Health Check Response Format for HTTP APIs

I'm continuing my journey of getting more familiar with HTTP APIs by reading related RFCs. This time, I read the Health Check Response Format for HTTP APIs at the suggestion of Stefano Fago. In this post, I'd like to summarize my reading.

Note that it's a draft. Moreover, it has been dormant for nearly two years and, thus, has been automatically expired. However, it's the closest to a specification on health checks and thus deserves some love.

Sample Data Visualization

Even though it's not a long read, it's a bit "dry." Fortunately, the specification offers a JSON sample. I copy-pasted it in PlantUML, and presto, it shows a visual representation of it:

What's in store for DevOps in 2023

Our 2023 "DevOps: CI/CD, Application Delivery, and Release Orchestration" report provides developers with the latest information on AIOps and MLOps, IaC, GitOps, automation techniques, and more.

Let's have a look at the proposed structure element by element.

The Root Object

At its simplest, the response is a JSON object with a mandatory status property:

Values can be:

• pass for healthy status: The value can also be ok (for NodeJS) and up (for Spring Boot) to account for existing health check libraries. The HTTP status code must be in the range from 2xx to 3xx.
• warn for healthy status but with concerns with the same HTTP status range
• fail to indicate unhealthy status: possible alternative values include error (NodeJS) and down (Spring Boot). The HTTP status code must be in the range from 4xx to 5xx.

One can add additional optional values:

• version: Public version of the service
• releaseId: Internal version of the service; for example, the version would be incremented for non-compatible change, while the releaseId could be the commit hash or a semantic version
• serviceId: The unique identifier of the service
• description: self-explanatory
• notes: Array of non-structured notes
• output: Plain error message in case of pass or warn.; the field should be left blank for pass

The links Objects

The links object consists of object pairs. Values are URIs, while keys can be URIs or common/registered ones: see RFC5988 for common values; e.g., self.

The checks Objects

Keys of checks objects consist of two terms separated by a colon, component name, and measurement name. The latter can be either:

• A pre-defined value: utilization, responseTime, connections, or uptime
• A standard term from a well-known source; e.g., IANA, microformat.org, etc.

Values consist of one of the following keys:

• componentId: unique id of this component
• componentType:
  • A pre-defined value, component, datastore, or system
  • A standard term from a well-known source; e.g., IANA, microformat.org, etc.
• observedValue: Any valid JSON value
• observedUnit: Unit of measurement
• status: As the parent object's status, but for this component, only
• affectedEndpoints: If the component is not pass, lists all affected endpoints
• time: Date-time in ISO8601 format of the observation
• output: As the parent object's output, but for this component, only
• links: See the previous section
• Any other non-standard value

I tried implementing the above with Spring Boot using a custom HealthIndicator. Here's the best I could come up with:

The current structure of the JSON response needs to be (easily?) customizable. You'd need to create your endpoint. I hope the Spring Boot team provides the option to generate a compatible structure.

Conclusion

The Healthcheck IETF draft is a great initiative to standardize health checks across the industry. It would allow monitoring tools to rely on HTTP status and response body without ad-hoc configuration on each service.

Unfortunately, the draft is expired because of a lack of activity. I'd love to see it revived, though.

To Go Further:

• "Spring Boot Actuator: Health"