Terapix Star Formation Region IC 1396, © 2001 CFHT
Teraminutes for the meeting of Friday November 26th 2004
Article
by HJMCC - Updated July 5th, 2005

Minutes for the Terameeting of Friday November 26th 2004

Present: Mireille, Emmanuel, Fred, Jean-Christophe, Yannick, Anthony, Gilles, Henry

Absent: Gerard, Laurent (on holiday)

Urgent matters:

-  1Gb of ram is missing from pix8: to allow Laurent to finish to the scamp jobs pending on this machine, we will temporarily remove memory from cosmix and add it to pix8. At the same time, the defective memory will be returned to Option+ and when it comes back, it will be returned to cosmix.

-  "Private" data processing at Terapix by Annie Robin and friends: these people have asked Yannick for an account on the TERAPIX machines to "test new reduction methods " and to try re-processing some stacks (it is not completely clear at this point what they want to do). Not sure if we have enough resources to provide this service to everybody: for the moment we will provide them with access to the "science" machines (cosmix and cosmix2 when it arrives) and not the production cluster to see how that goes. Eventually we will reassess the situation to see if this really works.

Release T0001:

-  Emmanuel will change the colour images TIFF to PNG.

-  Other matters. Mireille has problems ftp'ing from the TERAPIX machines, not sure of the origin of this problem.

Status for the next release T0002

-  Laurent has made 61 stacks in four days. (Yannick mentions we need to have some way of seeing all of the stacks together, because it is hard get an overall picture of the processing. Plus querying clix is very slow).

-  1. connection with the CDS: Can we install the USNOB1 locally? During a long spica run, stacks crash if there is a problem with the access to the internet or the CDS catalogue server (this problem simply comes from the fact that cdsclient hangs if there is no access to the CDS catalogue servers). Emmanuel suggests adding a cacheing service where scamp stores locally the query results for each field. Or could we add a general web cache at TERAPIX to store these results locally (JC will investigate this). Maybe could add a list of servers in scamp? But in the end, however this is not a top priority as we have more problems with stuff like instabilities in the NFS cross-mounting.

-  2. cross-mounting problems, stack processing notes: spica stops because the cross-mounting fails and then corrupts the database. The cross-mounting failure seems to happen every once or twice every 24 hours. The big problem is remove these false entries written in the database which is very time-consuming (Gerard needs to optimise this interface to allow multiple files to be removed). Perhaps Laurent should verify that the database content is correct before attempting to commit it in the database?

-  3. photometric problems: apparently there are two stacks which don't have any photometric reference. How is this possible? Not clear how we can interpret this: the blue labeled referenced image may not be there because the second pass isn't carried out? Is it possible that the two images have exactly the scale scaling value and could both be the reference image? What happen in that case? Laurent ?

-  4. comments from Emmanuel about the photometric errors. For stack from the very wide we see very clearly photometric offsets which probably come from flat-fielding errors. However, this might come also from the effect that the MAG_AUTO magnitudes for stars in the centre and for the same at the edges is not the same, further investigations are needed. From field-to-field the error 0.01, but over many fields this can add up to 0.1 magnitudes. It is also clear for the wide and very wide data fields we need as many photometric reference fields as possible (only one is clearly not enough).

-  5. Emmanuel and Yannick also suggests that we should try to impose several photometric reference images when the field of view become very large. Could this be implemented in Spica?

Defectix (Anthony)
-  Anthony has made some more tests with the machine learning in Defectix. Also some with Fred to install the multi-threaded version which should of course be much faster on the opterons.

Skywatcher: Jean-Christophe
-  VO table functionality fully implemented.
-  Yannick also suggests creating a new interface with one map of the sky for each filter.

Hardware (Fred, Emmanuel):

-  mix4: Fred has installed an OS but the system is completely unstable and crashes randomly. Several other operating systems have been tried but none works satisfactorily. (32 bit systems can't be tried because no support for the disks). Plan of action involves removing memory and 3ware cards and testing again. In the limit, the engineer from Option+ will come at the end of next week with a replacement motherboard if we fail to resolve these problems.

Other matters

PI data Seymour: g* 1 3 u* 20 i* Reductions done by Mireille seem to be okay, but there are some problems with bad CCDs for the u* images.

-  Laurent: rcp doesn't work on the mix2. Fred explains this is because the rcp daemon isn't installed on the mix2. Question for Laurent: do you really need rcp? scp is not good enough enough? Fred explains that rcp is non-standard and requires a lot of modification to system files (and of course presents more of a security risk).

-  Jean-Charles Cuillandre suggests that we should add an item that concerns the releases in the Terapix main page. We all agree we shoudl provide provide more information about the releases and perhaps indicate to the users any particular problem in ongoing processing. Emmanuel volunteers to look after this

Final important note
-  Headers of weight-maps need to be altered because the small 'e' (from exponent) is non-standard. Jean-Christophe will look after this.

Flash plug in Does not work well with IE. Hard to propose this tool for public outreach as it stands now. Fred will check the cost (Macromedia). If cheap, we may buy it and propose a student to work on this tool during a training period in order to develop new options/tools we need.

New hardware

-  Delphine has all the orders for the new hardware, which should have already been sent, including a replacement motherboard for mix2 to see if that resolves the problems with these machines.

Next Tera-meeting:

No tera-meeting next week because Yannick is in Gronigen.


Site Map  -   -  Contact
© Terapix 2003-2011