1Z0-184-25 Question 28
Select 2You have an Oracle Database 23c environment in Oracle Cloud Infrastructure that stores product embeddings in a VECTOR column named PRODUCT_EMBEDDING. After your data science team runs a bulk update on thousands of records to refine these embeddings, you notice that vector-based similarity queries are not returning the latest results. You want to ensure minimal downtime and that updated embeddings are reflected as soon as possible in queries. Which two approaches can help you meet these requirements?
- A
Run an online rebuild on the existing vector index to refresh it with the latest updates.
- B
Leverage the automatic synchronization job if the vector index was created with auto-sync parameters and schedule it more frequently as needed.
- C
Drop and recreate the entire table to guarantee that the next set of queries reflects the updated embeddings.
- D
Manually convert the vector data to a numeric array before running the UPDATE statements, then convert it back to VECTOR after updates.
Show answer and explanation
Correct answers: A, B
Explanation
In Oracle Database 23c, vector data can be maintained via standard DML statements. However, because vector indexing is often implemented as a domain index, additional steps may be required to refresh or rebuild that index after large-scale updates. You can either rely on an automatic synchronization job�if configured accordingly�or perform an online index rebuild to make the index aware of the new embeddings. For more information, refer to Oracle� documentation on vector indexing best practices and domain index maintenance for Oracle Database 23c.
- A. Correct.
Option 1 is correct. Performing an online rebuild of the vector index (for example, ALTER INDEX idx_product_embedding REBUILD ONLINE) ensures the updated embeddings are immediately recognized by similarity queries while minimizing downtime. This approach is a recommended practice when the default synchronization may not be sufficient or when extensive data changes occur.
- B. Correct.
Option 2 is correct. If your vector index was created with parameters that enable automatic synchronization (similar to certain domain indexes), Oracle Database can update the index in the background. Adjusting the frequency of this job ensures the new embeddings are recognized sooner. This can be particularly helpful for regular bulk updates where you cannot rebuild the index each time.
- C. Incorrect.
Option 3 is incorrect. Dropping and recreating the entire table is overly disruptive, impacting both downtime and data availability. Such an approach is rarely necessary just to refresh an index and goes against standard best practices for production environments.
- D. Incorrect.
Option 4 is incorrect. Converting VECTOR data to a numeric array before an UPDATE is unnecessary and does not address the issue of refreshing the vector index. This step would complicate your workflow without guaranteeing that new vector queries reflect updated embeddings.