...
- 2/20/2014 - Initial set-up of task
- 2/25/2014 - Begin trickleStream on Block 1 – entire P202 run list
- 2/28/2014 - Many problems: LSF seems to bunch up a large number of jobs before dispatching. This causes a shock wave of jobs causing a problem for AFS (reading the ~500 MB diffuse response file), and for the dataCatalog (the queries for FT2 file, and to determine the next file version for each data product). Therefore, reconfigure trickleStream to a mere dribble – no more than 300-400 jobs typically running simultaneously. This means a 1-day task has become a week-long effort.
- 3/3/2014 - Block 1 complete. Last week developed new scheme to move all dataCatalog queries from batch jobs into jython scriptlets. These changes will be integrated prior to processing any more data. Validity check: There are the same number of runs and the same number of events in the reprocessed P203 data as in the P202 data.
310,326,817 events in 29158 files. The run range for block 1 (== entire range of P202) is 239557414 through 405329691.
Configuration
Identical with P202-FITS except:
...
Note that diffuse response is calculated for 'source' and 'clean' event classes only.
Timing
- Each run requires approx 20-30 minutes of CPU time, depending on the machine-class being used. However, due to AFS and dataCatalog issues, block 1 running was restricted to ~500 or fewer jobs at a time.
...