Encountering the following error when creating a cutout (ERA5): <div class="snippe

PermissionError during temp file cleanup after cutout download (Windows) about atlite HOT 5 CLOSED

euronion commented on May 23, 2024

PermissionError during temp file cleanup after cutout download (Windows)

from atlite.

Comments (5)

coroa commented on May 23, 2024

A direct consequence for making the cutout preparation work for sarah was to let go of the cleanup tmp files automatically once the file handle is released (the weakref.finalizer(noisy_unlink, ...) bit), since dask is often configured to use multiple processes and then whether a file handle is released in one process is not a good indicator for whether we still need the tmp file. Thus, I was forced to switch to a create all files in a temporary directory and clean up the directory when the cutout is ready scheme.

What probably happens for you now, is that some exception happens during the preparation of the cutout. During the error handling I try to clean up the temporary files, but the file handles have not been released and windows prohibits the deletion of the files. The easiest way to debug this problem now, is to supply a tmpdir="<some_empty_dir_that_you_created>" argument to cutout.prepare, so that keep_tmpdir is set and the OSError does not overshadow the real exception underneath and we can find out who holds onto its filehandles in this situation.

Ideally all filehandles should be released before exiting a scope with error exceptions (ie using with or a finally clause)), as another way to mitigate it would probably be good to wrap the rmtree in an try-except clause and turn the OSError into a log message so it does not hide the true exception.

from atlite.

euronion commented on May 23, 2024

By specifying a tmpdir the operation runs without error:

$ cutout.prepare(tmpdir="./localtmp/")
INFO:atlite.cutout:Cutout uk-2011-01-2 not found in directory ./, building new one
INFO:cdsapi:Sending request to https://cds.climate.copernicus.eu/api/v2/resources/reanalysis-era5-single-levels
INFO:cdsapi:Sending request to https://cds.climate.copernicus.eu/api/v2/resources/reanalysis-era5-single-levels
INFO:cdsapi:Sending request to https://cds.climate.copernicus.eu/api/v2/resources/reanalysis-era5-single-levels
INFO:cdsapi:Sending request to https://cds.climate.copernicus.eu/api/v2/resources/reanalysis-era5-single-levels
INFO:cdsapi:Sending request to https://cds.climate.copernicus.eu/api/v2/resources/reanalysis-era5-single-levels
INFO:cdsapi:Request is completed
INFO:atlite.datasets.common:Downloading request for 1 variables to C:\Users\J. Hampp\Documents\GitHub\atlite\examples\localtmp\tmpybiq8b4h.nc
INFO:cdsapi:Downloading http://136.156.132.198/cache-compute-0003/cache/data5/adaptor.mars.internal-1568637651.461136-31786-5-0d10c442-82a6-4f46-8be4-a73443c1b3ec.nc to C:\Users\J. Hampp\Documents\GitHub\atlite\examples\localtmp\tmpybiq8b4h.nc (3.9M)
INFO:cdsapi:Request is completed
INFO:cdsapi:Request is completed
INFO:cdsapi:Request is completed
INFO:cdsapi:Request is completed
INFO:atlite.datasets.common:Downloading request for 3 variables to C:\Users\J. Hampp\Documents\GitHub\atlite\examples\localtmp\tmplpevg6_5.nc
INFO:atlite.datasets.common:Downloading request for 9 variables to C:\Users\J. Hampp\Documents\GitHub\atlite\examples\localtmp\tmpvoopfrne.nc
INFO:cdsapi:Downloading http://136.156.133.39/cache-compute-0012/cache/data3/adaptor.mars.internal-1568637651.4501736-25378-15-3562dc67-d581-4f22-a699-4d06068c8af5.nc to C:\Users\J. Hampp\Documents\GitHub\atlite\examples\localtmp\tmplpevg6_5.nc (11.6M)
INFO:cdsapi:Downloading http://136.156.132.210/cache-compute-0005/cache/data7/adaptor.mars.internal-1568702524.1244705-7942-5-a3a39405-ad59-4a37-aba5-4ea662c1747b.nc to C:\Users\J. Hampp\Documents\GitHub\atlite\examples\localtmp\tmpvoopfrne.nc (6.8K)
INFO:atlite.datasets.common:Downloading request for 2 variables to C:\Users\J. Hampp\Documents\GitHub\atlite\examples\localtmp\tmpcutwpaeh.nc
INFO:atlite.datasets.common:Downloading request for 4 variables to C:\Users\J. Hampp\Documents\GitHub\atlite\examples\localtmp\tmpaao4up9g.nc
INFO:cdsapi:Downloading http://136.156.132.198/cache-compute-0003/cache/data2/adaptor.mars.internal-1568637672.4669356-30780-7-f10d482b-f931-4644-9c61-209b361f0a76.nc to C:\Users\J. Hampp\Documents\GitHub\atlite\examples\localtmp\tmpcutwpaeh.nc (7.7M)
INFO:cdsapi:Downloading http://136.156.133.37/cache-compute-0011/cache/data8/adaptor.mars.internal-1568637651.4599202-22719-5-02ff42ba-49a2-420a-a401-8d45c3a3c8c4.nc to C:\Users\J. Hampp\Documents\GitHub\atlite\examples\localtmp\tmpaao4up9g.nc (15.5M)
INFO:cdsapi:Download rate 88.7K/s
INFO:cdsapi:Download rate 2.8M/s
INFO:cdsapi:Download rate 4.8M/s
INFO:cdsapi:Download rate 4.2M/s
INFO:cdsapi:Download rate 3.9M/s

