Core Minutes 18 December 2012

Core Minutes 12/18/2012

Present: Joanne Bogart, Jim Chiang, Richard Dubois, Warren Focke, Tom Glanzman, Heather Kelly, David Landriu, Leon Rochester, Alex Reustle, Tom Stephens

ScienceTools: (Jim)
- P7REP_*_V10 irfs added to ST LATEST and HEAD, a release tag to be made today.
- a couple of minor bug-fixes for python interface of Likelihood
- Steve F. is working on gtselect issue from Jean (STGEN-137)
- DMFIT model updates from Johann
- additions to st_facilities to support making psf available to GR from Johann.
See Science Tools Development Notes for details.

(Richard) Postponed trip to France for Jim and Richard is now planned for late January.
Root upgrade for ST From the strictly-ST point of view, there is no reason not to forge ahead, in fact this would be a boon for FSSC since Mountain Lion builds would then be possible. But then we would also need to upgrade Root for L1 and would need to build the skimmer against the new Root. There is no hope of getting all this done before shutdown, so it will be after.
FSSC: (Alex) Current work is internal to FSSC and mostly has to do with externals, not ST itself.
Reprocessing (Tom G.) 1000 runs (covering Oct. 6 to Dec. 11) are in progress. ROOT files have been generated. Expect to be done with FITS in a couple days.
Hardware and migration report (Tom G.)
- He has been watching new nfs servers (which had been xroot servers) and all seems well. The user disk is by far the most popular, but is surviving nicely.
- Bulk of batch machine order went out on Friday so naturally have not arrived yet. They will be rhel6 64-bit machines.
- Status of new fermilnx machines is unchanged: services are not yet migrated to them. Expect that soon after the holiday break.
SeeVogh/EVO (Richard) Users at NRL have been having trouble with EVO. It seems someone was stripping off certificates, but no one there has owned up.
Pass 7/8 (Leon) The FAKEOVL task in Lyon for 20-08-00 (with essentially empy fake overlay files) was exhibiting miserable efficiency. Tracy tracked this down to some code in RootIo doing a sleep when the read failed at the end of file. He fixed it by keeping track in Overlay of when it needs to wrap back to the first event, rather than letting RootIo ever get to end-of-file.

There is another issue with Overlay noted by Carmelo which predates 20-08-00: the merit files for corresponding overlay and fake overlay runs are out of sync. This problem pre-dates 20-08-00.

Leon has been working on a pipeline task to produce new overlay files and associated merit files (up till now we've been using a fixed collection of overlay events, all coming from early in the mission).

Warren has volunteered (or been volunteered) to be calibration librarian, for which Leon is grateful.
GR tags (Heather) Tracy's new Overlay tag is in the new GR 20-08-01. Meanwhile, she is running sys tests on 20-08-00. All but the slow ones are done and look OK.
Core talk:
- CVS interference (Heather) We had some between rhel5 and rhel6 checkout for 20-08-00. For the time being, Tom has shut off automated rhel6 builds. We don't have this problem with ST. There are a couple differences which may be significant:
  - ST has a smaller codebase than GR
  - ST builds are written to u35; GR builds are written to u52
  (Tom S.) CVS checkout is timing out, even though the time limit was increased from 30 minutes to 45, then to a full hour. By hand checkout time is about 20 minutes. There seems to be contention over the CVS disk, the CVS server, or u52 where the checkout gets written. He's tried staggering checkout start times (to the extent possible; controls are not as fine-grained as one could wish) and although it might be helping somewhat, the problem is still with it. We could temporarily switch GR builds to u35. (Richard) Where are the checkouts being done from? (Joanne) That is, what's the definition of CVSROOT? Local or remote? (Tom S.) checked and determined we have some of each for GR. For ST it's always local (except for Windows, where it must be remote). Tom will try switching to local wherever possible.
- Windows pow-wow (Joanne) Leon, Tracy, Heather, Richard and I met last week in order to assemble a to-do list of items which
  - would be sufficient to provide a usable, if not necessarily comfortable, environment for Windows developers
  - would be doable on an interesting timescale
  We may have succeeded. I've started looking into a couple of the items on the list; nothing to report yet.
- RM and macs (Heather) Jobs had been stacking up because certain units tests (pointlike, skymaps) routinely fail and hang up. In fact these tests fail on all platforms and package owners have not been responsive, so they (the tests, that is) will be disabled. (Tom S.) In theory the hung-up test processes should time out, as happens on other platforms, but for some reason the Qt timeout is not working on macs. Speaking of macs, is there a reason why we're only using bldmac02? (Heather) No reason she's aware of. bldmac01 should also be made available.
  
  (Tom S.) recommends that, if we upgrade the macs we go straight to Mountain Lion, skipping Lion. Otherwise by the time Lion is installed it will be nearly obsolete. (Heather) When she asked the Computer Center about this last summer they had no interest in Mountain Lion. Perhaps now they'll respond differently.

minutes index

J. Bogart, Last Modified: