Core Minutes 9/20/2011
Core Minutes 9/20/2011ScienceTools: Jim is on vacation; no report this week.
FSSC: (Eric W.) has mostly been spending time on the web interface, not on ScienceTools itself. There is a a push to find and install a fix for seg faults in gtsrcmaps. (Richard) It is connected with running out of memory. See thread "segmentation fault on gtsrcmap" in helpsoft list
Documentation: (Heather) We're doing fine with making updates to existing pages. Creating a new page is another matter. It can be done by hand, but it's awkward. (Richard) will ask around to get suggestions on modern web-authoring tools.
Pass7 reprocessing: (Leon) Cal moment analysis has improved significantly going from GR v15 to GR v17 and this has a noticeable effect. For example, comparing ultra-clean (that is, using Pass7 cuts on both old and new) about 10% of the events selected are different. And about 5%, though the "same" events, give different answers. Using new calibrations also changes things, especially the new cal calibrations. However it's still not obvious that the full reprocessing should be done. We'll be doing it for Pass8 in any case. He will put together some plots and send around for reactions.
(Richard) Just in case we do decide to proceed with reprocessing, discussions with Lyon continue. There is no problem finding space for the digis, but the requirement of about 1/2 petabyte for storing recon is being greeted with gasps. It would only be needed for about 6 months because network bandwidth is not sufficient to keep up. It may be possible to borrow the space just for that period.
GR/Pass8 (Tracy) Bill is able to work with the new GR release, but runs into some limits because of performance issues (and an unwillingness to break up the jobs and run them in batch on Linux, with which Tracy has some sympathy).
Tracy is matching trees to clusters in the Cal and discovering various things, all good.
VM projects: (Richard) First use has been at Toby's urging: Tom made a Linux VM with ScienceTools (and other additions as requested). Toby can then run it on the UW terminal server. How much time and effort does this take? Is Toby the only customer or might there be others some day? Is there a difference between using Virtual Box or VMWare? (Tom S.) There is no significant difference between them. Both essentially work out of the box. The effort goes into creating the stuff that will run in the VM. It takes a certain amount of time but isn't particularly difficult. (Richard) Could Toby do this himself? (Tom S.) Yes. Only issue is having sufficient Linux expertise.
(Richard) The other application to be looked at is "GlastRelease"; that is, some subset of elements involved in the processing pipeline so that, some day, we can freeze on a particular OS version or versions and not have to continually update just to stay runnable on a new OS. GlastRelease is a particular concern because of its dependence on external libraries which are not under our control. The alternative is to maintain sufficient staff to keep current.
People at Fermilab are looking into similar issues for the Tevatron; collaboration should be of benefit to all. Amber brokered a (management level) phone meeting with them. They are especially concerned about 1) connections to external servers; e.g. Oracle, HPSS 2) security. One major difference on our side is the possibility of moving to VM's while we're still taking data.
Heather will look into the staffing issue; Joanne will characterize processing elements and tendrils: dependencies on resources such as databases, storage, etc. Elements under consideration include everything from the half-pipe through ScienceTools.
(Eric W.) Even if ScienceTools are run in a VM (and hence frozen at some OS level) for pipeline processing, FSSC will continue to port them to new OSes.
Core talk:
WIRED (Leon) The major effort to get past show-stoppers is over and a success. Now people are using it and encountering lesser issues, such as the fact that if the user exits from WIRED when it is attached to Gleam, it doesn't exit Gleam gracefully. Dima is currently merging plug-ins from various experiments to avoid maintenance headaches. Once that's done he'll attack these problems.
Outstanding ntuple JIRAs (Heather) There are a couple:
SCons RM (Tom S.) Builds are running as they should, even on Windows! That may change after OS patches are applied tonight.
More build news (Heather) is keeping rhel4 builds of GR tags going in both CMT and SCons. She's tracking down problems with (SCons) vc71 GR HEAD builds. There are some updates, currently living only the UW terminal server, about to be committed to CVS, which should help not only vith vc71 but also vc90 builds.
obf (Joanne) and Tracy attended a flight software meeting which included some discussion of B3-0-0, among other things. They can build it on rhel6 as well as rhel4 and rhel5, albeit with some compile warnings which they plan to clean up. They were interested in helping us get away from the need for linking in extra libraries not in Tony's export area.
Backing up a bit — by the time of that meeting I had made obf externals based on both B1-1-3 and B3-0-0 and modifications to OnboardFilter so that the same source could use either version of the library. This required some changes in the SConsFiles package as well as to OnboardFilter so that a compile-time macro for obf version could be defined and available in OnboardFilter. However, for B3-0-0 I was linking against the library efc, which drags in quite a few other libraries. There is an alternate version for offline use, gefc, and also alternate, g-prefixed versions of the dynamically-loaded libraries gfc, dfc, mfc and hfc (concerning respectively gamma filter diagnositic filter, mip filter and hip filter) which we're suppose to use instead. I tried this out. The gamma filter stuff (controlled by library ggfc) went through but there was a problem with mip. It turns out gdfc, gmfc and ghfc were not quite right, however the problem is understood and JJ is working on a fix.
Other externals (Heather) has built G4 (both 8.0 and 9.4) on rhel5.
Announcements
SCons meeting this week at 10:30 on Wednesday.
previous
|
minutes index
|
next
 |
J. Bogart, Last Modified: