재시도 및 시간 제한 설정

In microservices, services can fail or become slow. To keep our applications running smoothly, we need to build resilience.

• Resilience means your system can recover from failures and continue to function.
• Spring Cloud Gateway provides tools to make your API Gateway more resilient.
• Two key strategies for resilience are Timeouts and Retries.

A timeout is a limit on how long an operation is allowed to take. If the operation doesn't complete within that time, it's automatically stopped.

• Timeouts prevent requests from hanging indefinitely.
• They free up resources (like network connections and threads) that would otherwise be tied up by a slow or unresponsive service.
• In a gateway, timeouts ensure that a slow backend service doesn't slow down the entire gateway or other requests.

Spring Cloud Gateway allows you to configure specific timeouts for routes. The ReadTimeout filter is commonly used to limit how long the gateway waits for a response from the backend service after the connection is established.

• This timeout is applied per route.
• It helps prevent a single slow backend from impacting the gateway's overall performance.
• The value is typically set in milliseconds.

Here's how you can configure a ReadTimeout for a specific route in your application.yml. This example sets a 5-second read timeout for requests to /service-a/**:

Remember, this is part of your Spring Boot application's configuration.

Retries involve automatically re-sending a request that has failed, hoping it will succeed on a subsequent attempt.

• They are ideal for transient failures: temporary issues like network glitches or a brief service restart.
• Retries should be used cautiously, especially for non-idempotent operations (actions that produce different results if performed multiple times).
• Spring Cloud Gateway can be configured to automatically retry requests to backend services.

Spring Cloud Gateway provides a Retry filter to enable automatic retries for failed requests. You can configure various aspects of the retry logic:

• retries: The maximum number of retry attempts.
• statuses: HTTP status codes that should trigger a retry (e.g., 503 for Service Unavailable).
• methods: HTTP methods that can be retried (e.g., GET, PUT).

Let's configure a Retry filter. This example retries requests up to 3 times if the backend returns a 5XX error or a 404, specifically for GET requests to /service-b/**:

While Spring Cloud Gateway handles retries via configuration, the core concept involves looping until success or max attempts. Here's a basic Java program demonstrating a retry loop:

Timeouts and retries often work together:

• A timeout can trigger a retry if the request doesn't complete within the specified time.
• If a request times out, the gateway might then attempt a retry, hoping the next attempt will be faster or the service will respond.
• It's crucial to configure these carefully to avoid endless loops or excessive delays. For example, a retry might happen *after* a read timeout, or a global timeout could encompass all retries.

When implementing timeouts and retries, consider these best practices:

• Idempotency: Only retry idempotent operations (e.g., GET, PUT) unless you have specific logic to handle non-idempotent ones (e.g., POST).
• Exponential Backoff: Introduce increasing delays between retries to avoid overwhelming a struggling service.
• Circuit Breakers: Combine with circuit breakers (covered in Lesson 10.1) to stop retrying services that are clearly down.
• Monitoring: Monitor retry counts and timeouts to identify persistently problematic services.

You're configuring a Spring Cloud Gateway route for a backend service that sometimes experiences brief network hiccups, resulting in a 503 Service Unavailable error. You want the gateway to automatically try the request again a few times before giving up. Which filter and configuration would best achieve this?

We've explored how Timeouts and Retries are fundamental for building resilient API Gateways with Spring Cloud Gateway.

• Timeouts prevent requests from hanging, freeing up resources.
• The ReadTimeout filter configures the wait time for backend responses.
• Retries automatically re-send requests for transient failures.
• The Retry filter allows fine-grained control over retry attempts, statuses, and HTTP methods.
• Combining these with best practices helps create robust microservice architectures.