Uploaded image for project: 'Data Management'
  1. Data Management
  2. DM-4706

Rerun and create a repository for CFHT astrometry test.

    XMLWordPrintable

Details

    • Story
    • Status: Done
    • Resolution: Done
    • None
    • obs_cfht, Test Data
    • None

    Description

      Understand, re-rerun, and recreate clean version of boutigny 's CFHT astrometry test for the astrometry RMS for two sample CFHT observations. This test is on the NCSA machines in
      /lsst8/boutigny/valid_cfht

      Create a repository for this test with an eye toward it becoming integrated in a validation suite for the stack.

      Attachments

        Issue Links

          Activity

            Yes, this is leading toward a key piece of the long term validation infrastructure. This Story is part of the Epic DM-3864: "Integration Dataset for metrics and regression tests - Part I":

            This present DM-4706 was covering specifically adding boutigny 's astrometric performance check.

            You're very right to call out the connection of this epic and `ci_hsc`. I've just added DM-4817 as a Story under Epic DM-3864 to call this out explicitly. My intention is to develop a few more tests here and then look through the `ci_hsc` in the next month.

            I'm honestly not familiar with the state of how much of `ci_hsc` is runnable in the current DM stack (e.g., building `obs_subaru` with `lsstsw` `a4d9de0` and `lsst_build` `015c01a` current fails on the `psycopg2` dependency).

            The idea of "validation" vs. "testing" is 2-fold. One: it's a test of the integration rather than of specific modules. But two, and the distinction from continuous integration, is that the validation effort will be runnable at a variety of scales:
            1. Quick 1-minute turn around for continuous integration.
            2. Perhaps nightly 30-minute scale processing to alert if there's been any more subtle regression in performance.
            3. Weekly hours-long processing scale to calculate more detailed performance metrics that are part of the goals of the DM stack.

            The `bin/check_astrometry.py` in `lsst_dm_stack_demo` is exactly the same code that's being further developed in `validate_drp`. The duplication here is exactly intentional. `lsst_dm_stack_demo` has been a very loosely "validation" that the stack kind of actually works after an install. There's no strong current intent to develop `lsst_dm_stack_demo` that much further as part of Epic DM-3864. We can decide later whether `lsst_dm_stack_demo` should explicitly import from `validate_drp`, but my preference would be to actually not create any additional dependencies for `lsst_dm_stack_demo` until we have a clearer idea of the purpose of the `lsst_dm_stack_demo`.

            wmwood-vasey Michael Wood-Vasey added a comment - Yes, this is leading toward a key piece of the long term validation infrastructure. This Story is part of the Epic DM-3864 : "Integration Dataset for metrics and regression tests - Part I": This present DM-4706 was covering specifically adding boutigny 's astrometric performance check. You're very right to call out the connection of this epic and `ci_hsc`. I've just added DM-4817 as a Story under Epic DM-3864 to call this out explicitly. My intention is to develop a few more tests here and then look through the `ci_hsc` in the next month. I'm honestly not familiar with the state of how much of `ci_hsc` is runnable in the current DM stack (e.g., building `obs_subaru` with `lsstsw` `a4d9de0` and `lsst_build` `015c01a` current fails on the `psycopg2` dependency). The idea of "validation" vs. "testing" is 2-fold. One: it's a test of the integration rather than of specific modules. But two, and the distinction from continuous integration, is that the validation effort will be runnable at a variety of scales: 1. Quick 1-minute turn around for continuous integration. 2. Perhaps nightly 30-minute scale processing to alert if there's been any more subtle regression in performance. 3. Weekly hours-long processing scale to calculate more detailed performance metrics that are part of the goals of the DM stack. The `bin/check_astrometry.py` in `lsst_dm_stack_demo` is exactly the same code that's being further developed in `validate_drp`. The duplication here is exactly intentional. `lsst_dm_stack_demo` has been a very loosely "validation" that the stack kind of actually works after an install. There's no strong current intent to develop `lsst_dm_stack_demo` that much further as part of Epic DM-3864 . We can decide later whether `lsst_dm_stack_demo` should explicitly import from `validate_drp`, but my preference would be to actually not create any additional dependencies for `lsst_dm_stack_demo` until we have a clearer idea of the purpose of the `lsst_dm_stack_demo`.

            Thanks for that; sounds useful & reasonable.

            To clarify the status of obs_subaru: we're aware that it doesn't build with lsstsw at present. The fix is all ready to go (DM-3518), but unfortunately we are blocked from merging by DM-4733. As soon as that's resolved we should be up and running and will press ahead with getting ci_hsc integrated (that's DM-4307). This is essential for the HSC port, so it's top priority from our point of view.

            From a quick look at the ci_hsc and validate_drp packages (I claim expertise on neither), it seems like ci_hsc has a more thought-through infrastructure for adding and running tests (based on SCons, which seems like it would be just what you want for this kind of thing), but the individual tests are pretty simplistic compared to the astrometry check in validate_drp. It sounds like it would be a great idea to investigate whether they can adopt a common approach before they diverge further. In other words, on DM-4817!

            swinbank John Swinbank added a comment - Thanks for that; sounds useful & reasonable. To clarify the status of obs_subaru : we're aware that it doesn't build with lsstsw at present. The fix is all ready to go ( DM-3518 ), but unfortunately we are blocked from merging by DM-4733 . As soon as that's resolved we should be up and running and will press ahead with getting ci_hsc integrated (that's DM-4307 ). This is essential for the HSC port, so it's top priority from our point of view. From a quick look at the ci_hsc and validate_drp packages (I claim expertise on neither), it seems like ci_hsc has a more thought-through infrastructure for adding and running tests (based on SCons, which seems like it would be just what you want for this kind of thing), but the individual tests are pretty simplistic compared to the astrometry check in validate_drp . It sounds like it would be a great idea to investigate whether they can adopt a common approach before they diverge further. In other words, on DM-4817 !

            nidever Could you kindly take a moment to review this issue?

            wmwood-vasey Michael Wood-Vasey added a comment - nidever Could you kindly take a moment to review this issue?

            nidever Could you kindly take a look at this ticket?

            If you are unable to review it, please let me know and I will ask someone else.

            wmwood-vasey Michael Wood-Vasey added a comment - nidever Could you kindly take a look at this ticket? If you are unable to review it, please let me know and I will ask someone else.

            This issue ended up being completely covered by DM-4707, DM-4708, DM-4709, DM-4814, DM-4827. I am closing this issue as Done.

            By the time the prototype was reviewed (DM-4709), and the generalizations to SDSS (DM-4708), and DECam (DM-4707, DM-4814), and then further development to standardize structure of the code (DM-4827), this present issue, DM-4706, was covered and reviewed.

            wmwood-vasey Michael Wood-Vasey added a comment - This issue ended up being completely covered by DM-4707 , DM-4708 , DM-4709 , DM-4814 , DM-4827 . I am closing this issue as Done. By the time the prototype was reviewed ( DM-4709 ), and the generalizations to SDSS ( DM-4708 ), and DECam ( DM-4707 , DM-4814 ), and then further development to standardize structure of the code ( DM-4827 ), this present issue, DM-4706 , was covered and reviewed.

            People

              wmwood-vasey Michael Wood-Vasey
              wmwood-vasey Michael Wood-Vasey
              David Nidever [X] (Inactive), Frossie Economou, John Swinbank, Michael Wood-Vasey
              Votes:
              0 Vote for this issue
              Watchers:
              4 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved:

                Jenkins

                  No builds found.