fix: add retry in read_parquet_metadata_distributed (#2196)

* fix failing tests and typing

* minor - adjust lf benchmark time

Author: jaidisido

awswrangler/distributed/ray/s3/_read_parquet.py

@@ -3,12 +3,14 @@
 import pyarrow as pa
 from pyarrow.fs import _resolve_filesystem_and_path
 
+from awswrangler import _utils
 from awswrangler.s3._read_parquet import _pyarrow_parquet_file_wrapper
 
 if TYPE_CHECKING:
     from mypy_boto3_s3 import S3Client
 
 
+@_utils.retry(ex=OSError)
 def _read_parquet_metadata_file_distributed(
     s3_client: Optional["S3Client"],
     path: str,

tests/load/test_databases.py

@@ -134,7 +134,7 @@ def test_athena_unload(benchmark_time: int, path: str, glue_table: str, glue_dat
     assert df.shape == df_out.shape
 
 
-@pytest.mark.parametrize("benchmark_time", [60])
+@pytest.mark.parametrize("benchmark_time", [80])
 def test_lakeformation_read(benchmark_time: int, path: str, glue_table: str, glue_database: str, request) -> None:
     df = wr.s3.read_parquet(path="s3://amazon-reviews-pds/parquet/product_category=Home/", dataset=True)