Describe the bug
The function get_analysis_extent in prepifg_helper.py iterates through a (sometimes very long) list of files and opens them all with GDAL through the open method of the RasterBase class. This sometimes results in conflict with the system's limit on the number of open files and results in a runtime error.
To Reproduce
Steps to reproduce the behavior:
Go to /g/data/dg9/INSAR_ANALYSIS/SURAT_S1_Asc/S1/PYRATE/ on VDI.
Run pyrate prepifg using the config file in that folder.
Expected behavior
The get_analysis_extent functionality doesn't require all of the raster files to be open at the same time, so it should not be restricted by the system's limit on open file handles.
Notes
This can be resolved if we move file opening into the inner loop of _get_same_bounds (currently a list comprehension which is getting the geotransform of each object), and closing the dataset after we are done. I'm not sure if there are other instances in the PyRate workflow where all of the interferograms are opened at the same time, but if so these should be corrected in a similar fashion, and we should move to a more 'on-demand' file handling model to avoid this type of issue in the future.
richardt94
commented
3 years ago
Describe the bug The function
get_analysis_extentinprepifg_helper.pyiterates through a (sometimes very long) list of files and opens them all with GDAL through theopenmethod of theRasterBaseclass. This sometimes results in conflict with the system's limit on the number of open files and results in a runtime error.To Reproduce Steps to reproduce the behavior:
/g/data/dg9/INSAR_ANALYSIS/SURAT_S1_Asc/S1/PYRATE/on VDI.pyrate prepifgusing the config file in that folder.Expected behavior The
get_analysis_extentfunctionality doesn't require all of the raster files to be open at the same time, so it should not be restricted by the system's limit on open file handles.Notes This can be resolved if we move file opening into the inner loop of
_get_same_bounds(currently a list comprehension which is getting the geotransform of each object), and closing the dataset after we are done. I'm not sure if there are other instances in the PyRate workflow where all of the interferograms are opened at the same time, but if so these should be corrected in a similar fashion, and we should move to a more 'on-demand' file handling model to avoid this type of issue in the future.