Teraminutes for the meeting of September 16th 2005

It's the boss's birthday!

Present: Chiara, Emmanuel, Yannick, Henry, Laurent, Jean-Christophe, Frederic

WIRCAM: Chiara

 The script that splits data cubes is ready and works,

 Weightwatcher does not make MEF binary FITS files correctly from the polygon files. It's important to create the masks to limit the number of spurious objects because these catalogues will be used to calculate the astrometric solution using SCAMP. There could be some kind of 'stop-gap' intermediate solution involving filtrage of objects in the rh-magnitude plane, but the best solution would be to fix Weightwatcher.

 First tests done with SCAMP with 2MASS and USNO catalogues do not show significant differences. As it was discussed in a previous meeting, we may use the 2MASS catalogue for photometric calibration. This is feasible by running SCAMP in a 2-pass mode or possibly by a modification in SCAMP. The 2-pass will be tested later by Chiara, but not a priority. A change in SCAMP is even a lower priority action.

 Chiara should define with Loic Albert the best dates to visit Hawaii; probably in November or December, when more data is available.

 Yannick reminds us that TERAPIX has promised to produce data reduction in PI-mode before the end of the year; however the camera is not yet ready.

Cluster: Fred

 Fred's overall priority at the moment is to resolve the problems of stability on the mix3,4,5,6 machines.

 On some of the pix, firmware for the RAID cards has been updated and now it's possible to control this hardware from the UNIX command-line. However, the web-interface doesn't work yet. The remaining pix will be upgraded as soon as possible.

 Fred still has to upgrade mix1 (low priority)

 Some problem of stability with the mix3,4: not sure what this is coming from. Fred has started to run memtester on mix5.

 There is same kernel on all machines; it's not clear at all what particular job causes the machines to crash.

 For example, Laurent tried make a stack on mix3, the procedure crashes during the rcp. Laurent will move to wget instead of rcp. Laurent will run his stack process (including the rcp tasks) without using RAID disks, but only on the system disk. This is a top-priority test.

 IRAF installation: Henry will define with Fred what needs to be done to install IRAF on the nodes, or at least mix1 and mix2 as absolute minimum.

 Emmanuel will move the 200GB Dennefeld data from pix6 to mix2 and will ask Riquebourg to process them before September 30.

Tests for T0003: Laurent

 Script which runs QFits seems to work OK; Laurent is ready to run QFits-in at any time.

 However, the script which produces the stacks is still being tested; this needs the data from T0002 to test.

 Laurent doesn't think it's a good idea to start the QFits-in whilst the machines are still unstable.

 Yannick would like the QFits-in to start next Thursday, so that any possible problems in QFits can be caught as soon as possible.

 Laurent still needs to do three things: make real stacks (stack definition here: organise data files and command lines for SCAMP+SWarp, SCAMP, SWarp and QFits-out); still waiting for the procedure with the .ahead; and write a procedure to grab the metadata from CFHT.

 Meeting next Thursday at 14.30 to define the software parameters for QFits in. The global action is : run QFITS-in and the Stack pipeline on a sub-sample of data, then discuss the results during the Thursday Sept. 22 meeting, modify the parameters of SExtractor, SCAMP, SWarp configuration files, if needed, then run the sub-sample again. Finally restart QFits-in.

 Date-obs in the headers in the stacked images will correspond to the observation date of the oldest images in the list.

Focal plane criteria: Henry

 Henry has still to look at the selection of LDACs to generate the .ahead files for the images which have a problem with the input astrometric solution produced by QFITS-in. Henry garantees he will provide tools to Laurent next week.

 For instance, to select observations with flag A or B, or based on² some other quality criteria.

 When .ahead files defining the focal plane are generated, they will be copied to all the machines.

Webservice: Jean-Christophe

 Data preparation function finished and tested.
 Second function which monitors the nodes has been finished.
 no need for PBS pro
 Should be ready for a demo at the end of next week.

PI data: Jean-Christophe, Laurent, Yannick

 PI data have been downloaded: Seymour, Tully and Albert (Cuillandre) ready for QFITS-in
 Should start QFits-in on these images over the weekend.
 Yannick needs the FITS tables that contain the CFHT-grade in order to TERAPIX-grade C the CFHT-grade 4 and 5 images prior to process.
 Yannick and Mireille will grade the QFITs-ed PI images and Mireille will process data with the following priorities: Tully, Seymour, and Albert.

Other matters: All

 Database backup still has to be done.
 Still not found the solution to the ganglia problem.

Status of SWarp and SCAMP: Emmanuel

 Problem with the multithreading in SWARP which produces 'lines' has been fixed. Another patch has to be applied for the public release.

Status of orders: Fred
 The order for mix7 has left, but has yet to arrive at the company.

 The new PCs for Yannick and Emmanuel should arrive before the end of September (waiting for the processor). Yannick has no PC until this date...

Next Tera-meeting: