Newsgroups: comp.parallel.mpi
From: zz@dart.cc.purdue.edu (Tony Z.)
Subject: p4_error: create_rm_processes: more slaves than msg queues
Organization: Purdue University, West Lafayette, IN
Date: 2 Aug 1997 00:21:09 -0500
Message-ID: <5rug45$1106@dart.cc.purdue.edu>

Hi,  could any one help me with the following problem?

I run a job on a set of SGI workstations with Mpich implementation.
When I run the job on 2 SGIs, the job works fine and completes. But
if I run it on 3 or more workstations, I get teh following error
messages:

========================
p0_10401:  p4_error: net_recv recv:  EOF on socket: 732
rm_6106:  p4_error: create_rm_processes: more slaves than msg queues 
: 2
rm_3800:  p4_error: net_recv recv:  EOF on socket: 555364
rm_l_1_3801:  p4_error: interrupt SIGINT: 2
bm_list_10402:  p4_error: interrupt SIGINT: 2
P4 procgroup file is /home/zz/test/hosts4.sgi.

==========================

where file '/home/zz/test/hosts4.sgi' specifies 4 workstation machine
names and executable path(all machine share same file system), like this:

===========hosts4.sgi============
sgi1 0  /home/zz/test/a.out
sgi2 1  /home/zz/test/a.out
sgi3 2  /home/zz/test/a.out
sgi4 3  /home/zz/test/a.out
=================================

And I start the job from machine 'sgi1':
==============
	mpirun -p4pg /home/zz/test/hosts4.sgi /home/zz/test/a.out
=============
right after it starts, the above error messages come out and the program
is interrupted.

If I remove any of 2 machines from the last 3 in the host list file
'hosts4.sgi', then run the above command, it just runs ok.


I tried an old version of mpich and the latest, the same problem.


If you have any idea of what problem it could be, please email me at
   zz@cc.purdue.edu


Thanks.

-Tony