Repository navigation
[Python] AWS Error NETWORK_CONNECTION during HeadObject operation: curlCode: 6, Couldn't resolve host name #40539
Description
Describe the bug, including details regarding any error messages, version, and platform.
Hi there!
We are using the PyArrow library to read files from an S3 bucket, and we're encountering an intermittent error:
OSError: When reading information for key '<REDACTED>' in bucket '<REDACTED>': AWS Error NETWORK_CONNECTION during HeadObject operation: curlCode: 6, Couldn't resolve host name
Please note that this error doesn't occur consistently, and the S3 bucket path is valid.
The reference code we're using is as follows:
import pyarrow as pa
import pyarrow.json as pj
uri = "s3://my-bucket/my-prefix/foo.json"
fs, path = pa.fs.FileSystem.from_uri(uri)
with fs.open_input_file(path) as f:
tbl = pj.read_json(f)
Error Details:
2024-03-12T13:06:05.616-07:00 [7]: with fs.open_input_file(path) as f:
2024-03-12T13:06:05.616-07:00 [7]: File "pyarrow/_fs.pyx", line 780, in pyarrow._fs.FileSystem.open_input_file
2024-03-12T13:06:05.616-07:00 [7]: File "pyarrow/error.pxi", line 154, in pyarrow.lib.pyarrow_internal_check_status
2024-03-12T13:06:05.616-07:00 [7]: File "pyarrow/error.pxi", line 91, in pyarrow.lib.check_status
2024-03-12T13:06:05.616-07:00 [7]:OSError: When reading information for key '<REDACTED>' in bucket '<REDACTED>': AWS Error NETWORK_CONNECTION during HeadObject operation: curlCode: 6, Couldn't resolve host name
Could you please provide any suggestions on how to handle such intermittent network connectivity errors while reading from S3?
Component(s)
Parquet, Python
Reacted by António Alegria, Jeff Picard, Hadrien Mary, Harel Oshri, Younglo Kim, elohmeier and Andrei Luca
Activity
[-]AWS Error NETWORK_CONNECTION during HeadObject operation: curlCode: 6, Couldn't resolve host name[/-][+][Python][AWS Error NETWORK_CONNECTION during HeadObject operation: curlCode: 6, Couldn't resolve host name[/+]pitrou commentedon Mar 14, 2024
Are you using AWS or do you set
endpoint_override? "Couldn't resolve host name" probably means you're having issues with DNS resolution of the S3 server(s) hostnames, which is quite unexpected with AWS...[-][Python][AWS Error NETWORK_CONNECTION during HeadObject operation: curlCode: 6, Couldn't resolve host name[/-][+][Python] AWS Error NETWORK_CONNECTION during HeadObject operation: curlCode: 6, Couldn't resolve host name[/+]anjali-chadha commentedon Mar 15, 2024
@pitrou Yes, we are using AWS, and not explicitly overriding
endpoint_overrideWe've only encountered this problem once among our numerous runs, so it's not a frequent occurrence. Currently, our approach to dealing with this issue is by increasing the default number of S3 retry attempts from 3 to a higher value.
However, we're uncertain if this is the most effective approach.
Do you have any recommendations on how we can better handle this on the client side?
pitrou commentedon Mar 16, 2024
Sorry, I don't have any recommandation. If increasing
max_attemptsworks, then it seems ok to me.antonioalegria commentedon Nov 4, 2024
I'm facing this issue when doing a scan_iceberg operation in Polars. It only happens with certain objects:
When reading information for key '' in bucket '': AWS Error NETWORK_CONNECTION during HeadObject operation: curlCode: 6, Couldn't resolve host name
Traceback (most recent call last):
File "", line 351, in scan_table
return pl.scan_iceberg(location)
^^^^^^^^^^^^^^^^^^^^^^^^^
File "/.venv/lib/python3.12/site-packages/polars/io/iceberg.py", line 143, in scan_iceberg
source = StaticTable.from_metadata(
^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/.venv/lib/python3.12/site-packages/pyiceberg/table/init.py", line 1693, in from_metadata
metadata = FromInputFile.table_metadata(file)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/.venv/lib/python3.12/site-packages/pyiceberg/serializers.py", line 113, in table_metadata
with input_file.open() as input_stream:
^^^^^^^^^^^^^^^^^
File "/.venv/lib/python3.12/site-packages/pyiceberg/io/pyarrow.py", line 270, in open
input_file = self._filesystem.open_input_file(self._path)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "pyarrow/_fs.pyx", line 789, in pyarrow._fs.FileSystem.open_input_file
File "pyarrow/error.pxi", line 155, in pyarrow.lib.pyarrow_internal_check_status
File "pyarrow/error.pxi", line 92, in pyarrow.lib.check_status
OSError: When reading information for key '' in bucket '': AWS Error NETWORK_CONNECTION during HeadObject operation: curlCode: 6, Couldn't resolve host name
hadim commentedon Feb 21, 2025
I am seeing the same error. What confuses me is that using
fsspecdirectly, everything works fine. Note that I am using a custom S3 provider (scaleway object storage) and so I set a custom endpoint URL in the AWS config file.Using:
hadim commentedon Feb 21, 2025
Current working but not ideal (verbose) workaround is:
harelo-cf commentedon Mar 4, 2025
@hadim I face the same issue and it is a new issue, I never had that before - same code that we always used now fails on table creation.
Did you manage to solve it?
@Fokko
thisisnic commentedon Oct 26, 2026
Original report answered as a client-side DNS resolution failure surfaced by the AWS SDK with
AwsStandardS3RetryStrategy(max_attempts=N)as the mitigation, and the later me-toos are different setups (Polars/pyiceberg, a custom endpoint set only in~/.aws/config) without a reprex; closing this as resolved. If anyone else encouters this, please open a new issue.