LSF Version 7.3 - Using Platform LSF HPC
Configuring LSF to Work with MPICH2
Make sure MPICH2 commands are in the PATH environment variable. MPICH2
commands include
mpiexec, mpd, mpdboot, mpdtrace, and mpdexit.
For example:
[174]- which mpiexec /pcc/app/mpich2/kernel2.4-glibc2.3-x86/bin/mpiexec
Add an mpich2 boolean resource to the $LSF_ENVDIR/lsf.shared file.
For example:
hmmer Boolean () () (hmmer availability)
lammpi Boolean () () (lam-mpi available host)
mpich2 Boolean () () (mpich2 available host) <====
End Resource
Add mpich2 to each host that an mpich2 parallel job may run on.
For example:
Begin Host
HOSTNAME model type server r1m mem swp RESOURCES #Keywords
qat20 ! ! 1 3.5 () () (mpich2)
qat21 ! ! 1 3.5 () () (mpich2)
qat22 ! ! 1 3.5 () () (mpich2)
End Host
Run lsadmin reconfig and badmin mbdrestart as root.
Run lshosts to confirm that an mpich2 resource is configured on all hosts on
which you would like to run mpich2 parallel jobs.
For example:
[173]- lshosts
HOST_NAME type model cpuf ncpus maxmem maxswp server RESOURCES
qat20 LINUX86 PC1133 23.1 1 310M - Yes (mpich2)
qat21.lsf.p LINUX86 PC1133 23.1 1 311M 635M Yes (mpich2)
qat22.lsf.p UNKNOWN UNKNOWN_ 1.0 - - - Yes (mpich2)
Configure and start an MPD ring.
If you want to start an MPD ring per job, this is the default and recommended
mechanism, and you do not need to do any extra configuration.
If you want to start an MPD ring for all users, use the mpdboot command as
root on all machines.
To check if mpdboot ran successfully, use the
mpdtrace command
[root@qat20 test]# mpdtrace -l
qat20_37272
qat21_52535
For MPICH2 1.0.3 only, add the following lines to $HOME/.mpd.conf for
all users.
[61]- cat .mpd.conf
MPD_USE_ROOT_MPD=Y <==========
secretword=123579a