Core Minutes 7/24/2012
Core Minutes 7/24/2012ScienceTools: (Jim) New tags of CRflux and EarthPhenom (new subpackage of celestialSources) are in GR HEAD and also in the tagged release along a new GR branch.
FSSC: (Alex) We're preparing a rhel6 VM for the Fermi solar workshop so that Windows and Mac users will have access to the full FSSL environment, including, for example, IDL licenses.
Astro Server and event selection (Tom G.) An inconsistency between certain data products developed during the P130 reprocessing (performed in April 2012) for the period 1 Aug 2011 through 18 Apr 2012. The event selection for FILTEREDMERIT files got out of synch with EXTENDEDFT1 and EXTENDEDLS1 files. This causes a problem for the Astro Server as in its current design a lock-step is required for events in all three data products. Two solutions are being considered:
There may be hints of other problems in MERIT event classification bits for data prior to 1 Aug 2011 (part of the original Pass7 (P120) reprocessing), but this must be substantiated. Investigation continues.
L&EO reprocessing (Tom G.) It's being discussed. Naturally a new task will have to be set up. There is also some question whether existing calibrations will be adequate. Leon believes tracker calibrations will be; Sasha is doubtful about CAL. Instrument response may have been changing relatively rapidly at that time.
Hardware: (Richard) The xrootd storage order is fully approved. It should be available by August, in plenty of time.
Our old, 4-core glastlnx machines will be replaced by 10 12-core machines.
More cycles (Richard) We may be able to get a significant quantity from NASA Ames, with 20,000 cores at their disposal. There is a hitch, however: their machines are running SUSE linux. Tom G. has been asking around — so far fruitlessly — to see if anyone has experience running our stuff on SUSE. (Tom S.) It would be easy enough to try it out with a VM running SUSE.
Pass8: (Tracy) A problem with Trunc64 simulation runs turned out to be due to misordering: TkrDigiTruncationAlg (which does the software truncation) was called after EbfWriter (which creates the data for OnboardFilter). A rearrangement of job options took care of it.
Johan encountered a problem with P8 overlay runs at Lyon affecting both rhel4 and rhel5 SCons builds but not with the corresponding CMT builds. test_Overlay had been disabled since it hadn't been brought up to date with the rest of the Overlay code. Tracy has resuscitated it and it's now working for CMT. Someone SCons-proficient needs to look at it there. [Update: SCons-specific bug was found and fixed]
Core talk:
vc90 and GR (Tracy) believes he's close; needs to slog through more code in the vicinity of the point of failure (end of processing of first event).
python for rhel5 64-bit (Heather) An old problem has reappeared: when remote users try to use the SLAC-built externals, the md5 library can't be found. She had hoped that this problem would go away with the move to a newer python on this platform (2.7.1 rather than 2.6.5) but it has not. The only known solution is for the remote user to install the library locally.
rhel4: (Heather) glastlnx14 is now fully repaired and back in operation. She had been hoping to shut off rhel4 builds tomorrowish, but, given vacation schedules and not-quite resolved issues with Overlay and SCons (see below) she'll hold off.
ROOT retries (Heather) Tom G. had reported a possibly-transient problem reading ROOT files, so she has added retries in RootIo, along with a couple new job options:
(Tom G.) The read failures will be a topic of discussion at the xroot meeting later this morning.
db migration fall-out (Michael) has been unable to log in with his account. It could be a password issue. [Update: it was, now fixed.]
relh4-to-rhel5 migration (Heather) We're in the process of validating L1proc on rhel5, comparing systests for the current production version and the rhel5 candidate. There are differences; Johan is looking to see if they're acceptable. He also would like to revisit the content of systests. Do they examine the right stuff?
(Heather) JJ has been contacted about upgrading the GRB tools to run on rhel5 and rhel6. He is willing to work on it and has access to rhel4.
Virtual machines (Joanne, Tom S.) Johann tried out Tom's version 2 of the GR "bare bones" VM. He installed just a couple more things (CVS, SCons) and was then quite happy with it as a development environment. Joanne will make a new appliance with these two additions. See draft instructions in Confluence.
CMT tags gone missing (Heather) noticed that recently they have not been created automatically with each SCons-style tag. It's not just a transient failure: there haven't been any for a while. One or both of Tom S. and Joanne will look into it.
rhel6 (Joanne) The only remaining known problem most likely concerns TMineExt and 64-bit. It might be easier to diagnose if ROOT exception handling were turned off inside TMine. (Heather) will send instructions.
previous
|
minutes index
|
next
 |
J. Bogart, Last Modified: