2015-11-30-MEMO-CSw¶

Offline:¶

• Tracker sign flip
  • Problem identified.
    • fibers were counted in the 'wrong' direction (wrt hardware design) in the reconstruction (code) & MC (description)
    • the two stereo-views had their rotations flipped in the geometry description
    • the rotation direction in geant goes the opposite way wrt reconstruction
  • Implementation & Testing in progress
    • Code:
      • Fix interpretation of mapping — counting of fibers to be consistent with h/w map
    • MC:
      • Fix fiber numbering in the geometry description (to be consistent with h/w and reconstruction)
      • handle geant-reco rotation difference in the gdml translation
    • C. Heidt has implemented
    • DR has re-processed a test run with the fixes, C.H is looking at the output
    • Fixes to go into production geometry & MAUS after testing

• Tracker momentum reconstruction
  • CHunt investigating: Energy loss handling in Kalman, field non-uniformity, field underestimation in MAUS

• Geometry
  • the EMR geometry has been fixed by RB (sensitive detector definitions in the GDML description)
    • FD has been working on structural improvements to the EMR reconstruction, will look at geometry next & validate

Infrastructure¶

• EPICS failover tests (partially) done
  • tested the failover of miceeccserv1
  • will need to be repeated with both backups once miceecserv nodes get static IPs

• CDB failover tested
  • procedure now documented
  • triggering the read-only-slave to become a read-write-master takes < 5 minutes.

• cooling channel currents for past runs now in CDB, so one can retrieve currents by run number
  • PMH has also started writing cooling channel information to CDB from RC now

• JM & DR to work on better automation and job submission/processing bookkeeping on offrec computer

C&M network reconfiguration¶

• Guideline: communication with critical devices should not go through any intermediate server
  • implies static IP addresses & avoiding nameserver resolution (communicate explicitly to a static IP )
  • lockout ssh connections into critical machines during running (except for superuser)
• requires
  • list of C&M machines -- with the networked devices they talk to (ACTION: PMH, PF)
  • configuration changes on C&M machines & networked h/w devices to assign a static IP (PMH, PF)
  • avoid nameserver resolution by listing names in local lookup table (/etc/hosts) -- sync has to be handled carefully (PF)
  • shared mounts to be referenced by IP, not name (PF)
  • code changes to communicate to devices by IP (not name) (PMH)
  • define PV to indicate running/testing mode (PMH), and script ssh-lockout based on PV (PF)
• PF documenting detailed procedure (in progress: https://micewww.pp.rl.ac.uk/projects/cm/wiki/Static_IP_implementation)
• Estimate ~40 hours (~20 machines + h/w devices), needs to be done when disruptions to C&M & network are acceptable