Health Check

A health check validates the status of an API and its dependent data before a new index is published in Sparque. It ensures that the required objects are available so that the indexing process can proceed safely.

• If the health check passes (i.e., objects are returned), the new index will be published.
• If it fails (i.e., no objects are returned), the new index is NOT created and the old index is used.

1. Execution timing: Automatically triggered before each new index is published.
2. API endpoint: Each API within a workspace must have an endpoint called /health_check.
3. Object verification: The health check verifies that the API returns objects. If one or more objects are returned, it is considered successful.
4. Health check failure: If no objects are returned, the index creation process stops and the old valid index is used.

1. Initial index execution: A new index is about to be published.
2. Triggering the health check: The health check is performed before the index is published.
3. Health check output:
   • Success: The /health_check endpoint returns one or more objects and the new index is published.
   • Failure: No objects are returned. The index creation is stopped and the system falls back to the old index.

The system will repeat above steps whenever a new index is triggered.

To enable health checking for your API, you need to configure a /health_check endpoint within each API that returns a relevant object. The assigned strategy allows you to specify the objects that Sparque must check before creating the index.

Configuration steps:

1. Define the /health_check endpoint:
   • Each API should expose an endpoint named /health_check.
   • This endpoint should return a valid response containing objects.
2. Configure required objects:
   • In the API configuration strategy, define which objects the health check should validate. This could include data files, configuration files, or other resources that are essential for a successful index creation.

This visualization shows an assigned strategy for the /health_check endpoint. It checks safety levels for each class (e.g., products = 1000, category = 200, order = 750). If any class has insufficient objects (e.g., if the index has 889 products instead of 1000 or more), the safety level is not reached and no objects are returned (Top N = 0), causing the health check to fail.

If the health check fails:

1. Investigate the response: Check in Sparque desk which objects are missing or invalid.
2. Correct the issue: Fix any issues with the data or within the assigned health_check strategy.

The health check is a critical safeguard in Sparque to ensure that new indexes are only created when the necessary resources are available and valid. By correctly configuring the /health_check endpoint, you can help prevent issues such as missing or corrupt data, ensuring smooth index creation and stable system operation.

• Each API must expose a health_check endpoint.
• Each health check must return objects for the index process to proceed.
• If one of the health check fails, the index creation process will be stopped and will be repeated within the next iteration.