Athena cannot query unsigned bytes (uint8) from parquet files

Unfortunately, Athena does not natively support unsigned byte (uint8) data types from Parquet files. This is a known limitation of Athena's data type handling.

The error you're encountering ("GENERIC_INTERNAL_ERROR: Value 174 exceeds MAX_BYTE") occurs because Athena is interpreting the unsigned byte values as signed bytes, which have a maximum value of 127. When it encounters values above this, like 174, it throws an error.

Changing the schema to smallint or int in Athena won't resolve the issue because the underlying data in the Parquet file is still stored as unsigned bytes. Athena is failing to properly interpret this data type.

Given that changing the data type in the Parquet files is not an option due to the large data size, you have a few potential workarounds to consider:

1. Use a custom SerDe: You could potentially create a custom SerDe (Serializer/Deserializer) that correctly interprets the unsigned byte values. However, this would require significant development effort.

2. Pre-process the data: You could set up a data pipeline that converts the unsigned byte values to a supported data type before querying with Athena. This could be done using AWS Glue or another ETL tool.

3. Use a different query engine: Consider using a different query engine that supports unsigned bytes, such as Presto or Trino, which can be run on Amazon EMR.

4. Use a different database: As you suggested, using a different database system that natively supports unsigned bytes might be the most straightforward solution if Athena's limitations are causing significant issues for your use case.

If querying this data is critical to your workflow and you need to stick with Athena, you may want to reach out to AWS support for further guidance or to request feature support for unsigned bytes in Athena.
Sources
Athena engine version 3 - Amazon Athena
Troubleshoot issues in Athena - Amazon Athena