Core Minutes 9/30/2014

• ScienceTools: (Jim) Since the last report in mid-August there has been much work on energy dispersion for Pass 8 as well as an upgrade to Likelihood to allow energy dispersion calculation to be turned on or off by source, work on orbitSim from Joe, and changes from Tom Stephens to use matplotlib rather than Root under some circumstance. There is a new ST release incorporating these updates: 09-34-03.

  See also Science Tools Development Notes.

• FSSC: (Joe) There is more to do for orbit simulator; he's working on that.

• Reprocessing: (Tom G.) Pass 8 is complete through April 30. There is now agreement to do another backfill through 9/30, to be started up soon.

  The astroserver has the Pass8 base data (source events). The extended class events, including transients such as GRBs, are being worked on but have been delayed due to issues of optimizing their organization within Oracle. As a fall-back, the old FITS skimmer is being dusted off and reassessed for possible use in obtaining data for limited time intervals such as GRBs.

• Hardware: (Tom) Fermi has 143 Tbytes of storage remaining on the old xroot servers. The new servers (~700 TB) arrived in July but critical cables just arrived last week (!). The Computer Center is busy running GPFS benchmarks on these servers as they are the only ones at SLAC with the necessary hardware to test the needed GPFS configuration. The new servers should be turned over to Fermi soon.

  (Richard) There are concerns about our nfs servers, some of which are quite old. There will be a meeting with Yemi to discuss migration to something newer.

  (Tom) wain25 (user/group disk server) has not been behaving well. A steady 40% of the cpu has been going to system usage, way too high. Last week Renata et. al. discovered several batch machines were keeping connections to wain25 for no good reason. These machines have been rebooted and now the base system usage is more like 10%, still higher than expected, but much improved. This server is chronically overloaded and is 6 years old, making it a prime candidate for migration.

• Data Catalog: (Tom) Brian is in the process of rewriting the whole thing. He's close to done; it may be deployed within a couple weeks. No immediate effect on Fermi running is expected, but ultimately we might want to take advantage of new features.

• Pass8: (Leon) Efforts in C&A to improve energy dispersion at low energies (down to 20 MeV) are bearing fruit and close to completion. There is also a package to do fish-eye corrections, primarily for transients at low energy, but useful up to 1 GeV for large angles. This software will be employed when making IRFs.

  Cosmic-ray analysis is another active area, much of the effort going into separating the electrons from the background (mainly protons). In principle, a clean separation gives you the electron signal and the proton signal. Obviously, background contamination and efficiency are going to be different for the two analyses, but the cuts can be similar (with the inequalities reversed!).

• Core talk:

  • rhel6 validation (Heather) At Johan's request she will retrigger builds of 20-09-10 for rh5-32 opt, rh6-64 opt and rh6-64 debug. rh5-32 debug already exists.

  • glastlnx07 (Heather) As preivously — it needs to be retired, which means we need to get WIRED working elsewhere. In progress.

  • Mac and Jenkins (Joanne) Up till now we have been unable to get the Jenkins daemon to restart automatically upon reboot. Apparently processes which start at reboot must be configured in a system directory, but we need this one to run as glastrm user. Last week I found a way to specify in the configuration that the process should be run as glastrm. Late yesterday I put that in and rebooted. It looked ok initially — process is running on the mac under glastrm user ‐ but the new ST release is not building there; it's hung up. [After the meeting we discovered an uppercase/lowercase issue which we were able to work around. Still some mysteries, though.]

  • Gui Installer (Richard) It's failing again after a brief period when it worked. Is it worth putting (Joanne's) effort into it or not? There is a work-around (command-line installer), but this kind of infrastructure decay is distressing. (Joanne) Doesn't believe anything in the code changed during the period it worked. In fact, I don't think the time period is necessarily a factor. It has worked for me on both Linux and Mac whenever I've tried it. More likely it is not a bug in the code itself, but some combination of environmental factors and transient conditions which will be difficult to untangle. The last time this came up Tony strongly implied it would not be a good use of my time to work on this problem. He suggested that, if anyone is to work on it, Max would be better candidate since he is already familiar with the code.

  • Workbook (Heather) Status continues to be unsettled. We are moving towards a hybrid arrangement with new material in Confluence and only updated pointers in the Workbook itself, which seems better than a wholesale move to Confluence. There is reason to believe user confidence in and knowledge of the Workbook is waning. (Richard) will discuss with Philippe, as primary user representative.

previous

minutes index

J. Bogart, Last Modified: