Page History

...

Job failure rate on the batch farm was 40% (3500 total jobs submitted, one per raw EVIO file).

Half of those failures was were due to filesystem errors and nothing to do with HPS.  These we were dominantly early runs (before SVT went to 0.5 mm) due to submission ordering of job submissions.

The other half were due to hps-java with runs before SVT bias was automatically recorded.  Those runs are not going to be very useful for SVT analysis anyway and require a new jar to fix.

Given these complications (and more), decision was made to let pass2 be and only resubmit failed jobs upon request.

...