download_file and flaky server (was: Trouble opening IERS table)

This causes NumPy to throw an error:

from astropy.utils import iers
dat = iers.IERS_Auto.open()

I am guessing the error is because an empty array is being generated from the IERS data and being sent to NumPy. Here is the error traceback and error message:

ERROR: ValueError: zero-size array to reduction operation minimum which has no identity [numpy.core.fromnumeric]
Error in sys.excepthook:
Traceback (most recent call last):
  File "/data/home004/warren.porter/anaconda3/envs/CrIS_geolocation_3/lib/python3.7/site-packages/astropy/logger.py", line 262, in _excepthook
    self._excepthook_orig(etype, value, traceback)
TypeError: 'NoneType' object is not callable

Original exception was:
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/data/home004/warren.porter/anaconda3/envs/CrIS_geolocation_3/lib/python3.7/site-packages/astropy/utils/iers/iers.py", line 660, in open
    cls.iers_table = cls.read(file=filename)
  File "/data/home004/warren.porter/anaconda3/envs/CrIS_geolocation_3/lib/python3.7/site-packages/astropy/utils/iers/iers.py", line 507, in read
    table = cls._combine_a_b_columns(iers_a)
  File "/data/home004/warren.porter/anaconda3/envs/CrIS_geolocation_3/lib/python3.7/site-packages/astropy/utils/iers/iers.py", line 470, in _combine_a_b_columns
    table.meta['predictive_index'] = np.min(np.flatnonzero(is_predictive))
  File "/data/home004/warren.porter/anaconda3/envs/CrIS_geolocation_3/lib/python3.7/site-packages/numpy/core/fromnumeric.py", line 2442, in amin
    initial=initial)
  File "/data/home004/warren.porter/anaconda3/envs/CrIS_geolocation_3/lib/python3.7/site-packages/numpy/core/fromnumeric.py", line 83, in _wrapreduction
    return ufunc.reduce(obj, axis, dtype, out, **passkwargs)
ValueError: zero-size array to reduction operation minimum which has no identity

Any assistance would be greatly appreciated. Thank you!

[Warren-Porter]

That sounds promising. I’m currently traveling at the moment and I won’t be able to test the update until mid-October, but I will let you folks know once I do so.

[timj]

Just a footnote here but LSST are hitting this exact problem with IERS A. Our problem is exacerbated by our CI system not having $HOME defined so every time we run multi-process pytest within the build we download the file again multiple times. Sometimes they are truncated and we were getting the exact same problem from line 470 of iers.py. @pllim thanks for pointing me at this issue. We were hoping not to have to deal with a local copy of the tables in Jenkins in the short term but it sounds like we may have to bump that up a bit.