EOF error when writing to S3

First, thanks for this awesome library! I was testing out some of the features yesterday and hit this error. The workflow was:

• Read a geospatial dataset off S3 using geopandas
• Create a geodataframe (a special case of a pandas dataframe)
• Write to S3 using aws-data-wrangler like so:

session = awswrangler.Session()
session.pandas.to_parquet(
    dataframe=final_gdf.astype({'geometry': str}),
    database='foo',
    table='bar',
    path="s3://bucket/path/to/file/",
    partition_cols=['partition_col'],
    compression='gzip',
    mode='overwrite'
)

It looks like it writes many files but eventually I hit an EOFError error:

---------------------------------------------------------------------------
EOFError                                  Traceback (most recent call last)
<ipython-input-40-0f6d02d0cf6c> in <module>
      5     partition_cols=['STATEFP', 'COUNTYFP', 'TRACTCE'],
      6     compression='gzip',
----> 7     mode='overwrite'
      8 )

/usr/local/lib/python3.7/site-packages/awswrangler/pandas.py in to_parquet(self, dataframe, path, database, table, partition_cols, preserve_index, mode, compression, procs_cpu_bound, procs_io_bound, cast_columns, inplace)
    555                           procs_io_bound=procs_io_bound,
    556                           cast_columns=cast_columns,
--> 557                           inplace=inplace)
    558 
    559     def to_s3(self,

/usr/local/lib/python3.7/site-packages/awswrangler/pandas.py in to_s3(self, dataframe, path, file_format, database, table, partition_cols, preserve_index, mode, compression, procs_cpu_bound, procs_io_bound, cast_columns, extra_args, inplace)
    632                                         procs_io_bound=procs_io_bound,
    633                                         cast_columns=cast_columns,
--> 634                                         extra_args=extra_args)
    635         if database:
    636             self._session.glue.metadata_to_glue(dataframe=dataframe,

/usr/local/lib/python3.7/site-packages/awswrangler/pandas.py in data_to_s3(self, dataframe, path, file_format, partition_cols, preserve_index, mode, compression, procs_cpu_bound, procs_io_bound, cast_columns, extra_args)
    685                 receive_pipes.append(receive_pipe)
    686             for i in range(len(procs)):
--> 687                 objects_paths += receive_pipes[i].recv()
    688                 procs[i].join()
    689                 receive_pipes[i].close()

/usr/local/lib/python3.7/multiprocessing/connection.py in recv(self)
    248         self._check_closed()
    249         self._check_readable()
--> 250         buf = self._recv_bytes()
    251         return _ForkingPickler.loads(buf.getbuffer())
    252 

/usr/local/lib/python3.7/multiprocessing/connection.py in _recv_bytes(self, maxsize)
    405 
    406     def _recv_bytes(self, maxsize=None):
--> 407         buf = self._recv(4)
    408         size, = struct.unpack("!i", buf.getvalue())
    409         if maxsize is not None and size > maxsize:

/usr/local/lib/python3.7/multiprocessing/connection.py in _recv(self, size, read)
    381             if n == 0:
    382                 if remaining == size:
--> 383                     raise EOFError
    384                 else:
    385                     raise OSError("got end of file during message")

EOFError: 

Looks like the connection is getting reset somehow? I’ll note that I’m testing this in a jupyter notebook running in a docker container.

Current configuration:

• MacOS with a Docker container running Debian (stretch)
• Python 3.7
• awswrangler 0.0.12

Any ideas what’s going on here?

[igorborgest]

Hi @koshy1123, thanks for contributing here!

Peculiar case, I’ve never faced that before, I will need more time to understand/troubleshoot that.

But in the meantime could you test the same without the parallelism (procs_cpu_bound=1)?

session = awswrangler.Session()
session.pandas.to_parquet(
    dataframe=final_gdf.astype({'geometry': str}),
    database='foo',
    table='bar',
    path="s3://bucket/path/to/file/",
    partition_cols=['partition_col'],
    compression='gzip',
    mode='overwrite',
    procs_cpu_bound=1
)

[koshy1123]

I got parallelization working after casting the column to be a str and forcing the dataframe to be an instance of a pandas dataframe, a la:

df = df.astype('geometry': str)
df = pandas.DataFrame(df)

Edit: This works some of the time 😕