Core Minutes 8/21/2012

• Reprocessing (Richard) Tom G. completed reprocessing of L&EO with new calibrations. (Johan) believes regular reprocessing is complete through July 31.

• Hardware (Richard) We took the opportunity of the scheduled power outage on Thursday to move to the new user disk. wain06 was replaced by wain25, which is twice as big. Should be transparent for users identifying their space via /afs/slac/g/glast/users.

  10 replacement boxes for the glastlnx servers are on order. We've also ordered a test Oracle database.

  We're still working on the rec for the batch farm. We will just be paying for the compute power; money for enhanced connectivity will come from PPA.

• Pass 7, Pass 8 (Leon) No news, really. He's been mostly reading and approving abstracts for the symposium.

  (Tracy) is also busy preparing for the meetings. He's battling a Fred issue: he can use it to display event, but cannot save the images in any shape or form. This is especially annoying because it did work before. (Leon) suggests using Hypersnap or similar product to capture parts of the screen.

  (Tracy) The goal of the Pass8 meeting is to recap progress to date on recon and make the point that the focus should now be moving to event-level analysis. There has been some work in this area already and results look very good.

• Workbook: (Heather) There is general agreement that we don't need the print version of the Workbook. She might send one last message to an appropriate list, alerting people to the proposed change.

• Core talk:

  • rhel4 decommission (Heather) We plan to shut down CMT RM builds on rhel4 next Tuesday, Aug. 28.

    The consensus on intermediate obf builds (in the B2.. series) on rhel4 is that they are not needed, thus lessening the work FSW has to do on rhel4.

  • rhel5 L1 validation (Heather) We might be there! Comparing rhel4 and rhel4 SCons builds with same sources looks good. (Johan) thinks so to. (Heather) The next step is for M.E. to try the rhel5 build with real data.

  • GR tags (Heather) made a final tag (20-04-02) prior to the massive tagging needed for non-installed headers on Windows. There are some issues with the opt build, apparently concerning th file Event/Event/RelTable/RelKey.h, which has been a troublemaker in the past.

    (Joanne) The non-installed-headers work is close to done. Steps are

    1. Make new tag of SConsFiles with backwards-compatible changes
    2. Make new tags of about 80 "regular" packages, incorporating mostly small changes to ensure that include paths on Windows have been adjusted so that compiles will still work when there is no central include area
    3. When LATEST build incorporating all of the above looks ok, commit further changes to SConsFiles to eliminate central include area on Windows (when building GR only) and tag.

    Steps 1. and 2. are done, but the LATEST build incorporating the changes has been a trial. The first attempt was bad because of a stale lock in flux so that the LATEST tag did not include all of flux. Brief (scheduled) outages of afs servers were going on during the second attempt, making externals unavailable at times, causing many compile errors. A third attempt begun late last night is still not pristine; problems seem to be centered around FluxSvc. But she's hoping to get this sorted out today. Progress and description of procedure may be found in Confluence. [Update: turns out I didn't follow the correct procedure to fix the original problem with flux. Build attempt #4 is in progress.]

  • SCons RM (Heather) There are a couple minor problems to be resolved:

    • The Windows distributions SCons RM makes end up with the wrong file protections so that the Installers can't access them. We've been chnaging the protections manually for now.
    • The environment variable LATCalibRoot is not getting the proper value on Windows so certain test programs fail. We don't yet know where the value is coming from — could be RM or it could be something that needs to be corrected for the glast user on glast-win04.

  • Virtual machines (Joanne) Nothing new to speak of. Users seem to be happy with bare-bones (Johan concurs). GR standard has been ready to try for a while but has been shunted aside. I hope to get to it as soon as Windows no-install work is done.

  • Python (Heather) Matthew Wood reported trouble with matplotlib in the python 2.7.2 we distribute with our RHEL6 ST builds. The issue was associated with Tk which was not fully installed when this version of python was initially assembled last year. Heather noted that indeed, we had started a 2.7.2-gl1, which was never fully set up. She has now finished adding in all the modules and has asked Matthew to give it a try. Assuming all goes well, we'll add this to ST.

previous

minutes index

next

J. Bogart, Last Modified: