OpenSearch creates an image for uploaded data and data synchronized from data sources. Reindexing is required whenever the application schema changes or full data is uploaded. Two scenarios trigger reindexing:
Scheduled reindexing -- Full data from a MaxCompute data source is synchronized to the application once per day.
Offline changes -- Schema modifications, index field changes, or attribute field changes trigger reindexing.
This topic describes how to configure scheduled reindexing for an OpenSearch High-performance Search Edition application.
How it works
Reindexing stages
Reindexing follows three stages:
Import data -- Synchronize full data from the data source.
Process data -- OpenSearch processes the imported data.
Build indexes -- Deliver the processed data to the build service for index building.
Each reindexing cycle takes a minimum of 30 minutes, even for applications with a small number of documents.
Version management
OpenSearch maintains up to two versions of an application:
Online version: The version that is actively serving search queries. All API operations that specify the application name use this version.
Offline version: A newly built version that is not yet published.
The following table describes the version lifecycle during scheduled reindexing.
| Event | Result |
|---|---|
| Scheduled reindexing starts | An offline version is created. If an offline version already exists, the scheduled task automatically deletes it first. |
| Reindexing completes | The offline version is published and replaces the online version. |
| After replacement | The replaced version is automatically deleted within 8 to 16 hours. |
In a manual reindexing task, the offline version must be deleted manually before reindexing. Scheduled reindexing handles this automatically.
Procedure
Log on to the OpenSearch console. In the upper-left corner, select OpenSearch High-performance Search Edition. On the Instance Management page, find your application and click Details in the Action column.
Click Reindex.
Specify a time for the scheduled reindexing task. The minimum interval is 24 hours, so the task runs once per day.
The OpenSearch console displays the progress of reindexing based on the status of each stage (Import data, Process data, Build indexes).
Usage notes
After an offline change, OpenSearch generates an offline version. Both the online version and the offline version then appear on the application details page.
When both versions exist, only the online version serves search queries. API operations that specify the application name use the online version.
When both versions exist, offline changes and reindexing are unavailable. Delete one version and publish the other to resolve this.
Deleting an application is irreversible.