Matthew, Nate, Craig, Pavan, Michael, Dominick, Nick, Susan, Satish, Chris, Ryan, Gavin, Kanika, Paola...
- The production session in the upcoming collaboration meeting will consist of one parallel session as usual. This will be a working session similar to a normal meeting. Matthew will give the plenary talk.
- Debug mode testing has been included in the FA testing. As expected a factor of around 3 slowdown is seen in some modes. It is less in others. For those interested a maxopt test of FA14-10-03 can be seen here and the same test repeated in debug mode can be seen here.
- At some point over the last week bad channel appeared for the ND. It was pointed out that we should version control these from the software so that results don't change over time.
Action Item: Matthew will follow up with Jon about versioning bad channels in our code.
First analysis simulation (Nate)¶
Over the weekend Nate submitted all of the FD genie files and all of the ND rock singles, as of the meeting all but 1,600 of the FD genie files are done. In detail these are: the FD random run samples in FHC, RHC, in the 3 flavour modes and in full 14 DB mode. Details will be announced as the files are finished.
Nate observed that the FD cry jobs exhibit a bad failure rate of around 30%, he is trying to understand why this is happening via scrapping log files. Preliminary results indicate that only 260 of these were actual art failures, leading to the conclusion that the remainder are SAM related.
Action Item: Nate will continue to investigate this.
After some discussion it was decided that the best course of action for the ND genie files would be to follow a "random runs" method as was done for FD. That is to say that we generate our sample to match the real configurations seen in data with the same PoT weighting. In order to do this and produce a sample of sufficient size we'll need to make more that 100 sub-runs per run (as there are only 25 good ND runs at the moment). This apparently only requires a simple change to make_sim_fcl. Ryan will implement this. Ryan will also follow up with Bernali and the DQ group to confirm that the appropriate run summary tables are available.
Nate further noted that there currently exist FD FHC genie non-swap files in two versions, v1 are test files made with v05 flux window and version 2 are the correct (flux v06) ones. He has datasets defined that include both types. He will remove it.
First analysis reco validation (Gavin)¶
Gavin has produced reco validation files from Nate's draining SIM datasets for validation. He ran reco cosmic and reco-genie on these files. He noted that we may not need the cosmic veto on the ND cosmic files but that is for the reco group to say. He also ran the reco validation histogram analysis on these files and put these files on a blue arc location. This information has been communicated to Peter.
He is also in the process of passing on the knowledge of how to run to Bruno who will help out with future productions.
All jobs were run onsite.
Keep-up pclist validation (Paola)¶
After some discussion it was decided that Evan's change that is needed for the pc-list validation requires a new release (as it is not simply a configuration change). Matthew will communicate this to Jonathan. When a new snapshot is available Paola can re-run and produce files for Evan, Evan will then need to sign off on these before keep-up can begin.
Tools for systematic sample production (Nate)¶
Last Friday Nate committed fcl files to the production scripts directory which drop the daq, photon transport and geant products and then re-run these based on whatever we want. This will be used, for instance, to put in a new geometry. However current Nate has no scripts to run the daq to daq step. Nate will continue to work on this and will produce a geo-jittered sample of the FA 14DB FD FHC swap & non-swap samples.
Large file renaming due to data logger problem (Jeny)¶
Postponed until Thursday
- There was some discussion about the procedure for routinely producing tags. An idea was floated about whether we could make snapshots automatically? Say on a Monday with an “R” prefix to differentiate them from the "S" snapshots? Can we then alias this to “S” if we need it? What about CVMFS, any tag for PID needs to be in CVMFS keep up should also be there - what about dropping the debug build? no slf5? This discussion will be influenced by the available disk space on CVMFS.
- Dominick will re-make the ND reco keep-up with some 0-th order calibration.
- Dominick will rename ProductionScripts to Production and ProductionScripts/fhicl to Production/fcl this week.