Core Minutes 2/12/2013
Core Minutes 2/12/2013ScienceTools: (Jim) Several small bugs were fixed since last week. See Science Tools Development Notes for latest ST news.
Reprocessing (Tom G.) No news. Still waiting...
Hardware (Tom G.) ftp-glast had several episodes of going away briefly, then returning along with one longer outage. (Richard) believes its a VM.
wain25 (user disk) has been unavailable at times; it looks like overloading. Nagios provides per-user information which could be useful in understanding this.
Pass 8 (Tracy) Last week's GR tag, 20-08-07, actually came out early Monday of this week because of ROOT io issues in ACD-land and AnalysisNtuple. There still is a problem reading data back. Headline-worthy features of this tag include
See JIRA LPATE-114 for details.
In what was probably a worst-case estimate (high energy MC gammas), output file size increased by a factor of 2.8. Some of this probably came from CAL since in the new recon there can be multiple clusters. Tree-based tracking writes four kinds of output. It's likely a couple of them (3d points; links between them) can be regenerated easily enough that it's not worthwhile to write them out. An attempt to do this will probably not be ready for the next GR tag; aim for the one after that.
Johan plans to use 20-08-07 to create data sets for the Collaboration Meeting. This makes sense since issues we'll be wrestling with next have to do more with performance than functionality.
(Leon) is working with Stephan Zimmer on creating a Pipeline Overlay job. They're getting close: they may have some files in time for the Collaboration Meeting.
Might ultimately just run it routinely in the Pipeline. Output, in fact resource usage generally, is modest. Will start with 4 precession cycles (53 days, about 800 runs) spaced over the mission to look for changes in background rates, dependence on magnetic fields, etc.
People should chime in with suggestions on how many overlay events we really need. There are > 200M periodic triggers in our data. For previous overlays we've used about 20M, which seems to be too few. 200M may be too many.
Core talk:
skimmer and Pass8 (Heather) The skimmer can handle merit but balks elsewhere. (Leon) put together his own skimmer which can skim digis, even though tskim doesn't. But the regular skimmer can only do recon and relations, not digi or MC (even though recon would seem to be more complicated). It could be that no one has ever tried to skim digis with SCons build and the new ROOT.
Next GR tag (Heather) will contain CAL cluster work and will use the new obf external B3-1-3. (Joanne) finished it off last week. Builds exist for redhat5 32-bit and redhat6 64-bit. Some routine code changes were required in OnboardFilter and OnboardFilterTds. Everything has been committed and tagged.
Mac (Heather) Yemi has spec'ed out a box for us. We can run Mountain Lion on it. (Richard) needs to check that we have the money to get a couple.
Mac, RM, and Jenkins (Tom S.) This morning Tony J set up Jenkins to use both bldmac01 and bldmac02, allowing up to two jobs per machine. Tom will try this out. So far he's only used Jenkins for LATEST; he'll move all mac builds over there soon.
Windows and u52 (Heather) Wayne is looking into the problem we've had with permissions on files written to u52 from Windows.
previous
|
minutes index
|
next
 |
J. Bogart, Last Modified: