From: Alberto Scotti <ascotti@whoi.edu>
Newsgroups: comp.parallel.pvm
Subject: PVM crashes
Date: Wed, 07 Apr 1999 11:01:53 -0400
Organization: Woods Hole Oceanographic Institution
Message-Id: <370B7361.8B3F24F4@whoi.edu>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Xref: ukc comp.parallel.pvm:8220


I am running PVM on a cluster of IBM dual pentium boards
running Linux 2.2.2.
The eight machines are connected by a 3COM hub (the system was set up
with spare parts...) and a ninth machine works as a gateway with 
the outside world. 
I am using PVM 3.3.11 at the moment. 
I am running an application which makes a quite intensive use of
the network. What happens is that after a while, suddenly one of
nodes ceases to respond. Here's the fact
1) It happens randomly, after about 24 hrs of operation.
2) The problem affects all nodes.
3) When a node crashes, it either dies without any message wathsoever
or can still be pinged but refuses logins.
Note that because of the hub, the number of collision is quite large.
Any help will be greatly appreciated.

-- 
**************************************************************
Alberto Scotti				Tel:508-289-2914
Dept. of Physical Oceanography		Fax:508-457-2181
MS 21					Email:ascotti@whoi.edu
WHOI
Woods Hole, MA 02543-1541