[GE users] FW: SGE Does Something Bad to rsh

Rayson Ho raysonho at eseenet.com
Fri Apr 22 23:21:30 BST 2005


    [ The following text is in the "iso-8859-1" character set. ]
    [ Your display is set for the "ISO-8859-10" character set.  ]
    [ Some special characters may be displayed incorrectly. ]

>We're using SGE 5.3p6 -- we should upgrade to SGE 6.03u soon.

update 4 will be released soon :)

>Somehow, SGE is using ssh instead of rsh.

I think the application is using /usr/bin/rsh, which is different than the
one shipped with SGE, and thus it is not accepting the -p option.

Can you use the command sequence from ("What rsh command is compiled into a
binary?") to determine what is compiled into the application??

http://gridengine.sunsource.net/howto/mpich-integration.html

If the application is linked against MPICH lib dynamically (use ldd to
check that), then you should replace "programname" with the dynamic MPICH
lib.

Rayson


> Can someone tell me what the (global) SGE
>config 
>should look like to use rsh?   
>
>Thanks
>
>-----Original Message-----
>From: Rossi, Angelo
>To: Research Technology Fellows
>Sent: 4/22/2005 4:28 PM
>Subject: SGE Does Something Bad to rsh
>
>Hello My Fellow Tech Fellows:
>
>Here is what I did:
>
>1) I recompiled the mpich library using the pgi compiler so that it uses
>rsh
>instead of ssh.
>
>2) I compiled a test program that uses mpi (the pingpong program) and
>linked
>it to the new mpich library.  Using the following command:
>
>/home/m254/rossi/mpich/bin/mpirun -nolocal -np 2 -machinefile
>./machinefile
>pingpongf90.x
>
>THIS WORKS BEAUTIFULLY.  NO PASSWORDS REQUIRED!!!! 
>
>"-nolocal" spawns parallel tasks on remote nodes defined in machinefile.
>
>3) Now try the same program using the sge batch engine.  Here is the
>batch
>script:
>
>#!/bin/bash
>#
>#$ -pe mpich 2
>#$ -cwd
>#$ -N pingpongf90
>#$ -j y
>#$ -S /bin/bash
>#
> 
>
>/home/m254/rossi/mpich/bin/mpirun -np $NSLOTS -machinefile
>$TMPDIR/machines
>./pingpongf90.x >pingpongf90.out
>
>IT FAILS.
>
>Here is what is in the error file:
>
>/usr/bin/rsh: invalid option -- p
>usage: rsh [-nd] [-l login] host [command]
>/home/m254/rossi/mpich/bin/mpirun: line 1: 18306 Broken pipe
>/home/m254/rossi/parallel_examples/./pingpongf90.x -p4pg
>/home/m254/rossi/parallel_examples/PI18222 -p4wd
>/home/m254/rossi/parallel_examples
>
>HELP!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
>
>Regards,
>
>Angelo
---------------------------------------------------------
Get your FREE E-mail account at http://www.eseenet.com !

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net




More information about the gridengine-users mailing list