[Halld-offline] Gluex osg production environment ready for testing

[Richard Jones]

Hello Mark and all,

After a number of interruptions, I finally finished the roll-out of a
container-based osg production framework for Gluex simulations. Please have
a look at the git project linked below. It contains two scripts that give
everything you need to try it. This is bound to be somewhat rough at first,
but let me know how it goes.

https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_rjones30_gluex-2Dosg-2Djobscripts.git&d=DwIBaQ&c=lz9TcOasaINaaC3U7FbMev2lsutwpI4--09aP8Lu18s&r=oyKV5joTkJsuRYv6hh48IMTw3i-IrYD-ZUAHHU0DdAY&m=oiF6sdfNBs8gUORck4A6M8slPAM9Dm7vn6Wy09F4Exs&s=cE6Q1E0JEk-3DhaAvGK0LfyCxH6S_ZskRDSwT6jZfXI&e= 

The sim-recon build that is configured for osg production right now is
based on version.2.11.xml from the http://halldweb.jlab.org/dist area,
which looks like the latest stable release. Still, it is not very stable.
This is probably because of the fast development over the past few months
of calibrations for real data, and the lack of corresponding exercising of
the code on Monte Carlo. A minimal checkout and run of the distro job
script submits a job with 40 slices of 250 events each generated with
bggen, simulated with hdgeant4, passed through hd_root for REST production,
and through hd_root again for plugins analysis. When I tested it, only one
of the 40 jobs crashed:

*** Error in `hd_root': free(): invalid pointer: 0x00002baa1c394ff8 ***

with 37 completing successfully, and 2 others bombing out of hd_root for
other reasons. At this point, I would like to pass things back to the
reconstruction/analysis gurus. I could work on diagnosing and fixing these
bugs, but as I discovered with the "hd_root hanging" bug last week, there
is a good chance that some of these bugs are known and fixed in somedev dev
branch of the sim_recon tree or jana repo.

Please have a look and try things out. Let me know if/when you run into
issues. I will update the HOWTO on the wiki once we get a little further
into the learning curve. Right now, you can rely on the README in the git
project linked above for docs.

Richard
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.jlab.org/pipermail/halld-offline/attachments/20170626/83cca4ea/attachment.html>