Failed to create S3 index, already exists #7576
Description
2025-10-17T16:13:43.026962Z WARN runtime::accelerated_table::refresh_task: Failed to load data for dataset s3tables_questions (glue): Failed to get data from the connector. External error: External error: Failed to create index in S3 Vectors. ConflictException: An index with the specified name already exists Ensure the dataset configuration is valid, and try again.Current logic is:
- Check if index exists
- If it does not exist, create one
- If we get error, return it
This can happen in parallel and there can be a race condition where two threads find that it does not exist but one creates it before the other, one is successful and one fails with this error.