MinIO: ObjectHashMismatchError when trying to upload object

Summary

The MinIO provider raises an ObjectHashMismatchError when trying to upload an object, for no clear reason 100% reproducible, code to repro the bug is provided

Detailed Information

libcloud: latest stable, 3.3.1 Python: 3.8.10 OS: Ubuntu 20.04

Here you can find a repo to repro the bug, using pytest (Actually there are 2 bugs to report): https://github.com/Wenzel/libcloud_bug

$ pytest -k test_demo_ObjectHashMismatchError_with_pyfakefs
clean_minio_db = <libcloud.storage.drivers.minio.MinIOStorageDriver object at 0x7f1410f3fdf0>, fs = <pyfakefs.fake_filesystem.FakeFilesystem object at 0x7f141021fa90>

    def test_demo_ObjectHashMismatchError_with_pyfakefs(clean_minio_db, fs):
        # create test file
        test_file = "/file1.txt"
        test_file_data = b"hello"
        with open(test_file, "wb") as f:
            f.write(test_file_data)
        # create container
        driver = clean_minio_db
        container = driver.create_container('test')
        # test
>       driver.upload_object(str(test_file), container, 'test_file')

/home/wenzel/Projets/libcloud_bug/tests/test_demo_bug.py:13: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
/home/wenzel/.cache/pypoetry/virtualenvs/libcloud-bug-IT-EyGKG-py3.8/lib/python3.8/site-packages/libcloud/storage/drivers/s3.py:545: in upload_object
    return self._put_object(container=container, object_name=object_name,
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

self = <libcloud.storage.drivers.minio.MinIOStorageDriver object at 0x7f1410f3fdf0>, container = <Container: name=test, provider=MinIO Storage Driver>, object_name = 'test_file', method = 'PUT'
query_args = None, extra = {}, file_path = '/file1.txt', stream = None, verify_hash = True, storage_class = None, headers = {'connection': 'close', 'content-type': 'text/plain; charset=utf-8'}

    def _put_object(self, container, object_name, method='PUT',
                    query_args=None, extra=None, file_path=None,
                    stream=None, verify_hash=True, storage_class=None,
                    headers=None):
        headers = headers or {}
        extra = extra or {}

        headers.update(self._to_storage_class_headers(storage_class))

        content_type = extra.get('content_type', None)
        meta_data = extra.get('meta_data', None)
        acl = extra.get('acl', None)

        if meta_data:
            for key, value in list(meta_data.items()):
                key = self.http_vendor_prefix + '-meta-%s' % (key)
                headers[key] = value

        if acl:
            headers[self.http_vendor_prefix + '-acl'] = acl

        request_path = self._get_object_path(container, object_name)

        if query_args:
            request_path = '?'.join((request_path, query_args))

        result_dict = self._upload_object(
            object_name=object_name, content_type=content_type,
            request_path=request_path, request_method=method,
            headers=headers, file_path=file_path, stream=stream)

        response = result_dict['response']
        bytes_transferred = result_dict['bytes_transferred']
        headers = response.headers
        response = response
        server_hash = headers.get('etag', '').replace('"', '')
        server_side_encryption = headers.get('x-amz-server-side-encryption',
                                             None)
        aws_kms_encryption = (server_side_encryption == 'aws:kms')
        hash_matches = (result_dict['data_hash'] == server_hash)

        # NOTE: If AWS KMS server side encryption is enabled, ETag won't
        # contain object MD5 digest so we skip the checksum check
        # See https://docs.aws.amazon.com/AmazonS3/latest/API
        # /RESTCommonResponseHeaders.html
        # and https://github.com/apache/libcloud/issues/1401
        # for details
        if verify_hash and not aws_kms_encryption and not hash_matches:
>           raise ObjectHashMismatchError(
                value='MD5 hash {0} checksum does not match {1}'.format(
                    server_hash, result_dict['data_hash']),
E                   libcloud.storage.types.ObjectHashMismatchError: <ObjectHashMismatchError in <libcloud.storage.drivers.minio.MinIOStorageDriver object at 0x7f1410f3fdf0>, value=MD5 hash  checksum does not match 5d41402abc4b2a76b9719d911017c592, object = test_file>

/home/wenzel/.cache/pypoetry/virtualenvs/libcloud-bug-IT-EyGKG-py3.8/lib/python3.8/site-packages/libcloud/storage/drivers/s3.py:922: ObjectHashMismatchError
------------------------------------------------------------------------------------------- Captured stdout setup -------------------------------------------------------------------------------------------
db2dfd078b96e1b0932dc354926aadb77adba2d8c19b8ccc875e7fe5163e8f46
----------------------------------------------------------------------------------------- Captured stdout teardown ------------------------------------------------------------------------------------------
libcloud_bug_objectmistmatch_miniodb
========================================================================================== short test summary info ==========================================================================================
FAILED tests/test_demo_bug.py::test_demo_ObjectHashMismatchError_with_pyfakefs - libcloud.storage.types.ObjectHashMismatchError: <ObjectHashMismatchError in <libcloud.storage.drivers.minio.MinIOStorageD..

It looks like there is a hash mismatch when upload_object is verifying the hash, but for no clear reason.

Thanks for maintaining libcloud !

apache / libcloud

MinIO: ObjectHashMismatchError when trying to upload object #1607

Summary

Detailed Information