placed sniffers at the client and server (MVS) end of the
networks.  We found that at certain times in the evening, the MVS TCP layer
would report a sliding window size of 0 to the client.  This is normal,
particularly under heavy loads.  The client would acknowledge this data.
Here is where things seem to go south, the MVS TCP layer then sends a
message indicating a tcp sliding window size of 65535.  We believe this is
an invalid packet and the TCP layer is intending to flag a size of 0,
again.  However, the client end accepts this as a valid packet and attempts
to send data, to which the MVS tcp layer responds with a sliding window
size of 65535.  This back-n-forth continues until someone drops the
session.

We have seen this in NT and AIX clients.

Our MVS IP support personnel applied PTF UQ14365 for APAR PQ12577.  This
PTF was not production, last week.  However, since it addresses this
problem only, we decided to apply it.

I hope this helps,
Daniel Thompson
Storage/Automation Management - USAA
thompsod AT usaa DOT com
<Prev in Thread] Current Thread [Next in Thread>