boto3 problems after updating to 2.1.x

It seems like I’m unable to get any of my cassettes to playback correctly after upgrading vcrpy from 2.0.1 to 2.1.0. In particular, this is the error I’m running into:

request = <SubRequest '_vcr_marker' for <Function test_datadog_failure_callback>>

    @pytest.fixture(autouse=True)
    def _vcr_marker(request):
        marker = request.node.get_closest_marker('vcr')
        if marker:
>           request.getfixturevalue('vcr_cassette')

../../../Library/Caches/pypoetry/virtualenvs/lattice-py3.7/lib/python3.7/site-packages/pytest_vcr.py:46:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
../../../Library/Caches/pypoetry/virtualenvs/lattice-py3.7/lib/python3.7/site-packages/pytest_vcr.py:87: in vcr_cassette
    with vcr.use_cassette(vcr_cassette_name, **kwargs) as cassette:
../../../Library/Caches/pypoetry/virtualenvs/lattice-py3.7/lib/python3.7/site-packages/vcr/cassette.py:96: in __enter__
    return next(self.__finish)
../../../Library/Caches/pypoetry/virtualenvs/lattice-py3.7/lib/python3.7/site-packages/vcr/cassette.py:64: in _patch_generator
    for patcher in CassettePatcherBuilder(cassette).build():
../../../Library/Caches/pypoetry/virtualenvs/lattice-py3.7/lib/python3.7/site-packages/vcr/patch.py:123: in _build_patchers_from_mock_triples
    for args in mock_triples:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

self = <vcr.patch.CassettePatcherBuilder object at 0x10780b5c0>

    @_build_patchers_from_mock_triples_decorator
    def _boto3(self):

        try:
            # botocore using awsrequest
            import botocore.awsrequest as cpool
        except ImportError:  # pragma: no cover
            try:
                # botocore using vendored requests
                import botocore.vendored.requests.packages.urllib3.connectionpool as cpool
            except ImportError:  # pragma: no cover
                pass
            else:
                from .stubs import boto3_stubs
                yield self._urllib3_patchers(cpool, boto3_stubs)
        else:
            from .stubs import boto3_stubs
            log.debug("Patching boto3 cpool with %s", cpool)
>           yield cpool.AWSHTTPConnectionPool, 'ConnectionCls', boto3_stubs.VCRRequestsHTTPConnection
E           AttributeError: module 'botocore.awsrequest' has no attribute 'AWSHTTPConnectionPool'

../../../Library/Caches/pypoetry/virtualenvs/lattice-py3.7/lib/python3.7/site-packages/vcr/patch.py:210: AttributeError

[fgblomqvist]

Sure, but what I’m arguing is that this is not vcrpy’s fault. You should open an issue with Airflow and tell them to update their boto3 (1.7.x is quite old), and until they do, you’re just gonna have to run an older version of vcrpy. If all Python libraries had to support really old versions of dependencies just because some libraries refuse to update, the ecosystem would never get anywhere.

[maxcountryman]

Alright. At the very least the >= specifier should probably be used since VCR breaks with too low of a version of boto3. But yeah thanks!

Our lock is requiring this: boto3 = ">=1.7.0,<1.8.0" which is believe is tied directly to Airflow.

We’re currently on Airflow ^1.10.