eraminutes for the meeting of 28th February 2006 Present: everyone. Fred arrived at 10.30. Henry left at 12.00

State of pipeline (Laurent)

 VW has to be restarted every few days because of the various problems with clix (which haven't been understood).

 Clix has worked well until the end of the Deep/Wide processing ( 20th January), since then the SPICA cluster has become unstable;

 For example: there are some processes which take 100% of CPU

 There are also other problems with accessing the fibre channel arrays, which causes the machine to lock up

 mix3 and mix4 are also unstable which causes a lot of problems for Laurent, as the coherence of the database must be verified after each crash

 one idea would to be to make a backup database; which Fred will try to implement, using clix2. According to Laurent software tools exist on the web that could send DB queries on clix2 if clix crashed.

 Another possible solution is to make a periodic copy between the fibre channel disks and the processing nodes; this could either happen regularly or right at the end of the processing
 It would take around 2.3Tb (compressed) to store all the very wide images; one idea might be to copy half the data and then continue.

 Note added after the meeting: it turns out there is enough disk space to implement this solution: all images will be copied to the fibre channel arrays at the end of the processing

 Very-wide will have to be restarted with the new version of SCAMP; Emmanuel will try understand the origin of the remaining problems prior to release. It is then expected we re-start the VW by the beginning of next week.

 Yannick, Laurent and Fred will meet to setup the T0003 summary table produced by Laurent.

Visit to CADC (Fred)

 Data products to be transferred to CADC: fits images, weight maps and catalogue (Step 1)

 final stacks and weight maps: the Qfits output and colour-colour plots and metadata will stay at Terapix. Quality flags (QFITS grades and comments) will also be transferred, CSV format acceptable.

 a discussion about the form and content of the catalogues provided to CADC

 CADC would like it to be clear in the name of the catalogue the detection image and the measurement image.

 we will make available a large catalogue which contains all the columns for all catalogues (and removes all the redundant formation). In order to clearly define values per filter, each column shoudl be defined as PARAMETER_FILTER (e.g. MAG_AUTO_u).

 Fred will provide a web page containing all the information on filenames which is the 'practical' version of ICD

 CADC ask to have a release reference number keyword in te Fits header (e.g. TPX_REL), a list of images used to produce the catalogue (i.e. the stacks).

 Fred suggests adding the effective area in to the header of the images, in the header of the .reg file and in the header of the catalog.

 the VeryWide release poses the smallest numbers of problems at the CADC

 Emmanuel suggests creating a new version of the chisquared image which manages better negative image spikes

Wircam (Chiara)

New data and stacks have been made available to JGC but he hasn't retrieved them yet; no feedback.

 Waiting for COSMOS- WIRCAM data which will be finished for the 3rd of March;

 Chiara is working on the documentation

PI Megacam (Mireille)

 PI data problems with M31 which was taken with short and long exposures; discussion with Mireille. It is likely Mireille had similar problems with Loic Albert' data. Seems a bug in the photometric rescaling step of SCAMP. Emmanuel will check.

 Brent Tully contacted Yannick on March 3rd about his data, so we will likely send data by March 7.

 Laurent has finished to Qfits 04BF28 (Dougados data), Yannick graded them. Astrometric and photometric calibrations and stacks are underway (Mireille).

WebService (Jean-Christophe)

JC is working on it. Yannick propose to relaase by March 3. Note: delayed, because Yannick had no time to check. Expected next week.

Visualisation: (Anis)

 Anis worked on the the best way to handle image contrast and luminosity with Shader or OpenGL. Though it seems easier with shaders, OpenGL seems preferable since shaders only work with NVIDIA boards.

Next meeting: MONDAY March 6: Salle du Conseil, starting 10:15 am