[comp.os.vms] How to trace cause of broken pipe

ccvj@luxor.latrobe.edu.au (Vicki Jordan) (05/03/91)

How does one find out what breaks a pipe?

The situation in question is nnptd running on a Decstation 2100 (Ultrix
4.1).  The first log message refers to a connection where localhost is
engaged in an <IHAVE> dialogue with its server (a VMS machine, running
ANU-NEWS 6-03, Multinet).   Unfortunately, this means that the site which 
feeds me has to repeatedly re-connect until all items have been processed.  
The second message refers to a readnews session from a client (a VMS machine, 
running ANU-NEWS 6-01, CMU-TCP).

04:18:49 localhost: 10238 nntpd: xvax.ucs.unimelb.edu.au disconnect: Broken pipe
12:57:13 localhost: 12395 nntpd: lust1.latrobe.edu.au disconnect: Broken pipe

I have run nntp with full logging, but get no more info than the "disconnect: 
Broken pipe" message.

At first I thought it might be a resource problem on the Ultrix side,
but increasing the no of user processes (150) has not helped.  The only
other thing of interest is that no broken pipes are logged from any of
our Unix client machines (RISC/os 4.52, AUX, AIX) which run either rrn or xrn.

I know this would seem to point a finger at ANU-NEWS, however, the VMS
site which feeds me gets a a message indicating it is the Ultrix machine
which is at fault:

%MAIL-E-LOGLINK, error creating network link to nude LU
-SYSTEM-F-LINKEXIT, network partner exited

Any help would be appreciated.
Since I have cross-posted to so many groups, please email and I will
summarise responses (and -hopefully- the answer!).

-----------------------
Vicki Jordan
La Trobe University
Bundoora AUstralia 3083

ccvj@lure.latrobe.edu.au