Version mismatch between the pyarrow version installed and the Arrow format produced by Snowflake. Network error caused a truncated Arrow stream to be received. Memory exhaustion on the client causing the Arrow deserialization to fail mid-stream. A pyarrow bug or incompatibility with the specific Snowflake query result type. Very large result sets exceeding available memory during Arrow deserialization

Upgrade pyarrow to the version recommended for your connector: pip install 'snowflake-connector-python[pandas]'.. Fall back to non-Arrow fetching: cursor.execute(query); result = cursor.fetchall() to rule out Arrow issues.. If the issue is memory-related, fetch in batches: cursor.fetchmany(batch_size) instead of fetchall.. Reduce result set size: add WHERE/LIMIT clauses or paginate large queries.. Enable connector debug logging to see the raw Arrow stream error: logging.basicConfig(level=logging.DEBUG).

Medium severityexecutionSnowflake →

Power BI Refresh Error:
255005

What does this error mean?

The Python connector successfully received an Arrow-format result stream from Snowflake but failed to deserialize it locally. The Arrow data is corrupted, truncated, or in an incompatible format.

Common causes

1Version mismatch between the pyarrow version installed and the Arrow format produced by Snowflake
2Network error caused a truncated Arrow stream to be received
3Memory exhaustion on the client causing the Arrow deserialization to fail mid-stream
4A pyarrow bug or incompatibility with the specific Snowflake query result type
5Very large result sets exceeding available memory during Arrow deserialization

How to fix it

1Upgrade pyarrow to the version recommended for your connector: pip install 'snowflake-connector-python[pandas]'.
2Fall back to non-Arrow fetching: cursor.execute(query); result = cursor.fetchall() to rule out Arrow issues.
3If the issue is memory-related, fetch in batches: cursor.fetchmany(batch_size) instead of fetchall.
4Reduce result set size: add WHERE/LIMIT clauses or paginate large queries.
5Enable connector debug logging to see the raw Arrow stream error: logging.basicConfig(level=logging.DEBUG).

Frequently asked questions

Is fetch_pandas_all() faster than fetchall()?

Yes — Arrow-based fetching is significantly faster for large result sets. Use it for analytics workloads, but fall back to fetchall() when debugging Arrow-specific issues.

Can I disable Arrow fetching entirely?

Yes — set cursor_class=DictCursor or use the non-Arrow fetch methods. Arrow is opt-in through fetch_pandas_all() and fetch_arrow_all().

Official documentation: https://github.com/snowflakedb/snowflake-connector-python/blob/main/src/snowflake/connector/errorcode.py