Newsgroups: comp.parallel.pvm
From: schiotz@oersted (Jakob Schiotz)
Reply-To: schiotz@fysik.dtu.dk
Subject: Re: pvmgs: possible BUG ?!
Organization: Physics Department, Techn. Univ. of Denmark
Date: 26 Sep 1994 12:54:30 GMT
Message-ID: <366ga6$1r4@news.uni-c.dk>

Daniel COHEN-LAROQUE (cohen@eurecom.fr) wrote:
: Hello PVMers,

: I am experiencing strange problems with pvm 3.3.3 on my
: Linux station. I wrote a master-slave style program wich makes
: use of group.

: The first run is ok, of course pvmgs is automatically spawned by the first
: task which registers a group. the second run crashes if I don't reset
: pvm from the console.(that means kill pvmgs)

: The problem is resolved if I manually spawn pvmgs from the console before
: executing the same programs. The I can run it as much as I want.

: WHERE DOES IT COME FROM ???
: Of course, this is not very important, but I'd like to know if it's commonly
: observed with other platforms.

I have NOT observed this problem running on Sun SPARCS, IBM RS6000 or
TMC CM-5E. But I observed one difference between first and second
run. Suppose I have 6 hosts, called h1 to h6, and I start a PVM
program using groups on h1. The programs joins the group and then
spawns 4 slaves. First time I run it I get pvmgs on h2 and the slaves
on h3,h4,h5,h6 but second time, when pvmgs is already running, I get
the slaves on h2,h3,h4,h5. Maybe your program is crashing because of a
problem on the host where pvmgs is started (h2 in my example). Then
when you run the program the second time you run the slave on the
"problematic" host.

Just a wild guess.

Jakob


--
Jakob Schiotz                    !  The noble art of losing face
Physics Department               !  May someday save the human race
Technical University of Denmark  !  And turn into eternal merit
DK-2800 Lyngby, Denmark          !  What weaker mind would call disgrace
schiotz@fysik.dtu.dk             !                  - Piet Hein
              ^^^ Note: new domain

