[Halld-offline] Containers Meeting Notes, April 13, 2018
Mark Ito
marki at jlab.org
Fri Apr 13 14:00:56 EDT 2018
Folks,
Please find the notes below and at
https://halldweb.jlab.org/wiki/index.php/GlueX_Containers_Meeting,_April_13,_2018
.
-- Mark
_______________
GlueX Containers Meeting, April 13, 2018, Notes
Present: Richard Jones, Mark Ito, Thomas Britton, Sean Dobbs, Zisis
Papandreou
* At Sean's suggestion, Mark wrote a HOWTO
<https://halldweb.jlab.org/wiki/index.php/HOWTO_use_the_GlueX_Singularity_Container>
on...well...how to use the current GlueX Singularity container to
access recent builds of the software without having to do the build.
o The current version *assumes* no CVMFS partition on the local
node, despite this being the recommended method of accessing the
software...patience y'all!
o Mark will note that the location of "/group/halld" on Oasis is
/cvmfs/oasis.opensciencegrid.org/gluex/group/halld in a future
version of the HOWTO.
* There are two versions of the container at present:
1. for when CVMFS is present (i. e.,
/cvmfs/singularity.opensciencegrid.org/markito3/gluex_docker_devel:latest)
2. for when CVMFS is not there
o This was only necessary because we have been invoking OSG jobs
in the mode where the user is automatically placed in the
container. If we invoke the container by hand, the bind option
can be controlled and two containers are no longer necessary.
Mark and Thomas will work on this.
* Mark has started looking at XROOTD
o building from source directly from a clone of the GitHub repository
o documentation on main site is a bit obtuse; pointers to
beginner-level docs appreciated
* The MC Thomas generated last week on the OSG using the GlueX
Singularity container has been used to demonstrate that the rate
dependence of reconstruction efficiency seen in the simulation
roughly matches that seen in the data
<https://halldweb.jlab.org/doc-private/DocDB/ShowDocument?docid=3583>.
* Thomas is gearing up to start a major bggen campaign on the OSG.
o Requested by the MIT group for their B boson analysis.
o Statistics to match 50% of the golden period of 2016.
o Will ship random triggers along with each job using the Condor
mechanism. If this proves to be a bottleneck, we can explore
using Richard's XROOTD servers at UConn for this purpose.
o Richard noted that statistics on the project would be
interesting, e. g. the number of OSG CPU hours consumed per unit
time as a function of time.
* Action items:
1. cvmfs bind instructions
2. go to self-serve singularity launch on OSG
3. fix issue on ifarm with singularity
Retrieved from
"https://halldweb.jlab.org/wiki/index.php?title=GlueX_Containers_Meeting,_April_13,_2018&oldid=86417"
* This page was last modified on 13 April 2018, at 13:58.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.jlab.org/pipermail/halld-offline/attachments/20180413/6bf204f0/attachment.html>
More information about the Halld-offline
mailing list