Core Minutes 1/8/2013

• ScienceTools: No report today due to SeeVogh intransigence. See Science Tools Development Notes for latest ST news.

• Hardware and migration report (Tom G.)

  • Wilko has gained root access to the new lnx boxes so we can get started on the transition. Brian will come up with a list of services and possible distribution on the boxes. He suggested the possibility of running multiple VMs in a single box to isolate services from each other.
  • The new nfs servers (which had been xroot servers) have been very reliable so far. Are there any objections to shutting down the old sulky servers? (Richard) If we had to move L1, would it move to another wain? (Tom G.) Yes. (Warren) That would require a new L1 version (Richard) so a CCD, but it would be expedited.

• Pass 7/8 (Leon) is finishing up putting together the overlay pipeline task. It will consist of 25,000 jobs, taking about 1/2 hour each. (Richard) That's considered "modest". (Leon) Would be nice if it were going by the Pass 8 Face-to-face, but that's not likely.

  (Leon) has been trying to understand whether uncertainty in calibrations is causing the discrepancy between PSF for MC and real data. But before he got that far, he ran into reproducibility oddities. The same job run on 3 different fell machines produces identical MC events, but reconstructed quantities are not always quite the same. The same job run twice on the same machine does produce identical results; this most likely rules out different random seeds or uninitialized variables being a factor. (Richard, Heather,..) We should find out if the fells are identically configured.

  (Tracy) Everything is fine; that is, the GR we have now is not perfect but is adequate for the upcoming (in two weeks) Face-to-Face in Paris. Shortcomings will be noted and addressed in the next (in theory, last) GR release. (Richard) When does P8 reprocessing start? (Tracy) "Soon" but I would expect an update on what that means during or shortly after Face-to-face.

• Reprocessing (Tom G.) Jean-Marc has updated the diffuse model. Once put into production the P202 (Pass 7) FITS files will be reprocessed. Only the FT1 files contain the diffuse response column, but the entire set of FITS files will be regenerated.

• Core talk:

  • Root upgrade (Heather) As long as there are still issues with the new ROOT (e.g., skimmer problem which requires update to data catalog bits; occasional "fail to write" errors), she will hold off upgrading ST and, especially, L1. (Tracy) thinks the "fail to write" errors are manifestations of memory errors for really really large events.

  • Bill support (Heather) Shortly before the holiday break she installed a CMT build of the new GR, 20-08-01, on glast-ts; that is, she unofficially tweaked requirements files to make such a build possible and then built. She will go ahead and commit the changes, which won't affect anything else.

  • CVS interference (Heather) will poke Tom S. to see if he has had a chance to change definition of CVSROOT to local for all GR linux builds. (Tom G.) had suggested monitoring surrey04, the server for the repository. It is unfortunately not easy for SCCS to do, but not impossible.

  • Passwords (Heather) The glast password on Windows needs to be changed. She will probably also change glastrm there, even though it's not normally used, and perhaps Linux passwords for these accounts as well.

  • Windows and SCons (Joanne) Leon came across a problem with all wrapper scripts which should be straightforward to fix. Meanwhile there is a simple work-around: edit the offending line in the file by hand. He also discovered an error in the Workbook doc, now fixed. But the serious outstanding issue is supersede, which is not working for him. Before the holidays I sent him an alternate version of makeStudio.py (utility in the SConsFiles package containing code to create project and solution files) in an effort to fix or at least diagnose a problem unique to him: files all.sln and allGleam.sln are not getting created automatically as they should. Ever since that time, even after backing out the change, supersede has been broken. (Leon) Except maybe it did work once.

  • Workbook (Heather) would like to make a new production version when Joanne is finished updating. (Joanne) is done. More extensive SCons documentation, especially for Windows users, will be coming before long, but it will be in Confluence.

  • GoGui (Joanne) has been working on getting rid of the "init" step, which tends to complicate usage of GoGui. As of yesterday it was working in the new arrangement, but not as responsively as a friendly interactive application should. She now more or less understands why and hopes to come up with something better after another day or two of work.

  • RM and macs (Richard) At some point we should look into new macs for the builds...reopen talks with Yemi.

previous

minutes index

next

J. Bogart, Last Modified: