Teraminutes for the Terameeting of 6th of May 2005

Absent: Laurent, Mireille

 Yannick also mentions that Gilles did not demount the rack at the end of last week, which delayed the installation of mix5. Yannick asks that his requests are followed promptly so that delays do not accumulate.

Hardware (Gilles, Yannick, Fred, Emmanuel)

NOTE: every time a "bon de command / bon de livrasion" this information should be copied in TRIPLICATE : one for Fred, one for Emmanuel, one for Gilles.

 4x400 (Hitachi, spares): 2x400 (Seagate for EFIGI) and 1x250 for the cosmix2 system disk came today. Gilles has these disks. Fred asks that the serial numbers of these spare disks should be added to the TERAPIX web pages.

 7 disks of 250 giga arrived 14 days ago for pix10 in order to organise backup.

 The alphas have been disconnected but they are still in the salle informatique. To dispose of this equipment we need authorization from the IAP which we don't have.

 We will decide on WEDNESDAY what kind of bay fibre channel / raid array to buy to store incoming / outgoing data

 Unfortunately, mix5 (which has arrived last week) is unstable: Fred is not able to install the system without crashes: it seems there is a problem with memory. Fred will investigate, Emmanuel is ready to help.

 The ftp server will arrive next week.

 3ware card from mix4 (mix2) returned to 3ware:

 cosmix3: Yannick notes that there will be 3000-5000 euro available from the PNC funds for this machine: not sure yet if we will be able to afford a machine with a large RAID array

Photometric offsets between the fields (Henry, Yannick, Emmanuel)

 Photometric offsets have been computed by Henry using galaxy counts and the location of the stellar locus

 Yannick is reassured that in general, the size of these offsets is small (for all fields with two exceptions it is less than 0.05 magnitudes).

 Yannick proposes that these offsets be presented in a table to the users but that we do not apply them to the catalogues or images (as they are small and it is possible (though not likely) that this signal could be of cosmological origin.

 Question from Emmanuel: In some cases why are the offsets always negative for a given field? Could this be it cosmic variance?

 Henry will apply the offsets calculated from the galaxy counts to the images / catalogues and then recalculate the minimisation of the stellar locus to see if the offsets have been reduced

 Henry will also investigate if the 'outlier' fields are in a certain location on the sky

Validation of data (Jean-Christopher)

 Jean-Christophe has validated around 100 images each day,

 Yannick notes the images noted "MISSING" are re-processed but at the end of the processing, if these observations still have completeness plot missing, what happens? If they are re-processed, we will never converge; if they have higher priorities than all in the queue each time, this is even worse.

QUESTION for Laurent: how are images put in the Spica queue when the completeness plot is missing? Is it done automatically , after checking the png/ps plots are not in the DB? Or is it after reading Yannick's comment "MISSING"? What happen if the plot is still missing after a second pass?

 Yannick has started re-noting the deep which were marked "D."

 Yannick leaves for China at the end of next week. The processing of the deep images, which will serve as reference for the other fields, has not terminated yet, and it is not sure that this will be finished before Friday: this is a serious problem. [Yannick away from Friday 13, afternoon to Sunday 22]

Defectix Anthony

- Problem from last week fixed (reading swarped images) - Gilles has prepared a script to test defectix

- Yannick suggests that first it would be better to test that the mask generated by the swarped image and the mask generated by the non-swarped image (MEF images on input)

 Anthony will update the users manual for defectix.

Other matters: - There are 12,000 images in total; we will need to re-run qualityFITS on all images for the next release. This is because CFHT have changed the flat-fields and reprocessed images: for this we need to re-run quality fits on all available data.

 Is qualityFITS optimised for machines with multiple processors? There are several single-threaded processes in the pipeline.

 We will try to have T0003 for the middle of September (August doesn't seem to be a realistic target).

Next Tera-meeting: