waitSurprisingly simple, eh? Even the defaults seem ok. See man wait.
Run multiple processes on the same node using "&" in the usual way then wait (in the same shell script file as the &) for them to complete (or fail etc.) May need to adjust -n parameter to LSF bsub to increase the number of CPU cores it allocates.
Use module avail to see other packages that module can load for you.
limit coredumpsize 0(With unix bash use ulimit -c)
By default failing runs can generate huge files containing core dumps. Often the error message give a good enough idea what went wrong so telling Unix not to generate core dumps saves time.
bpeek 72890[1] bpeek: No match.This is an error with the unix shell command line interpreter rather than bpeek. The work around is to simply wrap the argument to bpeek inside double quotes:
bpeek "72890[1]"
HOST LSB_BATCH_JID=70590[1] LSB_JOBID=70590 LSB_JOBNAME=allgenes[1-1] LSB_HOSTS= LSB_JOBFILENAME LSB_JOBINDEX=55 LSB_JOBINDEX_STEP=1 LSB_JOBINDEX_END=100 LSB_OUTPUTFILE=xtrain2nd_1
Article in SIGEvolution.
(Slides).
Technical report on using initial experiences of Emerald
RN/12/08.