Core Minutes 24 April 2012

Core Minutes 4/24/2012

Present: Joanne Bogart, Jim Chiang, Warren Focke, Tom Glanzman, Heather Kelly, Michael Kuss, David Landriu, Leon Rochester, Alex Reustle, Tom Stephens, Tracy Usher, Eric Winter

IFC Meeting: (Heather for Richard) The meeting was yesterday. The computing budget was not on the agenda, which is probably good news.
ScienceTools: (Jim) fixed a couple bugs in Likelihood and satisfied a request to delay reading in certain data until required (performance enhancement). He further [above and beyond the call of duty] fixed a bug in the JulianDate class in astro which had been affecting output of the test program in some circumstances. See Science Tools Development Notes for details.
FSSC: (Tom S., Eric W.) Eric is leaving! — but, fortunately for us, not going very far. It will probably take some months to hire a replacement. Alex will be picking up some of his duties.

(Eric) was informed last week that Scientific Linux 6 machines (like the one on his desk), user write access to /tmp would be turned off as a security precaution. I heard second-hand this morning that it may only be a rstriction against creating executables in /tmp. If general user access is denied, the Science Tools will fail spectacularly. If the restriction is only on executables, we should be fine. This situation is evolving, and I don't know what the ultimate outcome will be.
Pass7 reprocessing: (Tom G.) It's humming along, now about 45% complete. If it continues at the current rate it would be done in early June or perhaps late May but, given all the interruptions we've had to date, it doesn't seem likely we'll escape without more.
New hardware for old: (Tom G.) The 13 sulky nfs servers are due to be replaced. They're old (5-7 years), slow and, more critically, short of memory. They handle about 100 Tbytes total We plan to replace them with a smaller number of more powerful machines, probably old xrootd servers which have been replaced by new Dell machines. We need to to arrange it so that heavy users don't all end up on the same server.

glast linux servers are also in need of replacement. The older ones are 5 years old; the remainder are nearly 4 years old and, like the sulky servers, they've had hardware problems. These will be replaced by new Dell compute servers.
Pass7 news: (Leon) would like to get the truncation data-handling code now in a branch off of L1 incorporated into L1. The code will, if so requested, convert data from the new truncation scheme into the old form. Otherwise, it does nothing. (Heather) Is sys tests output adequate to test this? (Leon) Would also be a good idea to compare behavior with that of L1 when fed real data. In fact, it would be good if systests included such a run. (Heather) would like to integrate this code with the current production L1 which uses obf version B1-1-3, not the one using B3-0-0 which she is currently trying to validate [see below].

(Leon) has been trying to understand what's going on with magnetic latitude. The relevant code is f2c-generated, hence nearly inpenetrable.
GR validation: (Heather) has been comparing a rhel4 build of GR 17-35-24-gr27 adjust to use obf B3-0-0 with the standard L1 (17-35-24-lp22) using obf B1-1-3. M-E has also been looking at it. There are some differences in the OBF output for MIP and HIP filters. Expert opinions are solicited as to their significance.
Too many branches: Everyone would like to see fewer. Leon and Heather will work on eliminating the branch off of the L1 branch as discussed above. Heather will also look into merging P8 back into HEAD. (Tracy) suggests P8 developers could hold off on CVS commits for a day or two to make the merging easier and faster. (Heather) There is a known issue with HEAD concerning Event display, both WIRED and Fred: the connection is lost after the first event. (Leon) It's maybe just as well to make a tag with this so that Dima can work on it more conveniently.
Core talk:
- MySQL migration: (Heather) The upgrade and move to mysql-node03, destination for the calib dbs and mood mirror, is currently hung up because of issues concerning EXO databases also resident on that machine. The move of our other databases to mysql-node01 is still waiting on OpsLog.
- RM news (Tom) There was one recent tagging failure. Its handling by RM demonstrated that the code was still not behaving as intended (that is, RM still tried to create builds with the flawed tag). This should now be fixed; we'll find out for sure the next time there is a tagging failure. Otherwise, there have been no problems with RM itself. There are compile and test program failures on Windows.
- GR vc90 (Joanne) As of SConsFiles-00-21-00 SCons-generated project files will install headers, xml files, etc. This addresses the first issue of last week. But the second one (getting SCons to create an all.sln file) is a much harder nut to crack. It was working briefly yesterday; then I did some straightforward clean-up (eliminated commented-out lines, eliminated a function definition for a function that was never invoked... I don't think I did anything else) on a crucial python file and it stopped working.

minutes index

J. Bogart, Last Modified: