Minutes for the terameeting of October 1st 2007

Present: Mireille, Fred, Gregory, Henry, Mathias, Yannick

Yannick's summary document

List of all Megacam FITS images forming the 2 T0004 3x3 W1 and 3x3 W4 ugriz patches that will be used by CFHT as a re-masterised set of data test (preparation of T0005 input images)

Megacam PI (Mireille, Emmanuel, Mathias, Yannick)

 Mireille did the astrometric calibrations of 05AF14 and 05AF09 as 07AF18 and started Swarp-ing. Swarp terminated for the 15 stacks of 07AF18, and still in progress for 04AF09 and 05AF14 using the new astrometric calibration.

 Annie Robin' data astrometric calibration with 2MASS: Emmanuel found CDS recently changed the format of their archived 2MASS catalog. The magnitudes and the errors on magnitude read by Scamp are therefore wrong and turn out to be almost zero. This explains the very high external chi2 value found by SCAMP and the negligible weight of the internal solution. The relative accuracy is then expected to be poor, as found on Annie Robin's data. The efficiency of the matching process is likely degraded also. Emmanuel will modify Scamp accordingly once set in Urbana.

 Emmanuel asks Mireille to run a 2-pass Scamp by using the GSC2.2 catalog instead of 2MASS during the second pass. This will also check if the 2-pass process does work well.

 Mathias will come to Terapix for a full week to learn how to process MegaCam PI data. Mireille and him will agree on a slot in October or early November.

WIRCam PI (Chiara):

 07AF34 (WIRDS Kneib): Chiara started to download these data yesterday. In progress. She asked Henry to check the if the other RUNID's requested by Willott and Kneib ( http://clix.iap.fr/spica/pi_processing.php ) are available.

 Henry will clarify the status of COSMOS D2 stacks that comprise data from 3 PI's, Kneib, Sanders and Willot. It is important to understand what could be shared between them in order to avoid conflicts of interests , misunderstanding and eventually embarrassing position of Terapix with respect to its servicing tasks to all.

 Bouvier data: we understand from Chiara's email that she is willing to take over the WIRCam+MegaCam join processing (Chiara could you please confirm?). Fred upload Bouvier' MegaCam data and started QFITS-in (46 images). Mireille will then grade each image prior to leave the whole set to Chiara.

DB T0005 (Gregory, Fred, Yannick)

 A long discussion on item 3 of Emmanuel's summary regarding the URI usage and the transparent access to files spread over all Terapix raids. Fred pointed out the best way to handle raw images on the raid disks. He summarises this as follows

*** Each image and its location protocol/node/directory (i.e. URI) is listed in one single data base

*** Only scripts are permitted to handle these files and to update the data base accordingly

*** A daily cronjob checks the database consistency (by checking if all files listed in the data base do exist on disks, checking where it is listed and by checking that all physical files - listed by a find command - are listed in the data base).

*** An email should be sent in case of inconsistency.

 Gregory pointed out the impressive potential of Django. He stressed Django can be fed by a pre-existing database, but most up to date fields, like "polygon" . So, Gregory cannot guarantee that at this stage Django will be sufficient to handle all fields.

 Fred will provide the script to handle URI with Condor before the end of this week.

Cluster/Hardware (Fred)

 Fred will replicate mix8 on mix7 before then end of the week. He will send an email before. He will also stop several raids (like those on fcix2) that seem to have corrupted xfs .

 The missing pieces of hardware for the new rack arrived today. Fred will install the rack before the end of the week. He will then remove pix2 and cos1/2/3/4 from the cluster.

 About the following question from Chiara: "Fred a installe condor sur fcix3 fcix ftpix fcix2 : est-ce que il y a encore besoin de separer les deux clusters terapix et Condor pour c3? Sauf si la modification fut faite effectivement au cours des derniers jours, le default de cexec etant le cluster 'condor', les logiciels installes ne le sont pas par default sur toutes les machines sur lesquelles Condor tourne." . Fred prefers the two clusters be separated since we are still in a test phase. Regarding the software, Fred checked that all nodes have the same version of software , but mix9 and mix10 for known reasons.

 Fred started the investigation of hardware option (Barcelona CPU option) and will set the PUMA this week.

Misc.

 Chiara reported she met Mireille Louys at ADASS who asked the status of QualityFITS public release.

Next Tera-meeting: TBD. NO terameeting next week because of very busy week (ESA Cosmic Vision and SNAP meeting)