Dear SWAN team,
I discovered that EOS is not mounted on spark clusters, thus, using eos files with spark one must have to add (for the COMPASS experiment) " root://eoscompass.cern.ch/" as prefix for all file paths inside the eos/experiment/compass folder.
This works like a charm, however I had recently to move from my experiment folder to my cernbox. I tried to add the prefix “root://eosuser.cern.ch/” this time, but this does not work. I also changed my $EOS_MGM_URL variable (if that has to do something with it) but same issue.
[vbrian@lxplus779 vbrian]$ root -l root://eosuser.cern.ch//eos/home-v/vbrian/outputs/dvcs2016-P09slot5.1-08October/SWAN_WS/dvcs2016-P09slot5.1-08October.root root  Attaching file root://eosuser.cern.ch//eos/home-v/vbrian/outputs/dvcs2016-P09slot5.1-08October/SWAN_WS/dvcs2016-P09slot5.1-08October.root as _file0... Error in <TNetXNGFile::Open>: [ERROR] Server responded with an error:  Unable to open file /eos/home-v/vbrian/outputs/dvcs2016-P09slot5.1-08October/SWAN_WS/dvcs2016-P09slot5.1-08October.root; No such file or directory (TFile *) nullptr
Then my question is the following, what prefix should I use then, or what variable should I set up in order to access files on my cernbox, using spark clusters ?
NB. I use the PyRDF module, and removing the prefix and using the local backend (not spark) it obviously works. When using the spark backend, nothing works anymore besides working on my experiment folder.