Core Minutes 4/23/2013
Core Minutes 4/23/2013ScienceTools: (Jim) No news this week.
FSSC: (Tom S.) Event server has been updated: we added weekly all sky files that have the same events as the photon data (FT2 files, SOURCE class and above) but contain all the data columns form the extended/event data (FT1 files).
In spite of continuing problem with 32-bit Fedora, FSSC will proceed with next release. 64-bit, which is what most Fedora users care about, is ok.
Pass 7 Reprocessing (Tom G.) No movement in reproc itself; work to understand IRFs, different event classes, etc. is ongoing.
As part of support for new IRFS, Johann submitted an MC task which was throttled by a 400-cpu limit. This (default limit) is a feature Brian recently added to the pipeline. It can be overridden (either for all tasks in a group or per task), but that takes some doing.
(Richard) When might we be ready to start up the Pass 7 reprocessing task? (Tracy) might have an estimate by the end of the day.
Hardware (Richard) Purchase req for new file server has been submitted. This is a bulked up server — 340 TBytes per server rather than 90. We had just barely enough money for one.
Pass 8 (Tracy) GR 20-08-12 is the Recon freeze candidate. By default it still uses standard hit errors but has job options allowing you to switch to candidate alternatives. It includes code to eliminate very large events and it's effective: tails are gone. There is a slight hit on high-energy efficiency, but people are ok with that.
Bill is investigating error estimating alternatives and thinks we should switch. It's just a jo parameter.
Leon found and fixed a bug; still to be committed.
The only development area at this point is incorporating alignment into pattern recognition. (Leon) has a first round of all code needed to introduce the alignment corrections at the patrec stage. He's having problems updating CVS because this project has been going on for a while, with most of the work being done on Linux without benefit of elegant diff output. This work needs to be integrated with Tracy's fixes. Preliminary look: this won't change the results very much. He doesn't think it should hold up the freeze. It can be added in later if necessary.
(Tracy) doesn't foresee any special difficulties in integrating their changes. We need to shift gears and work on non-recon things (e.g. output size) before reprocessing starts in earnest. At some point need to meet with Tom G. to figure out details of running. (Richard) When? (Tracy) nothing to prevent attempts now. We have a candidate. (Richard) When will you be ready to run for real? (Tracy) had been assuming 2 weeks after recon freeze. (Richard) Say mid-May.
Core talk:
Root compression (Heather) Stefan tried experiments with compression, all the way up to level 9, the most (normally we're at 1). Didn't see any big differences in time reading files; perhaps will in write. We could try a (new) alternative algorith, LZMA. We would need to modify RootIo a bit to use it. Heather is working on it. Expect to see a hit in read-back; maybe also writing.
Fred lifetime (Heather) SCCS is moving to Ruby 1.9. Could affect Fred, of which there are still at least 2 users: Leon and Tracy. (Leon) would like to keep Fred around if possible. (Heather) No problem if it works for this new Ruby. But there will be no new development; we'll just keep it around as is. (Tracy) There are still features of Fred which we will apparently never get with WIRED, although WIRED is better in other respects. (Tracy) Would also like to see it set up on fermi-ts as well. (Heather) intended to and maybe already did. Ruby is there. If it's not in path by default she can fix that.
fermi-ts; Windows and SCons (Heather) Any news from Bill? (Tracy) hasn't heard anything. Presumes he is either happy or off doing something else. (Heather) will install new build incorporating Joanne's addition to create alluserAlg.sln file.
CMT builds (Heather) We don't need LATEST any more, but we still occasionally need to make L1 branch tag builds. These happen on the old SLAC boxes. She expects Toby will be asking about shutting down glast-ts.
RM (Tom S.) No news about Jenkins; he needs to talk to Tony. He did figure out how to make the tests not pop up the alert dialogs but it requires changing the code on all the tests and I need to determine a central place to put the header file that needs to be included.
Moving Moot db (Richard) Any closure? (Joanne) Has not heard anything recently from Tony. Currently the db is hosted on glastlnx06 and is accessible only from the SLAC internal network. There is a mirror on mysql-node03 which is accessible (read-only) from everywhere. glastlnx06 is going to be shut down soon. Tony will move the db to one of the new fermilnx nodes but was agonizing over how to handle the mirroring. Apparently no one uses the db remotely (no response to my mailing list query, at any rate) so it should hardly matter to us what is done about mirroring. Those who access the db locally, including L1, do so via the aliasMOOD. That will move to the new node along with the db. Of course there will be an interruption during the move; M.E. is aware of this.
Pass8 worksheet (Joanne), mysteriously, while working on a pipeline job to make overlay files, Stephan ran into a problem with the Pass8 worksheet in extFiles v0r12. The xml parser complains about something which is in fact improper (an unescaped & character which should have been written as &). All versions of this worksheet have had this incorrect syntax; it's odd that it ever worked. (Heather, Leon) Preferred procedure to fix is:
Leon will handle step 1. (Tracy) Only Bill used IM; he may have dropped it by now.
previous
|
minutes index
|
next
 |
J. Bogart, Last Modified: