I prepared a small number of scripts to ease the submission of jobs on 
lxplus

Not really straightforward for the moment, but...


1) take ORCA_6_0_2
2) take the tag TkPsi_after_pre2 of Tracker and TestBeams
3) IF NEEDED, take from COBRA_6_0_3 Utilities/CHBook4 and modify 
CHObject.h . Sorry, there is nothing I can do there, I have no rights on 
that area...
4) compile libs and executables
5) go to .....TkPSIMay2002/test

There you have 3 files:
batchR2.job
batchTT6.job
batch.job

You have to:
1) modify the batchXXX.jobs you need
    basically, only modify the variables
exe_pg  (the full path to the executable)
scram_dir (where to set scram)
files_dir (location where to find the SIM_X etc)
params (parameters for the job, default = process all the events)

Note that the dirs must be reachable from the batch node, so I really 
advice about using /afs

2)  use batch.job
In it you have to modify only the queue (1nh); but I guess it can stay 
like this. At PSI 1 job processing 1 file needed ~ 3 min, but on the 
slower lxbatch cluster it needs more than the 8 min available in the 8nm 
queue.
So, decide the range of runs to be analyzed, the launcher and the 
destination directory, which can be reached with rfcp

Something like
batch.job 40010 40030 batchTT6.job lxcms68:/tmp/.

will submit the runs 40010-40030 using the default 1nh queue, using 
batchTT6.job as the launcher and (rf)copying the result on the temp disk 
of lxcms68 (to do so, you need to have the same (afs) account id on the 
lxcms68, which is available till long ago for all the people involved in 
the TB, even if you didn't know...)

Hope it is reasonable... My experience is that you can have ~20 jobs 
running at the same time; but since you are slower, and you have to wait 
for castor etc etc I am not sure it is really faster. It will 
nevertheless be useful in the future, when the server will be off, and 
for casual users which we do not want to host on the server...

        Tommaso