Need default value for JANA_RESOURCE_DIR for farm jobs

[faustus123]

This is prompted by a set of farm jobs that hammered the halldweb server with downloads of the magnetic field map. Farm jobs should be getting this and other resources from /u/group/halld/www/halldweb/html/resources . Normally, this is controlled with the JANA_RESOURCE_DIR environment variable or configuration parameter. We should check if JANA_RESOURCE_DIR is not set and if so, check if /u/group/halld/www/halldweb/html/resources and if also true set it. This will avoid this issue which is really only a big problem for big farm jobs.

[sdobbs]

I've pushed a change of the type you've described in the branch "resource_dir_fix".
One thing to consider that it's not only a problem for farm jobs, but also for people running batch jobs off-site, if they've not properly set JANA_RESOURCE_DIR. But I'm not sure if a more general fix is feasible.

[markito3]

The problem with many requests on the farm goes like this: the default location of the local cache of files is /tmp (if I remember right). On the farm each node has its own /tmp, so the cache getting filled does not help for the next job which is in general on another node.

Normally, on a single computer, /tmp is just fine. So the solution is to sense whether the user is sharing disk with others interested in these files, and if so, putting them into an agreed upon location. But that is pretty hard to sense, and there may be no agreement, thus the environment variable.

We could issue a warning: JANA_RESOURCE_DIR not set. Try /group... at JLab. A warning would not be visible to the user during farm execution, but would hopefully be noticed during set-up and initial testing of the farm jobs.

Being at JLab, and knowing the correct location and seeing as how the problem may happen commonly, it is tempting to just set it (at JLab). It is hard to argue against that approach. I will put the environment into the gluex_env_jlab script at any rate. That is only helps; it is not a solution.