Core Minutes 10/2/2012

• SeeVogh: Yet another attempt today. We now may have a majority on SeeVogh [including me!], but by no means everyone.

• ScienceTools: (Jim) There are three recent items:

  • Progress on the implementation of a scheme to set correct IRFs automatically when running tools downstream of gtselect. It's now committed; not yet tagged.
  • Julie requested ability to select on minimum zenith angle. See JIRA STGEN-133.
  • Update of FT2 files to include lambda parameter.

  See also Science Tools Development Notes.

• FSSC: (Alex) is working on an update to add NumPy and SciPy. It's all right on Scientific Linux; he's running into snags on Mac.

  (Tom S.) is still wrestling with Mountain Lion and ROOT. Current production version of ROOT (5.26) doesn't compile. The new one Heather is about to release (5.34). However he had to use a Mac port of gcc 4.3 to do it. At run time, certain symbols can't be resolved. This is likely a problem with his machine's configuration.

• Hardware (Tom) glastlnx servers are on order; not yet on site.

  There has been talk of a "high availability rack" so that essential services can weather outages. This could include mail servers, web servers, business servers, afs space...

• nfs partitions (Tom G.) wain32, still part of th xroot cluster, is being drained. It will become an nfs server. wain31 is in the process of being commissioned for ISOC.

• Reprocessing: (Tom G.) We're waiting for IRFs and new calibrations for the next interval (post July). (Leon) We get new CAL calibrations every 2-3 months and new TKR dead strips every 2 months. The current suite of calibrations is good through September.

• New FT2 (Leon) Will make FT2 including the lambda parameter and recalculating other magnetic variables with the new, more accurate model. Jim has added the parameter to the template so now it's just a matter of finding an appropriate version of GR and pushing the button. (Heather) In particular need correct versions of astro and fitsGen packages.

• Pass 8 (Leon) Current issues:

  • Re-animating the Gleam job that skims periodic triggers and generates overlay.root files Tracy and I converged on a small set of fixes... it's running again, and mostly committed and tagged. (FYI: it takes ~ 2 hours to skim and convert the ~10K periodic triggers in a run.)
  • The code in AnalysisNtuple that automatically removes MC-related variables when the jo parameter realData = True stopped working recently, because of compiler pickiness. It's fixed and about to be committed.

  (Tracy) Many new things will go into the new GR tag (20-06-00), now scheduled for October 3rd (middle of the day or later). See a list on the Pass 8 agenda page, Sept. 27 entry. New tags are still coming in. We may delay a day or two for CalValsTool work. With so many new things, it's likely some time will be needed to pick up the pieces.

  (Heather) the new tag will use a new TMine release, 3.2.8.

• Core talk:

  • Pre-generating .pyc (Toby) accidentally generated .pyc files in the release build area. This raises a couple issues:

    • He was able to write there because he belongs to group glast-RM. It would be better if he couldn't and didn't; there is no need for him to be in that group.
    • If the byte-compile were done as part of the build process it would save time when the code was run, both for batch jobs at SLAC and for remote users.

    (Joanne) This came up before. I think we concluded that .pyc files generated at SLAC would not necessarily run successfully when exported. But we could make them at SLAC if there is a way to do it after the tarballs are created, and remote users could create them after downloading a distribution. (Tom S.) also remembers the earlier conversation and the conclusion of potential problems when exporting the files. He believes he could tack on the generation in RM workflow after the tarballs are made, as part of the finalize step.

    (Toby) That leaves the remote-user case. (Joanne) would start with documenting how users can do the byte-compile themselves.

    (Heather) Also stemming from this experience - we will clean up membership of glast-RM.

  • lsf replacement (Tom G.) Next step will be to approach Amber's group about pilot projects for the two leading contenders: Univa/Grid Engine and SLURM.

  • GlastClassify and extFiles Recent versions of GlastClassify include large Pass 8 worksheets (xml files) which seem to be implicated in recent RM problems: timeouts, CVS lock files, etc. We have therefore moved these files out of GlastClassify and into the externals (in particular, into the external extFiles, but it's not clear whether this has had much effect on RM problems.

    The files are still kept in CVS, but in a separaate module, pass8Analysis, which is not a GR package. Developers can check it out and check in changes as usual, but making a release entails making a new extFiles tag.

  • ROOT upgrade (Heather) The upgrade is a little stickier than originally anticipated because Gaudi has to be rebuilt against the new ROOT. The upgrade will therefore follow the upcoming Pass 8 tag rather than preceding it.

  • TMine and 64-bit (Heather) Johann may have tracked this down. It is apparently a compiler issue, not a 64-bit problem as had been thought.

  • SCons RM (Tom S.) Fixed the variant=NONE issue with TMine so that proper builds will now be created.

  • Windows developer environment (Joanne) Leon ran into a strange problem having to do with the creation of the master solution file, all.sln. Tracy and Joanne both went through the "same" procedure successfully, but in Leon's case a path argument to a function seems to have been mislaid. Fortunately there is a work-around: call the python program which creates this file by hand with the proper arguments. (Leon) Overall, SCons/vc90 is working and I'm getting used to it.

  • VMs (Joanne) Another user has appeared on the horizon with useful suggestions about what additional services might be built into VM, or at least documented so that users could add them themselves.

previous

minutes index

next

J. Bogart, Last Modified: