Matthew, Joseph, Paul, David, Chris, Jonathan, Susan, Nick, Gavin
- New dataset request: FD cosmics matching FD NuMI run list. Tian from nu_e estimates that they'll need around ~7 subruns (why don’t we start with 10). This is the same number of files as Joseph ran for NuMI so if we run ~1/4th and the bottle neck is the FTS then it should take ~1/4th the time (but will need more nodes). Subruns 0-10 including top up is 36k files. Need to check on the raw2root status of these.
- New dataset request: ND genie MC with varying PoT per pulse - make_sim_fcl switches in place? Ping Ryan
- New dataset request: Calibration shifted files - need fcls.
Anything else anyone knows of? No
- David will tag tomorrow
Epoch discussion (all)¶
No documentation so far. Need to formalise this. Initially use letter for analysis, but what is important is when things change as part of the running. Top-up etc can be forgotten in time as they are arbitrary. Joseph will email the group again and start a wiki
FTS optimisation (Chris)¶
Chris brought up a new VM novasamgpvm04 on Friday to replace gpcf028. He and Dominick are working with Robert to test the meta-data changes.
Run history bug (Matthew)¶
Still working on getting test files running. FNAL ones work, FZU ones died. Draining set had weird errors. Will continue to validate his outputs and then progress to a full sample (in parallel with debugging offsite).
- All MC done.
- Joseph reports that non-top-up reco done except for files with crashes as is the PID done. Top-up reco is running (~50%), he’ll then run PIDpart and estimates all that will be done in 3-5 days. The plan is to then start cosmics. He has done this with a new fcl that essentially doesn't make the pid-part files. He'll commit this.
Not much going on. Will re-do ND data as only ~60% of that seems to have shown up in SAM.
- Bruno will start running on the ND data. * Gavin reports that the ideal conditions mix-caf is all done.
ND genie RW (Gavin)¶
Going slow, FZU only getting 200 nodes. Running a bit quicker than he expected, ~6h, 5 days on 1k nodes