ScienceTools: (Jim) gave the report; see it in Confluence.
FSSC: (John) has been recovering from hardware problems in the data server; no other news. (Toby) would like to be able to select, e.g., just diffuse class. (John) Others have requested this. It's #2 on his list of things to do.
(Richard) FSSC made use of its dual injest capability. (Tom) "Dual ingest" refers to the ability of FSSC to simultaneously accept normal Level 1 output and new, reprocessed data. The two data collections are kept separate. The public will see only a single collection. This scheme should significantly reduce the FSSC downtime associated with the introduction of reprocessed data.
GR: (Tom) is gearing up for event classification reprocessing with Pass7.2. This will be done three times in sequence: once for v02, once for extra diffuse, and once for data clean. [late update from Tom: It just got underway! This is a trial intended for extensive validation. It is expected that we may need to run the whole thing again.]
(Tracy) The first in a series of bi-weekly Pass8 EVO meetings will take place this Thursday. They will include an Italian contingent in addition to the Californians. See overview and plans in Confluence.
L1: (Richard) The L1 slow-down problems mentioned last week were largely but not entirely cured after Thursday, when several machines were rebooted and certain other changes were put into effect, but exactly which actions cured what is not certain. (Warren) The MySQL connection failures appear to have been associated with a particular AFS server. When an AFS volume with certain libraries was moved to a different server, the connection problems went away. Bad configuration (too small cache) on rhel5 machines affected bandwidth. Now that that has been fixed, worst case is better, but typcial bandwidth is still a factor of 5 or so worse than on rhel4 machines. (Tom) The bad cache configuration also applied to taylored rhel5 desktops. (Richard) Silver lining department: SCS's willingness to commit serious resources to solving these problems has been gratifying.
Documentation: (Chuck) sends the following:
I'm making progress on the rmfit portion of the gtgrb - rmfit tutorial. I have successfully installed the IDL Virtual Machine (VM) on SLAC Public, and these instructions can be easily translated for installation on desktop computers.
I ran into a stumbling block when trying to install rmfit in that it requires Fortran 95 (or later ?) to perform Make.mfit.sh. Unfortunately, the new noric RHEL4 machines appeared to have only f77. Late yesterday, I discovered an f95 version residing on noric05, so hopefully will be able to peform a successful install today. From there, it should be a relatively straightforward path to finishing the tutorial.
GR (Heather) For L1 processing we're now at v15r47p12gr09. It uses ROOT 5.20 and a newer tag of rdbModel implementing retries when initiating a MySQL connection.
See this JIRA concerning time translation in astro. (Warren) When timestamps are converted from MET to UTC while making FITS files, a timestamp in the last second of the day (rare, but it has happened) gets put into the following day, with hour set to -1. (Toby) will look into it ASAP.
(Heather) Picked up most of the latest package tags available at the time when the latest GR tag, v17r37p0, was built. Includes Leon's tracker geometry/mass model work.
She made a HEAD build off the GR v17r35p1 branch (pass 7.2 branch) which includes the ability to turn off specific (ROOT) branches when reading, in an attempt to address reprocessing performance problems. (Tom) has not yet had a chance to try it.
rhel5 builds (Heather) Do we know how important it is to get rhel5 going for Lyon? (Richard) They can now use rhel4 binaries. They also have a new person to help out with Pipeline there, in fact the new person had a part in discovering what was needed to make rhel4 binaries usable.
RM (Navid) Concerning package updates needed for 64-bit rhel5 — there had been some misunderstanding of who would do what but that has been cleared up. Eric will talk to package developers about necessary changes. After these are implemented we'll see how much more work there is to do.
(Navid) is done documenting the CMT RM. An SCons RM doc is in progress.
SCons and Windows (Heather, Toby, Joanne) It's time for another attempt at using SCons for development on the UW Terminal Server. Joanne will revise list of necessary items to install.
GR and SCons (Joanne) Only 5 packages to go! A discussion of remaining issues is available in Confluence. (Heather) is well on the way to being able to build Gaudi for the vc90 compiler. Unfortunately the newer version of Gaudi she's working with depends on ROOT, so every time we move to a new ROOT version Gaudi will have to be rebuilt. However the good news is that the number of externals Gaudi depends on is small and all can already be built on vc90. She is hopeful that Gaudi and other externals not yet available for vc90 will be built by the time she has to go on leave.
ASP and SCons (Joanne) It's very close. There is a new CVS module, ASP-scons. After checking it out and setting up GoGui to use a recent ScienceTools LATEST build as its base installation and the checkout of ASP-SCons as supersede, all packages built successfully. There are still some minor problems with wrapper scripts to be worked out; then the build will be ready for Jim to test.
previous | minutes index | next |