Big run: As always, follow progress on the Big Run Checklist. (Richard) About half of the 5 days' worth of Level 1 processing is done. The remainder will be done after the outage on Thursday.
Stress testing demonstrated that
The main bottleneck concerns an afs buffer used to merge recon files together, among other things. SCS will look into providing multiple buffers to reduce the contention.
Sky Tom is getting ready for Interleave. We expect to start generating the sky after the collaboration meeting.
GRB grid There has been a memory problem with these runs. (Heather) A likely reason has been found: the increased basket size put in a while back.
Re-recon: (Leon) Progress is being made. (Toby) We need to read in a digi file and run recon, but not in the context of MC. There is no need for FluxSvc or the way it manages pointing information, in fact it's incompatible with this mode of operation. The job options file Gleam/src/jobOptions/pipeline/readigi_runrecon.txt does approximately the right thing.
CEL, Interleave: (Heather) Before leaving on holiday, David had committed updates to rootUtil. This has since been tagged v1r3 and after conferring with David, it was agreed that the Interleave updates would be best served by using this latest version of rootUtil. Heather has started re-activating the CEL code Tracy introduced in Interleave some weeks back. Richard wondered if we had a good understanding how we will keep the file trail straight in the CEL files for the Interleave. Eric Charles originally created some standalone test applications that can be used to generate CEL files given input mc, digi, recon, etc. David has ported versions of these applications into rootUtil. Our best bet may be to run one of these tools to regenerate CEL files are necessary.
Disk clean-up: (Richard) See details in Confluence.
Science Tools report: Jim went through the Science Tools Update for Feb. 26.
Data handling: (Dan) There was some fall-out from the stress testing for Data Handling as well: Wilko is looking into issues with xrootd while Dan concerns himself with Data Catalog.
Max gave a preview of the ASP page.
Computer Center outage: (Richard) This Thursday at 9 AM SCS will take down many services for hardware work better done now than left until even closer to launch. In a perfect world they would be done by about noon, but assume it will take the day. There will be no batch farm, no CVS, no u33 user disk, no calibration database, etc. Confluence will also be down but we've asked that they put that off until 10 AM if possible so it will be available for morning meetings.
Documentation: (Chuck) is happy to be back [the feeling is mutual, Chuck] but has some catching up to do. He's working on the next newsletter and Science Tools updates, which he expects will take 1-2 weeks. Several other projects are lined up behind those.
GR news: (Heather) Late last week Eric C. made updates for ribbon attenuation. It's in the tag, v13r12 and appears to be fine. Heather is collecting updates for yet another tag. If there are things you know of which belong in a production release please point them out to her.
GCRCalib Exceptions are occasionally thrown from G4Propagator as invoked from GCRCalib. Johann has added some exception-handling which should help in pinning this down. We await the experts' (Tracy, Claudia) return.
Composite Event Lists, Skimmer David Chamont is on holiday. Before he left he made a new RootUtil tag which extends the write interface of CEL. He has also completed a first draft of a bridge between CEL and the Skimmer.
ROOT 5.18 Navid has installed it at SLAC. Heather is hopeful we will be able to upgrade to it with minimal disruption.
SCons status: (Heather) We had a meeting last week and there are substantial minutes to prove it. One issue which ought to be settled soon is how to deal with all the LATEST-.. tags now being generated for each LATEST build [all packages in ScienceTools get another tag every time there is a LATEST build]. They clutter up the view from ViewCVS and also from MRvcmt/MRStudio when one asks to see all tags associated with a package. There are at least three ways one might deal with this:
3. seems overly drastic and possibly disruptive (though there was general agreement that it would be adequate for RM to check every half hour rather than every quarter hour for new tags). (Jim) suggested we amend 1. to make a cut by number rather than by time (keep the N most recent LATEST tags); Navid also prefers this. 2. is also a possibility but probably a little more work than 1. (Heather) voiced some concern over losing old LATEST builds; sometimes they can be useful. (Navid) The LATEST builds would not be lost, nor the information in RM web pages referring to them; what would be lost is the ability to check out an arbitrary LATEST build with a single cvs checkout command, specifying a particular tag. (Joanne) Perhaps, when one of these LATEST tags is made and applied to particular tags of the contained packages, a list could be kept of those tags, essentiallly containing the same information as is now in the container package CMT requirements file. If just this list were saved, it would be possible to check out all the code belonging to the possibly defunct LATEST tag at any time.
(Navid, Heather) Windows builds were not working because, for packages building shareables, no .lib file was created. The SCons-recommended way to deal with this is to make use of (Microsoft-specific) __declspec syntax to describe which symbols need to be exported, so this is what Navid would prefer, to keep our use of SCons as vanilla as possible. (Joanne, Toby, possibly others) are concerned about the volume of changes which would need to be made to source code to implement this. Doesn't seem like something we want to do, especially so soon before launch. (Toby, seconded by Jim) believes it would not be difficult to get SCons to run the ROOT bindexplib tool or other similar thing as part of the link step, as he subsequently explained in email to the Infrasoft list.
(Richard) Contrary to earlier reports, Tracy finished 6th! Currently he's in London on more Laser business. He'll be back on Monday.
previous | minutes index | next |