Core Minutes 6/26/2012
Core Minutes 6/26/2012ScienceTools: No report today since Jim isn't here, but see Science Tools Development Notes for the latest news. Since last week there have been a couple package tags and, on June 21st, a new ST tag incorporating all the newest package tags except for Likelihood and pyLikelihood.
Outages: (Heather) An update over the weekend of Crowd which affected Confluence users with 2 accounts (external and SLAC). There aren't many such but Heather is one. She was unable to edit Confluence pages. Her problem has been fixed. Any other individuals affected should contact apps-admin@slac
glastlnx03 and glastlnx12 crashed early Thursday morning. glastlnx03 is now back up but has not been given a clean bill of health. The reason for the crash of that machine is unknown but we're hoping it stays up because we need it in order to see systest results. It would be prudent to grab any plots of interest and save them somewhere while you can. glastlnx12 is going to get a new system disk.
FSSC: (Alex) They have the newest ST release and are unpacking it.
xroot: (Tom G.) Some minor issues have come up and have been resolved, e.g. one concerning a server which was returned online in read-only mode.
We're requesting another 450 Tbytes of xroot storage, targeted for September.
Pass7: (Tom G.) Trunc64 reprocessing testing and comparisons are in progress.
(Heather) Leon asked for a CMT (VS 2003) build which, with some difficulty, she was able to produce.
Pass8: (Tracy) Aiming for a new GR at the end of this week. (Heather) will make a new HEAD today, then incorporate more new tags later this week.
(Heather) attended the last Pass8 meeting. The SCons rhel4 build has been validated; rhel5 is still in the works.
Virtual machines: (Heather) One of the issues which came up at the Pass8 meeting was the difficulty those using Fedora (and there are several) are having running rhel5 SCons builds. rhel4 builds are ok because of the gcc34 compatibility libraries. There is nothing comparable for gcc41, the rhel5 compiler. (Joanne) Johan mentioned virtual machines as possible strategy. It just so happens I had recently written up my progress so far. The focus there is broader (everything we need for Fermi offline which might have to be frozen on some OS — presumably rhel6 — rather than solely development environment) but for developers I was thinking of making a GR appliance. Doing it first for rhel5 is only a very small detour. Furthermore, progress can be made in this area right now without guessing about unknowables, like the API for the SLAC batch system in 2016. Johan is enthusiastic; he would like to devote some time to it at the next C & A meeting. Tom S. has offered to investigate to what extent the OS can be stripped down to minimize the size of the appliance.
relh4 migration (Heather) Johan has asked that we keep rhel4 builds going through August. That might be possible, but we'll probably have to lessen our dependence on rhel4 in some other respects earlier than that: stop CMT builds and shrink the list of those able to log on to glastlnx14.
rhel6 externals (Heather) Johann has been working on them. All that's left is TMineExt. (Joanne) The most recent tag, 03-02-07, builds successfully on all supported platforms (rhel4, rhel5, snowleopard, vc71 and vc90). It looks like the newer gcc compiler on rhel6 doesn't like something, most likely having to do with templates. Could we ask Alex D. to look into it since he's familiar with the code? (Heather) We could!
rdbGUI (Heather, Joanne) Leon ran into a problem with it recently. The version of rdbGUI you get by default on rhel4 is very old and doesn't have the fix needed for connecting to MySQL 5. Joanne will rename to the February build. This one has a small display problem (column widths are not adjustable), perhaps because it's linked against a newer version of FOX.
Leap second (Tom G.) One is coming up so we need a new ST (already taken care of; see above) and GR.
Core talk:
Absences (Heather) Michael Kuss is out until July 1st. Email access will be spotty. Richard is in Scotland, returning the week after next.
DB migration (Heather) We're still waiting for grants issues to be sorted out before databases CMT RM, opslog and flight can be moved to mysql-node01.
SCons RM (Tom S.) Running smoothly. (Tom S. — can't speak but can type) I've enabled rhel6 builds for TMineExt. I'll get the rhel5 LATEST builds going tomorrow. I've also been working on getting the TMineExt latest builds to move beyond build 99. There is "feature" in the way we store the tags that causes problems when we add a digit to the version number that I have to work around.
GR vc90 (Joanne) No news this week. She's still waiting to hear from users.
Suddenly big builds (Heather, Tom G., Joanne) M.E. noticed that builds in recent months (since about mid-April) are 2 or 3 times bigger than they used to be. This only affects opt builds. The reason is that newer opt builds are compiled with -g (debug symbols). This is needed to make sense of core dumps and is also required if someone needs to debug the executable used in production (the non-opt debug version of the same tag may behave differently). At this time there is no hard evidence that we're taking a noticeable performance hit at run time because of the larger binaries. The other issue is just the time and space needed to copy the release, as is normally done for L1. However it isn't necessary to copy the whole thing. Less than half of it is needed at run time. L1 might be better served by starting with the user release. (Warren) will look into that.
SeeVogh (Tom G.) has been working closely with Wayne. Wayne can get SeeVogh to work on his 32-bit rhel6 system, but not on Tom's 64-bit. They have made various comparisons between the machines without turning up anything suspicious. (Tom S.) I had it working on my 64bit Scientific Linux 6 machine (a rhel 6 derivative) but I do have many of the 32bit libraries installed. (Tom G.) is intrigued and will pass this on to Wayne.
Announcements
previous
|
minutes index
|
next
 |
J. Bogart, Last Modified: