I'm running 7.1.2. lamboot with -prefix option. The local machine
spawns the lam daemon ok but the remote machine can't find the
liblam.so.0. The lam binaries are fully qualified so I believe I don't
have to set PATH to point to the LAM executables. I've also tried
exporting LAMHOME in .profile but that it still can't find liblam.so.0.
I thought the -prefix option in lamboot will pass the install path to
hboot to boot remotely....
This is my lamboot command:
/afs/tda/sti/r620/prod/linux24/tools/tb/lam/bin/lamboot -d -prefix
/afs/tda/sti/r620/prod/linux24/tools/tb/lam
/afs/tda/home/yamend/20686.mynodes
This is the output of debug option during lamboot ... trying to boot on
remote node:
n-1<20811> ssi:boot:rsh: attempting to execute: /usr/afsws/bin/rsh
158.140.144.47 -n '( ! [ -e ./.profile ] || . ./.profile;'
/afs/tda/sti/r620/prod/linux24/tools/tb/lam/bin/hboot -t -c
/afs/tda/sti/r620/prod/linux24/tools/tb/lam/etc/lam-conf.lamd -d -s -I
'"-H 158.140.144.43 -P 35978 -n 1 -o 0"' -prefix
/afs/tda/sti/r620/prod/linux24/tools/tb/lam )
ERROR: LAM/MPI unexpectedly received the following on stderr:
/afs/tda/sti/r620/prod/linux24/tools/tb/lam/bin/hboot: error while
loading shared libraries: liblam.so.0: cannot open shared object file:
No such file or directory
What can I do to have remote nodes boot using the prefix directory?
Also, how is LAM_DEPENDENCY_PREFIX define used? If I copy the output of
compile into the directory pointed to by this define, the remote node
boots (by using that path to find liblam.so.0). LAM_DEPENDENCY_PREFIX's
path is the -prefix option I used during configure step.
Thanks,
YoungHui
|
