Production status for Run10

Speaker : Gene Van Buren ( BNL )

Talk time : 12:35, Duration : 00:10

Following the You do not have access to view this node, as Software and Computing phone meeting. Some further details...


Production of all file streams was started and finished by August 20. Jobs with excessively long running time were killed and their production postponed. These represent approximately 7.5% of the jobs, but 35% of the data [44M out of 127M] as there is a increased chance of getting "stuck" in the longer jobs. This is likely the same issue brought up in RT ticket 1989, to be discussed Tracking problem discussion, RT 1989.


Production of st_physics is essentially done (only ~20 jobs are still running) today. 53M (out of a possible 56M) successfully produced (less than 3M events, or less than 5% of the data, affected by the excessive running time issue).


Priority productions on RCF, excluding (for now) the hung (very slow) jobs:

  1. AuAu11 stream data
  2. CuCu MuDst reproduction
  3. AuAu39 "phase I", reproduction with updated TOF calibs
  4. AuAu200 priority runs (final list not yet provided)
  5. You do not have access to view this node

Other Run 10 data has lower production priority. Notes:

  • (a) probably requires a Effect of stream data on database performance, a 2010 study, but we must be running Run 10 data productions only on the farm
  • (c,d) are also candidates for a Run 10 DB snapshot
  • (b) is big enough that it needs a significant fraction of the farm
  • (e) is about half done, and will take until the end of this week using about 200-300 job slots at a time
  • (c,d) are contingent upon a "go" from calibrations (TOF & TPC), though (c) is You do not have access to view this node