boundless=True - hangs in docker container rio read

[tonybutzer]

Expected behavior and actual behavior.

I expected to read a band from a geotiff - window it and use boundless=True - inside a docker container - the read hangs forever. - if i remove the boundless parameter everything works as expected. This only happens in a docker container. The same python rasterio on an EC2 instance runs fine. I have built ubuntu and fedora based containers - both exhibit this behavior.

Steps to reproduce the problem.

Dockerfile

from condaforge/mambaforge

env tony 1.0
workdir /home
copy environment.yml /home
run mamba env update -n base --file environment.yml

(water) [ec2-user@ip-10-12-70-10 fail_mamba_forge]$ cat environment.yml
name: ts2023
channels:
  - conda-forge
dependencies:
  - python=3
#jupyter
#xarray-related
  - rioxarray
  - zarr
#other
  - fsspec
  - s3fs
  - folium
#aws
  - boto3
  - awscli
# gaass-py
  - pandas
  - numpy
  - rasterio
  - matplotlib
  - geopandas
  - tqdm
  - geojson
# stac VERSIONS are Important Here - bummer!
  - pystac
  - pystac-client
# CPU
  - pip
  - pip:
    - loguru
    - python-dotenv

For example: a brief script with required inputs.

(water) [ec2-user@ip-10-12-70-10 00_docker_version]$ cat test_docker_rasterio_read.py
import os
import rasterio as rio

os.environ['AWS_REQUEST_PAYER'] = 'requester'
os.environ['GDAL_DISABLE_READDIR_ON_OPEN']='EMPTY_DIR'
os.environ['GDAL_HTTP_MAX_RETRY']='10'
os.environ['GDAL_HTTP_RETRY_DELAY']='3'
os.environ['GDAL_PAM_ENABLED']='NO'


image = 's3://usgs-landsat-ard/collection02/tm/1995/CU/013/013/LT05_CU_013013_19950107_20210424_02/LT05_CU_013013_19950107_20210424_02_SR_B1.TIF'

width=256
height=256

with rio.open(image) as ds:
                    window = rio.windows.Window(0, 0, 256, 256)
                    out = ds.read(1, window=window, boundless=True, fill_value=0, masked=True).filled()

print(type(out))
print(out.shape)

Environment Information

• Output from: rio --show-versions or python -c "import rasterio; rasterio.show_versions()"

(base) rio --show-versions
rasterio info:
 rasterio: 1.3.3
     GDAL: 3.5.3
     PROJ: 9.1.0
     GEOS: 3.11.1
PROJ DATA: /opt/conda/share/proj
GDAL DATA: /opt/conda/share/gdal

System:
   python: 3.10.6 | packaged by conda-forge | (main, Aug 22 2022, 20:35:26) [GCC 10.4.0]
executable: /opt/conda/bin/python
  machine: Linux-4.14.330-250.540.amzn2.x86_64-x86_64-with-glibc2.34

Python deps:
   affine: 2.3.1
    attrs: 22.1.0
  certifi: 2023.11.17
    click: 8.1.3
    cligj: 0.7.2
   cython: None
    numpy: 1.23.4
   snuggs: 1.4.7
click-plugins: None
setuptools: 65.6.3

• rasterio version (python -c "import rasterio; print(rasterio.__version__)")
• GDAL version (python -c "import rasterio; print(rasterio.__gdal_version__)")
• Python version (python -c "import sys; print(sys.version.replace('\n', ' '))")
• Operation System Information (python -c "import platform; print(platform.platform())")

Installation Method

condaforge/mambaforge docker image - however this fails on manual conda/mamba installs as well

[sgillies]

@tonybutzer What do you see if you upgrade Rasterio and GDAL to the latest? There have been many bug fixes since the versions you report above? And if you put CPL_CURL_VERBOSE=YES in your environment?