Troubleshoot ScaNN indexes in the columnar engine
Stay organized with collections
Save and categorize content based on your preferences.
This document describes commands for diagnosing and troubleshooting columnar engine index operations. Vector indexes use the Approximate Nearest Neighbor (ANN) search strategy that enhances performance and requires a balance between query latency and recall. See Choose a vector index in AlloyDB AI for more information.
Ensure index integrity
To verify that an index in the columnar engine is consistent and doesn't have any corruption or errors, especially when you experience issues related to index performance or data retrieval, use the following SQL command:
SELECTgoogle_columnar_engine_verify_index(index=>'INDEX_NAME');
Replace INDEX_NAME with the name of the index.
Enforce index creation and suppress errors
To debug or analyze the structure and data within an index, you can dump the contents of a specific index within the columnar engine using the following SQL command:
SELECTgoogle_columnar_engine_dump(index=>'INDEX_NAME');