If you want to run over some rarely used files, it is very likely that they have to be first staged. While running this is done in sequence, so the file is only staged at the moment the algorithm wants to use it. This is of course very time consuming. But there is a possibility to pre-stage the files manually and in parallel.
Create first a list of the castor file names advanced by -M and store it in variable. So if you have the two files
/castor/cern.ch/grid/lhcb/data/2011/RAW/FULL/LHCb/CALIBRATION11/86753/086753_0000000001.raw /castor/cern.ch/grid/lhcb/data/2011/RAW/FULL/LHCb/CALIBRATION11/86753/086753_0000000002.raw
and you store them in $FILELIST
echo $FILELIST
should give
-M /castor/cern.ch/grid/lhcb/data/2011/RAW/FULL/LHCb/CALIBRATION11/86753/086753_0000000001.raw -M /castor/cern.ch/grid/lhcb/data/2011/RAW/FULL/LHCb/CALIBRATION11/86753/086753_0000000002.raw
. Afterwards stage them by
stager_get -U <stagejobname> $FILELIST
where <stagejobname> is an arbitrary name use to later control the staging process.
To look if the files are already staged, type
stager_qry -U <stagejobname>
if the staging is still going on the file is marked with STAGEIN, if it is staged it is marked with STAGED.