IPP Progress Report for the week 2009.08.30 - 2009.09.05

(Up to IPP Progress Reports)

Eugene Magnier

Most of the IPP team were at the meeting at Harvard last week, so not a lot of progress was accomplished. We had a major incident on the MHPCC cluster, losing the RAID that hosted the home directories. The RAID was recovered eventually, thanks to Brad, Cindy, & Gavin, and in the processed we learned that the drive bays are mis-labeled. In the mean while, we took the opportunity to re-assign the home directory machine temporarily to ipp053, with the goal of moving to a pair of mirrored 1U compute nodes. We'll need to buy some raid hardware to make that a easy-to-recover setup, but it will mean a more reliable home directory in the long run. This also gives us the chance to fix or address the drive bay in ipp004 that is inactive (we've been running at raid 5 as a result for > 2 years). There is still some re-organization work needed, which was made difficult to arrange with nearly the whole team gone to Harvard, but that should be finished by early the coming week.

For my part, in addition to recovery from too much travel, I've been working on the extended source analysis. I'm now fairly happy with the radial profile extraction + petrosian analysis, and I'm incorporating it back into the psphot code base. I'm also making some tests and fixes at a low level of issues that have been reported lately, including some trouble converging on a good 2D psf model. the coming week, I will continue to focus on psphot.

Paul Price

  • Cluster woes + Checked operations pantasks scripts into repository
  • MOPS support: branches/pap_mops/ + Revised publish_file.pl to support revised ppMops, following "ICD lite" + System working again + Republished ThreePi_SouthernRegion.090818 and ThreePi_NorthernRegion.090818 to MOPS + Merged into trunk
  • Destreaking: fixed doCopyExtraExtensions in streaksremove --- wasn't actually doing the write, wasn't setting compression appropriately
  • Detection efficiency: branches/pap/ + Got first cut of code done + Discovered psphotSignificanceImage doesn't seem to be using correct smoothing size: decreased sensitivity, but not too bad? + Finally worked out how to calculate rough limiting magnitude + Working now, next concern is formatting for output
    @FAKE.EFF F32 0.884 0.838 0.34 0.046 0.026   # Efficiency fractions
    @FAKE.MAG F32 -1 -0.5 0 0.5 1   # Efficiency magnitudes
    FAKE.REF         F32       -7.570797        # Efficiency reference magnitude
    

Bill Sweeney

During the week ending August 28 bill worked on improving the performance of the postage stamp request processing. This led to a factor of 10 improvement for typical MOPS request.

He spent the week ending Septmber 4 at the PS1SC conference in Cambridge.

In the upcoming week he will be 'releasing' the IPP postage stamp server to the consortium.

Heather Flewelling

3 weeks ago

  • culled a few million raw science files. I hope we never have to do

that again. Learned a few more things about nebulous in the process.

2 weeks ago

  • contacted Natalia. Gave her instructions on compiling ipp and some

links to read on dvo for when she gets an account on the manoa cluster.

  • worked on dvo stats for smf/cmf. Started writing macros to make dvo

more idl like, but also useful for the dvo stats I'm working on. Should be useful for other ipp-users.

  • investigated other raw images files with >2 copies
  • investigated raw image files with <2 copies that are not replicated

(caused by ipp008 being pulled out of nebulous and the raw image files not having the proper xattr bit set)

  • flew to boston for conference

last week

  • Attended panstarrs conference in boston
  • ate lobster for the first time
  • met with harvard
  • installed ipp on my mac laptop
    • forced cfitsio, libtool, and libjpeg to install, coped glibtool to

psLib and psModules directory (renamed to libtool, thanks to a helpful hint from our consortium members from NCU), and hacked a few of the header files to make it work. Probably forgetting a few other things. Not pretty, but it works.

  • spent thursday and friday in michigan (go blue!!)

Bill Giebink

  • Continued working on psLib tap tests throughout week
  • Restart and maintenance of DIMM on Tuesday
  • Observed with Natalia Wednesday night

Chris Waters

  • Changes to ppImage to allow optional masking of unfit burntool trails as well as pixels repaired by burntool.
  • Harvard meeting.
  • Began investigation of spurious transient sources noted by Armin Rest to determine if they are really caused by burntool, and whether ppImage masking could fix this.