Minutes for the Terameeting of 8th of December 2005

Present: Yannick, Mathias, Anis, Fred, (for part of the meeting), Mireille, Henry, Laurent

First resolution, Fred will arrive at 13.00 every day from here on!

Status of T0003

 Usage of machines: currently the release has started, but only mix4 (deep) and mix6 (wide) are being used. Not optimal! Is there any way to process more stacks on mix3, which is not being used at all? (mix5 has some hardware problems and mix7 arrived too late to be integrated into the cluster).

 Uneven distribution of jobs between the machines is due because the data was not distributed correctly at the beginning.

 Laurent started work on a method to move stacks from one machine to another, but this has not been finished. Need further tests prior to be implemented. If this works, then it will be possible to proceed during the release.

 Laurent stressed again that while we could also move the Qfits-in from one machine to another, this solution is not feasible while T0003 is under way.

 Some problem with the priority of the wide stacks, which are not processed in the correct order; Laurent will investigate.

 The astrometric solution for the wide is worse than before; we need to compare in detail the astrometric solution in T0002 and T0003. Henry and Mathias will do this by compraing ldacs from T0003 and T0002 stacks using SCAMP with the -SOLVE_ASTROM NO option.

 Mireille has already made a plot in gr-ri for one of the wide fields, which looks the same as the T0002; there is a small offset wrt to t02, but this is less than 0.05 magnitudes.

 Mathias will produce tracks ug-gr and gr-ri and ri-iz for the wide survey fields computed using correct metallicities. They will replace the Pickles data used for T0002 and T0001.

 There is a problem with some of the wide stacks because there are u* images in the wide stacks with poor seeing which have a very bad initial astrometric solution; this causes problems in the global solution. These images were taken by CFHT as 'photometric' exposures to calibrate pointing that were observed during no photometric conditions.

 We will re-evaluate the status of the release on Tuesday and make a decision. We need Mireille, Laurent, Henry, Emmanuel, Mathas (if possible) and Yannick for this meeting.

Status of machines (Fred)

 Many tests were made on mix5, and it seems that there is a real hardware problem with the machine. Many tests with swarp produce errors in the logs. Unfortunately, it seems to be some complicated problem not coming from a particular memory barette, but rather the interaction between the memory and CPU, and is therefore very difficult to isolate.

 mix5 will be sent back to option+ as soon as possible once Fred has copied all the data to mix7. Yannick point out that mix7 is also for EFIGI so it is recommended that TERAPIX shouldnot to use all CPU and disk space on this machine

 mix3 is still connected a broken UPS, need to stop this machine and plug it directly into the power supply so the UPS can be sent back for repair.

 yannix: 3ware raid card doesn't work correctly in yannix, no way to correctly mount the raid array on the machine; we should send this machine back on Tuesday if Fred has not been able to resolve the problem in order to avoid wasting too much time on this problem.

 need to buy new disks to replace the ones removed from the RAID arrays

 Naming of objects in the release. Should we produce names for each each object in the catalogues? Easy to implement, but there will be a non-trivial problem of maintaining constant naming conventions for successive releases.

 It is decided that merged catalogues will always include all filters, even if one otf them is missing. If so, the data value will be "-99".

 Need to buy a new battery for JC's laptop.

Comparisons with the Sloan Digitital Sky survey (Mathias)

 Mathias has made a matched, merged catalogue for the full wide survey
 Photometric and astrometric comparisons made with the sloan survey
 Some zones show an offset. For the i* band, there is no offset over full 9 deg2 at less than 0.05 magnitudes. In g, r there are some zones with larger offsets

 For the astrometric solutions, there is a large offset at the borders, indicating that it's really important having a pre-survey which covers a larger area than the main survey.

 Mathias will re-do these tests but only using objects OUTSIDE masked regions (currently all regions were used). Bright stars could be producing some of these effects.