There’s more information here: AttributeError: 'DataFrame' object has no attribute 'to_spark', but we’ve had trouble with S3 timing out on long pipelines. You might try using stage_locally. If you share the hail log file and the stack traces, we might be able to help further.