OpenSearch:Technical operations and maintenance FAQ

Inaccurate statistics for large datasets

• Each search node ranks results in two stages. First, it coarse-ranks up to one million initial results. Then, it selects the top 200 for fine-grained ranking and returns the top 5,000 results.

• For large datasets, OpenSearch randomly samples a subset of the data for sorting and returns estimated statistical values. These values are approximations, not exact figures.

Data synchronization latency

• Most data updates to a primary table synchronize within 10 seconds. Updates to a secondary table take longer, and real-time synchronization is not guaranteed.

• To update data in both a secondary table and its corresponding primary table, update the secondary table first, and then the primary table. This triggers a refresh that updates both the secondary data and the index, reducing the total synchronization time.

• Alternatively, after updating the secondary table, wait for about one second. Then, use the updated information to find and update the corresponding records in the primary table.

Handling datetime and timestamp data types

When you import data from an ApsaraDB RDS table, OpenSearch automatically converts datetime and timestamp fields to milliseconds since the epoch.

To preserve the original time format, convert these fields to the string data type in ApsaraDB RDS before importing.

Connect to OpenSearch using PrivateLink

Procedure

1. Log on to the Virtual Private Cloud (VPC) console.

   Select the target region, click the Interface Endpoint tab, and then click the Create Endpoint button.

2. Create an endpoint.

   On the Create Endpoint page, configure the VPC, security group, and zone. Keep the Endpoint Type as Interface Endpoint. For Endpoint Service, select Other Endpoint Services. In the text box that appears, enter the endpoint service name, for example, com.aliyuncs.cn-hangzhou.opensearch. Click Verify, and then click OK to create the endpoint.

3. Check the connection status.

   View the connection status on the endpoint details page:

   Initially, the State is Creating, the Connection Status is Connecting, and the Bandwidth limit is 3072 Mbps.

   After the endpoint connects, the statuses change:

   The State is Available, the Connection Status is Connected, and the Bandwidth limit is 3072 Mbps.

4. Test the connection.

   Copy the Domain Name of Endpoint Service. On an ECS instance in the target VPC, run the following command to test connectivity:

   # curl ep-foo.opensearch.cn-hangzhou.privatelink.aliyuncs.com/v3/openapi/apps/foo/search
   {
     "status":"FAIL",
     "errors":[{
         "code":4016,
         "message":"Authentication info was required for the request"
       }],
     "request_id":"d164fa75cec6cb360c13693acd806941"
   }

OpenSearch endpoint services

Region	Endpoint service name	Zone
China (Shenzhen)	com.aliyuncs.cn-shenzhen.opensearch	E and F
China (Hangzhou)	com.aliyuncs.cn-hangzhou.opensearch	I and H
China (Beijing)	com.aliyuncs.cn-beijing.opensearch	K and L
China (Shanghai)	com.aliyuncs.cn-shanghai.opensearch	E and F