search

elaird / supy

analyze events stored in TTrees in parallel
8 stars 7 forks source link

Missing Events at FNAL #197

Open betchart opened 9 years ago

betchart commented 9 years ago

This may not be a supy bug, but sometimes when I run an analysis on a dataset via the FNAL batch system, the resulting supy files are missing some events; it is as if they were never passed through the TChain, and no error is generated. The only protection I have against this problem is manual checking of the output. It would be safer to have the option to put the number of expected events in the dataset description, and a related option to ask supy to check that the correct number of events were processed.

betchart commented 9 years ago

This issue can be caught by specifying an expected number of events, new in #198 . A common difference between expected and actual number is 100000, pointing to a probable explanation that whole files can be skipped, perhaps if they fail to open. Such silent failure could be causing problems for other analyses if it is a ROOT or pyROOT issue.