Core Minutes 10/16/2012
Core Minutes 10/16/2012ScienceTools: (Jim) There are two new items:
See also Science Tools Development Notes.
FSSC: (Tom S.) The latest version of ST now builds on all supported linuxes; work on macs is in progress. Expect a release next week.
Reprocessing: (Tom G.) It's now done up to October 6th. Fits file generation is in progress, in fact almost done. Then the new data needs to be ingested by astroserver.
(Tom G.) The FSSC discovered a problem with two LS1 files from the ongoing P202 reprocessing: an EVENT_ID was repeated ~1000 times. The underlying problem is still not understood but is believed to be a silent ROOT error which causes internal file corruption. There were no indications of problems (i.e., no warning or error messages) in any of the ROOT file generation steps. However, makeFT1 did spew out a lot of error messages but went ahead and created the (corrupt) FT1 file from a corrupt MERIT file. Five additional files (out of 22,090) were discovered to have these error messages in the log files. Jim kindly added some more checking to makeFT1 so that now it will return a non-zero exit code should this happen in the future. The rarity of the problem makes it difficult to reproduce. Heather is following up with ROOT-TALK.
Hardware: (Richard) The 400+ Tbytes of xroot space is installed and in use.
(Tom G.) The first two (out of four) nfs server migrations are done. Next up are the partitions used by ISOC operations. The last one affects partitions used by RSP, ASP and so forth. The new server (wain32) is drained and ready but we need a plan for the migration.
(Tom G.) Ten stand-alone Linux servers are on order, due to arrive on Oct. 24th. They will replace the existing 25 or so glastlnx machines. A couple will be dedicated to FastCopy; others will host other servers.
Pass 7 (Leon) Is back to working on Overlay after a hiatus while he was dragged away for something else. Nothing to report today.
(Tracy) B3-1-0 update for new truncation scheme was uploaded on schedule today and activation is planned for Thursday. (Joanne) The export libraries from flight software showed up shortly after last week's meeting. She used them to make an obf B3-1-0 for rhel5 last week. Once back from BNL she finished updates to OnboardFilter, OnboardFilterTds and SConsFiles needed to use the new external. The new stuff builds and test programs run. Output should be compared to production and then all that remains before Heather can make a new GR tag is routine stuff — tagging the new packages, installing B3-1-0 in the standard place. (Heather) is wondering how to integrate into P8. Branch off of last good tag (before TMine issues) or just put it at the head? (Tracy) The newest stuff is ok as long as we don't try to use the Pass 8 worksheet.
Pass 8 (Tracy) The strange error loading certain libraries observed last week turned out to be due to cfitsio. We needed a debug build. But since then there have been some TMine issues. The first was an array indexing error when you read in a new worksheet. The error was visible on Windows since bounds checking is routinely done there in debug builds. On Linux it failed silently. Alex fixed that one, but now Gleam crashes on init because it's using too much memory (4 Gbytes). On Friday the Pass8 analysis was turned off. GR tag 20-06-02 works as long as it stays turned off. We probably will have to do it in a separate job step. (Heather) then we don't need TMine in GR any more? (Tracy) sees reasons for hanging on to it even if we go to a separate job step.
Core talk:
GlastRelease recommendations (Heather) wrote up proposals on several troublesome topics connected with developing and releasing new GR production releases: sticking to a sensible time frame, schedule and procedure when making a new release; using System Tests effectively, formalizing requests for updates via JIRA and so on. [If you haven't already done so, please follow the link and read the contents now! ed.]
(Tracy) would like to see Systests run on HEAD builds, as a prerequisite for making a real tag. (Heather) agrees, but there are a couple practical matters to attend to: there needs to be enough disk space (she'll come up with an estimate and talk to Tom G. about it) and we have to make optimized HEAD builds.
(Joanne) is not yet convinced we should merge TMine packages back into GR; Ipros and cons should be weighed.
Concerning JIRA, I have never been able to find a good way to filter messages. The total number I receive is rather high; I end up dumping them into a dedicated mail folder and ignoring them. There must be a better way.
Pass8 developer environments (Tracy) P8 Linux developers have expressed a willingness to try out new code on glast-ts to find out about and fix some kinds of Windows-only glitches before tagging. (Heather) Will they need support there? If so, where does it come from?
(Joanne) We also (even more) need traffic going the other way; that is, Windows developers should be accumulating experience and expertise in the Linux environment. I'm well aware that support for Windows is far from ideal now; improvements will be coming, if anything, slower in the future because I have less time to devote to it.
(Joanne) In recent TMine episodes VS debug builds had a clear advantage in that bounds-checking is done by default. gcc apparently has a similar facility. We need to try it out. Depending on what we learn we could incorporate it into all debug builds or add a feature (e.g., scons command-line option) allowing easy creation of an additional kind of debug build.
lsf replacement (Tom G.) This project is on hold for the time being. (Richard) The reason is lack of manpower within SCS to install test versions of candidate replacement systems.
ROOT upgrade to v5.34.01 (Heather) hopes to get back to it and finish it off. Once it's done we ought to be able to get rhel6 64-bit builds going.
previous
|
minutes index
|
next
 |
J. Bogart, Last Modified: