Newsgroups: comp.parallel.mpi
From: rdaoud@magnus.acs.ohio-state.edu (Raja B Daoud)
Subject: Re: Running MPI without rsh/rexec access
Organization: Ohio Supercomputer Center
Date: 24 May 1995 22:11:03 GMT
Message-ID: <3q0atn$o8o@charm.magnus.acs.ohio-state.edu>

Stuart D Blackburn  <sdblackb@uncc.edu> wrote:
>since earlier this spring. It is a pain to do because I don't have rsh
>or rexec access to machines, only rlogin. PVM does provide a work around
>for this by allowing you to specify a manual start of pvmd on your remote
>machines.

This can be done in LAM, though we don't encourage users to hand-boot
the multicomputer (this is how we debug the booting tools :-) ).
Instead of using the "lamboot" and "wipe" tools, the user would have
to login to each machine and start the lam daemon (using "hboot") there,
and later, login to each machine and kill it (using "tkill") locally.

To boot a 3-node system (e.g. tbag, alex, hawkeye), type on each
machine respectively:

{tbag}% hboot -vc conf.lam -R "-n0 -o0 -N3" -I "alex 1 hawkeye 2"

{alex}% hboot -vc conf.lam -R "-n1 -o0 -N3" -I "tbag 0 hawkeye 2"

{hawkeye}% hboot -vc conf.lam -R "-n2 -o0 -N3" -I "tbag 0 alex 1"

This tells each host that there are 3 nodes (-N3), gives each one
a nodeid (-nX), and lists the other host names and their nodeids (-I).

To terminate LAM, type "tkill -v" on each machine.

See the "LAM for C programmers" doc, page 58, "Low Level Booting".
If you have questions, contact us at: lam@tbag.osc.edu

--Raja

-=-
Raja Daoud				raja@tbag.osc.edu
Ohio Supercomputer Center		http://www.osc.edu/lam.html
-=-
MPI users conference, June 22-23 @ Notre Dame.