[                                        ] | 0% Completed |  0.0s

C:\anaconda\envs\atlite\lib\site-packages\dask\core.py:119: RuntimeWarning: divide by zero encountered in true_divide
  return func(*args2)
C:\anaconda\envs\atlite\lib\site-packages\dask\core.py:119: RuntimeWarning: invalid value encountered in true_divide
  return func(*args2)

[########################################] | 100% Completed |  0.4s

The cutout is properly prepared.
It seems to me some file handles are not released properly before the finally clause.
For now this is an acceptable workaround, but maybe we can pinpoint and resolve the cause.

from atlite.

coroa commented on May 23, 2024

Should be fixed in v0.2. Re-open if the problem persists.

from atlite.

euronion commented on May 23, 2024

I can confirm that this works for now in my limited test case.

from atlite.

euronion commented on May 23, 2024

Cheering a bit to early.

This introduced a minor bug:
If the cutout is already fully created, then

atlite/atlite/data.py

Lines 199 to 202 in 09714d5

 if not missing_features and not overwrite: 

 logger.info(f"All available features {cutout.available_features} have already been prepared, so nothing to do." 

 f" Use `overwrite=True` to re-create {cutout.name}.nc and {cutout.name}.sindex.pickle.") 

 return

returns, but first the finally of the block

atlite/atlite/data.py

Lines 227 to 236 in 09714d5

 finally: 

 # ds is the last reference to the temporary files: 

 # - we remove it from this scope, and 

 # - fire up the garbage collector, 

 # => xarray's file manager closes them and we can remove tmpdir 

 del ds 

 gc.collect() 

 if not keep_tmpdir: 

 rmtree(tmpdir)

get's executed.
Problem:
In this case ds is never defined, raising

    230         # - fire up the garbage collector,
    231         # => xarray's file manager closes them and we can remove tmpdir
--> 232         del ds
    233         gc.collect()
    234 

UnboundLocalError: local variable 'ds' referenced before assignment

Maybe wrap the del ds inside its own

try:
    del ds
except NameError:
    # that's ok.

from atlite.

PermissionError during temp file cleanup after cutout download (Windows) about atlite HOT 5 CLOSED

Comments (5)

Related Issues (20)

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent

	if not missing_features and not overwrite:
	logger.info(f"All available features {cutout.available_features} have already been prepared, so nothing to do."
	f" Use `overwrite=True` to re-create {cutout.name}.nc and {cutout.name}.sindex.pickle.")
	return

	finally:
	# ds is the last reference to the temporary files:
	# - we remove it from this scope, and
	# - fire up the garbage collector,
	# => xarray's file manager closes them and we can remove tmpdir
	del ds
	gc.collect()

	if not keep_tmpdir:
	rmtree(tmpdir)