Monday Morning Meeting 1 August 2016

  • DIAL-IN NUMBERS & PASSCODES:
  • IP: 192.33.117.12##8110
  • Phone: (434) 817-6524

Minutes in blue.

Attendance

  • Socorro: Jeff, Martin, George, Lindsay, Susan?, Juergen, Tak, Kumar, Urvashi
  • CV: Andy, Darrell, Anand, Dave
  • Garching: Sandra, Justo, Julian
  • SCO:

News / Meetings / Visitors

  • Anthony Moraghan and Kuo-Song Wang in Socorro early this week, visiting from ACDC. Will be talking with Jeff and Susan about CARTA.
  • Last week Jeff was involved in ALMA ICT meeting. ALMA is moving toward using git.
  • Pipeline Review last week. Review of pipeline and pipeline operations. The review committee was the ALMA Management Team. Software came out looking pretty good. Operations seems to need some improvement.
  • Adding content creator acct to plone.
  • Justo has accepted lead developer position for telcal; will transition to that position at end of year.

Build, Release

  • started unification of build & test storage
  • Today is feature freeze.
    • Aug 15 is code freeze
    • Mid-September is release

Verification Testing

  • Due to Ngc2403 test failing, no stable yet - Justo believes this is fixed as of 7:30 AM ET (r37825)
  • Waiting for a stable to confirm Tsdbaseline is working, to add it back to the critical tests
  • alma-m100-hpc regression added (this uses tclean) to the test framework
    • Running on RH7 (VM) and on the cluster
  • Worked on POP goals

Validation Testing

Architecture

HPC

  • More parallel testing of the ALMA pipeline. I sent a report which shows that there are failures in hif_makeimages, therefore the run seems successful but it is not. I'm now investigating the created images, but I believe they are empty. Contrary to what Jeff said that James had similar problems but switching to a different OS fixed it, I don't think the problem is related to OS. I run my tests in CV's lustre. More results tomorrow.
  • Re-run the EVLA pipeline to verify Julian's fix in quicksort. The pipeline runs to completion in parallel.
  • Discussed how to use the new mpi4casa feature to re-set memory limits settings at run time from python in the pipeline context

Development

Pipeline

AOB

  • Erik Rosolowsky needs a content creators account on plone for writing CARTA documentation. Anand will follow up with the helpdesk.

Developer Reports

Monday Meeting
  • Sanjay Bhatnagar
  • Sandra Castro
    • CAS-8842: "frequency selection ignored in flagdata": Verified that this failure is not in flagdata as it also happens with the ms tool alone. Sanjay looked further and saw that this MS had CHAN_WIDTH incorrect values, which confused MS selection.
    • We have a student (Recep Balbay) working with us for the past month and half. He has developed a prototype of a tool that re-indexes the sub-tables and sanitizes all meta-data as well as it removes unused rows after an MS is split using an intent selection (this is reported in CAS-7829) and in other places. The student has done the initial testing, but we need other MSs that show this problem. Have posted in the ticket asking Todd and Crystal to provide us with datasets.
  • Lindsey Davis
  • Pam Ford
    • Further work on plotcal->plotms poln selection and flagging: use of George's new VisCalGlobals::calParSlice function for supported cal tables.
    • Tested CAS-8018 (avg channel number) in plotms, reported bug to Justo; related ticket CAS-8853 is same bug I think.
  • Kumar Golap
  • Justo Gonz├ílez Villalba
    • Finish cross-validation of UVContsubTVI vs uvcontsub3 including high fit orders, non-trivial channel patterns and line-free selection, identified all differences which are related to different handling of channelize flags and weights. Next step is integration in mstransform.
  • Jim Jacobs
  • Jeff Kern
    • ICT Admin Meeting
    • Pipeline Review
    • Talk to the NRAO SysAdmin Meeting
  • Susan Loveland
    • Vacation
  • David Mehringer
    • (CAS-8855) implement multithreading in ClassicalStatistics::_doMaxMin(): WIP. Working with Julian on optimization.
    • (CAS-3660) importuvfits can't handle gaps in Antenna tables: Discussed difficulties in testing with Jen, because not straightforward how to create another data set with this issue, so given it is an exceptional corner case that has been tested on the data set given, suggested it be resolved.
    • (CAS-8848) implement multithreading in ClassicalStatistics::_createDataArray(): RTT.
    • (CAS-8839) request to remove channel averaging from exportuvfits: IR. Discussed with Juergen who said he would make request to stakeholders.
    • (CAS-8837) implement multithreading in ClassicalStatistics::_createDataArrays(). RTT.
  • George Moellenbrock
  • Dirk Petry
    • worked on and readied for test:
      • (CAS-4885) concat+split creates incorrect FEED subtable entries
      • (CAS-7636) listobs failure on VLA ephemeris table
      • (CAS-7545) streamline importfits "defaultaxes" argument syntax
      • (CAS-8799) concat breaks feed table?
      • (CAS-8801) concat doesn't combine SOURCE_ID correctly
    • tidied up scheduling of tickets assigned to me
      • (CAS-6803) Upgrade importfitsidi to full VLBI support
        • on hold but already partially covered by work by Mark Kettenis
      • (CAS-8323) Create script only dependent on CASA/CASACore/Python to update measures tables
        • in contact with people from ASTRON to work this out; some progress
  • Martin Pokorny
    • Continuing CASA I/O testing; tracking down causes of some read performance characteristics
    • [VLA: CBE work (I/O) for VLASS support]
    • [Vacation]
  • Urvashi Rao
    • Made a local casa build for a few local users to try out widebandpbcor with the new EVLA models.
    • Fixed a bug with .pb creation in parallel continuum.
    • User support + non CASA work (summer student)
  • Darrell Schiebel
    • worked on generation of CASA dependency graph (for Structure101/Understand C++)
    • met with Anthony from AC/DC
  • Ville Suoranta
  • Julian Taylor
    • changed casacore quicksort to introsort (gh-469), fixes gencal crash (CAS-8785)
  • Takahiro Tsutsumi
    • Added auto-adjustment of the input threshold for automask (does less aggressive masking for the first iteration and also when the rms change is less than 10%
    • Worked on to find a work around for an extra automasking in interactive mode - the current interactive GUI reads from mask image on disk so automask need to run before that. But the decision whether it needs another minor cycles happens later.
    • Pending: CAS-7863 (new asteroid models). ready to check in but waiting for Bryan's response
Friday NAOJ Meeting
  • Kanako Sugimoto
  • Wataru Kawasaki
  • Masaya Kuniyoshi
  • Takeshi Nakazato
* Renaud Miel * Renaud Miel
Topic revision: r19 - 2016-08-01, JaredCrossley
This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding NRAO Public Wiki? Send feedback