Minutes from 27 May 2009

2009/05/27 CSEP Meeting

Attendees: Phil M., John Y., Masha L

1. csep-op:
   • Should enable backup copy to scec-02
   • Matlab failure due to missing .matlab/R2007a/history.m file:
     1. John found a switch to disable history that should be placed in ~/.matlab/R2007a/matlab.prf file:
        • CommandHistoryHistoryOptionsR14=I0
     2. Masha will update Trac page https://intensity.usc.edu/trac/csep/wiki/SetUpTestingCenter with configuration option, so other testing centers are aware:
        • Under 'Configuration Files' paragraph
   • DNS servers at USC failure that prevented publishing of results for California one-day models for test date of 2009-04-23:
     1. John updated /etc/hosts file to prevent such failure in the future
     2. Masha re-processed the test date successfully
   • Updated to V9.4.1:
     1. Unit and acceptance tests succeeded
     2. Failure to re-process due to the Trac ticket #152: Re-processing with --disableExistingDataStaging option fails if links to the existing observation catalog file exist:
        • Majority of votes to go for V9.4.2 release with the fix
   • V9.4.2 release candidate was installed on csep-cert as of Friday, May 22, 2009:
     1. Includes fix for Trac ticket #152 and new acceptance test for the fix
     2. Passed all unit and acceptance tests
     3. Masha copied some of operational data for failed reprocessed test date to csep-cert, reprocessed successfully
     4. Observed slowdown in R-test due to implicit import of scipy.stats by cseprandom.py module:
        • Fixed in SVN
        • Nightly build on csep-cert succeeded
        • Nightly Dispatcher's processed OK and were faster
     5. V9.4.2 to be installed today
2. Matt reported 5 failed acceptance tests for installed release candidate V9.4.1:
   • John traced down the problem to QuakePy which sets backend to 'PS':
     1. Fabian recommended to disable QuakePy plotting by commenting out some module import's
     2. Fixed the problem
3. Tasks for CSEP V9.7 release:
   • Peter Shebalin's EAST three-month alarm-based model for California
     1. Masha need to re-create EAST forecast on csep-cert since it was changed to use M>=4 version of the model
   • Alvaro Gonzalez's one-day alarm-based model for California
     1. To be part of CSEP V9.10.0
   • Yan's Global long-term and short-term models using CMT:
     1. Yan's forecast is using 0.5 degree grid vs. CSEP 1.0 degree grid:
        • Discussed what changes are needed to get model under test
        • Estimated about 1 week of Masha's time
   • Trac ticket #148: Introduce acceptance tests for the viewer of testing center results
     1. John and Masha are working on setting up an initial set of test cases using ClientForm.py, HTML parser (BeautifulSoup, may be) and Python's unittest testing framework
4. Jeremy distributed mini_CSEP
   • Danijel suggested to re-write Matlab evaluation codes in Python:
     1. Danijel will survey the group about the need for such distribution
5. Hardware:
   • John built new csep-cert, will need to install CruiseControl before switching the server
   • csep-devel already uses new hardware, will need to update packages