Minutes for Tera-Meeting of Friday 19th March 2003. Present: everyone, except Mr. Laurent Domisse.

SCamp (Emmanuel) :

 Main problem at the moment comes from cross-identification, it is difficult to carry out accurate pattern matching at the edge of the field. How to fix this? One solution: select only brightest sources in the catalogues or make matches in flux ranges. For the global astrometric solution, there is still another bug which might be a numerical error.

 For further testing: will use .ldac files from the Pre-survey, for photometric/astrometric calibration and testing. In parallel, work will start on the photometric calibration code, using a similar algorithm developed for EIS. Will make use of phot_c and phot_k photometric keywords supplied by CFHT, as well as flagged photometric images, to compute the solution. In order to check the full process, validation will be carried out using the 1x4 deg2 W3 data that Mireille processed earlier with Astrometric/Photometrix.

 The QRunID keyword as the instrument Keyword for the SCAMP configuration file.

 Emmanuel will provide the tools for XML output and PLplots, and the package will be delivered to Laurent for SPICA implementation.

Defectix (Anthony):

 Halo generation works, for the minute try with only one CCD. For an entire MEGACAM image, learning mode can take an entire weekend. More soon.

SkyWatcher, datatransfer/Snoopix (JC):

 Nothing new. Some problems with Jerome Bouvier data, problem with CRVAL (z-band).

 Jean-Christophe did the packaging of the Microsoft SDSS HTM library. Spica packaging: SWarp, SExtractor packaging easy, other stuff requires more work.

 Question from JC: what about the corrupted headers produce by SWarp? Henry confirms that this problem is not Opteron-specific, it occurs on *all* machines. Emmanuel says that is a bug, will attempt to look at is as time permits. Header corruption appears to occur on *any* swarped image.

SExserver (Gerard) :

 Idea is that this software should be installed in the place where there are big images to produce automatic catalogues. Idea is to have a first prototype in 1-2 weeks. Gerard will have the opportunity to talk to Francois Bonnarel this week end Mark Wenger the week after.

 Dbclient: make tests on mix1 for 64bit readiness, but for the moment stay with clix.

Data processing:

 Data Sophie Maurogordato (Gilles): No progress. However, scripts are in place once all machines are running again. NEED to get the master flat field and mask and to run QualityFITS on Sophie's data to get the weightmaps.

 Data transferred to TERAPIX: Half of the Bouvier data has arrived. There is apparently another 30 cosmos i* images on their way.

Status MDA proposal (Emmanuel):

 Resume deadline: 21 proposal deadline, 29th

Cluster software upgrade (Fred, Emmanuel, Henry)

clic2 on clix: *many* installation problems relating to hardware incompatibilities. These hardware incompatibilities have forced use to use the old 2.4.21 kernel. Unfortunately, there are still problems mounting all the fibre channel disks.

Opterons: hardware problems are (probably) resolved, after kernel upgrade to 2.6 and firmware upgrade (Henry successfully carried out several swarp jobs lasting several hours). These machines are now running Gentoo Linux (and not Mandrake 9.0 like the other systems). 6 GB ram available on each machine, plus two raids of 1.7TB. However, there could be problems integrating these machines in the cluster as they are running a different version of the kernel (and not all software is fully ported to AMD64). All data on the mix1 raids will be removed and the raids configured as RAID-0 for 64bit database testing (Henry confirms this is OK).

To add the Opterons to the cluster requires that we buy a new switch, as there are not enough ports on the old switch to accommodate all machines. Approximate cost: 1000eur, TTC. THE SWITCH should be ordered next week.

cosmix: need to do a system re-install. To do this, we will install a new system disk, as currently the system is on the RAID partition (which is shared with user data, which would have be to be backed up and restored). Fred CANNOT guarantee that all data will not be lost on the raids during system upgrade.

Summary and priorities for Fred: Astrometrix install / fibre channel / /opterons/ cosmix [note added in minutes: astrometrix now successfully installed].