Core Minutes 5/8/2012
Core Minutes 5/8/2012ScienceTools: (Jim) No news, at least no items on the development page. He's started looking into geomagnetic field code to see what it would take to extrapolate beyond 2010.
FSSC: (Tom S.) Servers were recently upgraded from Scientific Linux 4 to SL6. There were some minor hiccups but everything is ok now.
Pass7 reprocessing: (Tom G.) Now about 68% complete. Failure rate has been low. The number of machines available is far from constant: anything from 600 to 2000. If we continue at the current rate we might finish with this step by June 1st or 2nd. It will take another couple days to backfill (reprocess data which have come in since the bulk reprocessing started). Then we embark on the FITS file generation step. With a good tail wind we could be all done by mid-June.
Thursday outage: (Tom G.) There are three separate, more or less concurrent, outages:
EVO retirement (Richard) EVO will go away by the end of this calendar year when government funding stops. It will be replaced by a commercial product, SeeVogh. Heather, Anders and I have tried it out. See details in Confluence. He plans to investigate other systems as well, e.g. WebEx, GoToMeeting, ReadyTalk
Pass7 news: (Leon) There isn't any except for the geomagnetic field calculation; see above.
Pass8: (Leon) is encountering some difficulties just getting set up — for example, the event display which, after some patches to HepRepSvc, is now working for Tracy, does not work for Leon.
(Tracy) We believe the Event Display problem is now fixed. That was the last remaining new-Gaudi issue. Therefore, we should be ready for a base GR tag. Then we can check in code (from Bill, Philippe, me) which has been waiting on the completion of the merge. We would like to make another tag incorporating the new stuff by Friday, but that's quite ambitious. We should run sys tests on that tag and make a dataset for people to look at. We're getting close to the time when we'll be "done" with Recon and can move on to examining backgrounds.
(Richard) Is the increased CAL recon time going to bite us? (Tracy) It could probably be reduced by looking carefully at the code and making it more efficient, but it's not a critical issue. The average time has gone up somewhat, but it doesn't suffer from outliers the way TKR recon does.
Documentation: (Heather) We now have a Workbook mailing list and the links for feedback in the Workbook point to it. To see it you need to refresh the appropriate frame; refereshing the page from your browser doesn't always do it. It would be nice so somehow force it not to be cached.
The ISOC web pages, orphans it seems, need some minor updates. They live on the V disk in a place only accessible to a few accounts.
GR validation: (Heather, Joanne) JJ is actively looking into possible differences between B1-1-3 and B3-0-0 of handling of MIP and HIP filter bits. [Late update: he didn't find any. The ball is in our court. He thinks it likely that data are being incorrectly unpacked or presented by our code, some of which is specific to obf version.]
L1proc GR validation apart from comparing filter bits continues, where we have backed off to B1-1-3. See details in the Confluence page. System tests have also been run, revealing may differences, however the random number seeds are completely different. Heather is working to find a way to make then the same in the two versions.
Core talk:
DB migration: (Heather) Arash has requested that we limit IP addresses from which we access the various databases (for write) as much as possible. For the most part (RM's, calib, mood) we only need universal access for read; writes can be restricted to SLAC nodes. This issue is still under discussion for fsw databases and for OpsLog.
RM news (Tom S.) No news. Everything is working!
(Heather) would like a script to clear out builds within a range as disks fill up. (Tom S.) RM can be configured to keep only a certain number of builds of each type (LATEST, HEAD, Release). Currently only LATEST are automatically purged: last 10 are kept. (Heather) We could purge HEAD builds in this way. She would prefer to do release builds manually after archiving. Is it true that, when deleting a build, RM does not get rid of the Windows zip file? (Tom S.) Yes, He will look into fixing this.
GR vc90 (Joanne) As of last week I was able to create a moderately sensible-looking "all" solution file for an artificially small set of packages but hadn't tried it in VS. It was ok but more verbose than it needed to be, so I trimmed it down and also added code to create an "allGleam" solution file. It includes projects for Gleam, test_Gleam, and all libraries. The full-size versions of these files worked in Studio up to a point: I discovered a (long-standing) "feature" (all projects use the same workspace) which caused failures or strange behavior when more than one project produced a file with the same name. That is easily changed by modifying one line in the project file. It all works transparently for operations VS knows about, like compiles and links, but not for our custom commands. So far I've fixed up the bindexplib command (pre-link step), but rootcint is not quite right. swig may also be broken but that probably doesn't matter for GR.
previous
|
minutes index
|
next
 |
J. Bogart, Last Modified: