[Halld-offline] Offline Software Meeting Minutes, January 7, 2014
Mark M. Ito
marki at jlab.org
Thu Jan 8 08:31:00 EST 2015
Please find the minutes at
GlueX Offline Meeting, January 7, 2015
* *FIU*: Mahmoud Kamel
* *FSU*: Aristeidis Tsaris
* *JLab*: Mark Ito (chair), David Lawrence, Paul Mattione, Kei Moriya,
Nathan Sparks, Simon Taylor
* *NU*: Sean Dobbs
Review of Minutes from December 10
We looked over the minutes
of the last meeting.
* Kei is using the gxproj1 account for offline monitoring jobs and
Mark is using gxproj2 for Data Challenge 3 (DC3).
* Mark was able to get the stand-alone version of Dmitry's Run
Conditions Database (RCDB) running. He will circulate instructions
on how to do that. He will also approach the Computer Center on
installing it on halldweb1.
* Paul mentioned that he could not find documentation on the version
management system that Mark presented last time. Turns out it does
not exist yet. Paul suggested that the documentation be featured in
the "getting started" section of the wiki.
* Mark presented a new version
of the main "Offline Software" wiki page. It is still under
DAQ and TTab Plugins Converted into Libraries
David reviewed for us the email
he sent today. Now it is no longer necessary to specify these plugins as
JANA command options.
HDDM Versions and Backward Compatibility
We reviewed principles
that Richard proposed for future changes to REST format. The issue is
preservation of backward compatibility, being able to analyze old REST
data with new code. The scheme does the preservation at the cost of
complication in the element names for those that are changed from the
elements they replace. We did not decide if this approach should be
enshrined in policy, but will discuss it further in the future.
Commissioning Run Review
We reviewed tasks and progress from the recent run.
Offline Monitoring Report
Kei presented issues
from the offline monitoring reconstruction jobs during the run. The slides:
* Offline Monitoring Summary
* Disk Usage For Each Week
* 2-track Skim Output
* Number of Files Processed
* EVIO Statistics
* Errors found:
o DAQ plugin
o Too Many FDC Hits
o Insufficient Buffer Space
o Bad Alloc
o Mismatch in Trigger Bank
o Unknown Module Type
o F1TDC Block Header
* Looking Ahead
Commissioning Branch-to-Trunk Migration
Most of the development that went on during the run was checked into the
commissioning branch. A lot (but not all) of this code now has to be
moved the trunk, in particular those changes that improve the
reconstruction in general. Simon has been managing the commissioning
branch and will look into doing this transfer.
There was interest during the run for REST-formatted data for high-level
analysis. Kei has already produced these files (see his talk above).
They can be found in
A lot of recent analysis has been based on raw data with corrections
done on raw or reconstructed quantities to get better results. At
present these corrections are not reflected in the REST data. In order
to make future production of REST data more useful we need to move
corrections and calibrations into the standard reconstruction.
Sean will coordinate production/update of calibration constants and
capture of correction algorithms through the Calibration Working Group.
The goal is to make a push for some substantial progress and then to
re-make the REST data set.
Paul brought up the a private discussion we had on the email list on
"data reconstruction trains". The idea is to have a single set of jobs
do reconstruction on raw data and have several "cars" attached to the
jobs for specialized purposes. Each individual project would then avoid
having to fetch the data from tape and pay the CPU price of
reconstruction. However such an effort requires significant coordination
and management. We did not move to start designing a system right away
since (a) the need is only prospective at this point and (b) a lot of
the savings for high-level analysis can be achieved with useful
Storage of Data Taken between Runs
David pointed out that there will be cosmic and test running in the
coming months before the spring run and we need to decide where in the
tape directory hierarchy they should be stored. The consensus was to
simply create a new run period directory in parallel to
"RunPeriod-2014-10" (used to store data from the commissioning run just
ended), something like "RunPeriod-2015-01".
Mark M. Ito, Jefferson Lab, marki at jlab.org, (757)269-5295
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Halld-offline