Spring Cloud Gateway Hoxton.SR5服务重启后无法恢复显示问题咨询

Hey there, let's dive into why you're hitting this issue with Spring Cloud Gateway Hoxton.SR5 and how the version downgrade fixes it.

First, Let's Align on the Version Context #

• Hoxton.RELEASE ships with spring-cloud-gateway-core 2.2.0.RELEASE, which works perfectly for your use case.
• Hoxton.SR5 bumps the default spring-cloud-gateway-core version to 2.2.3.RELEASE—and it's the changes between 2.2.0 and 2.2.3 that are causing your problem.

Key Changes in 2.2.1-2.2.3 That Trigger the Issue #

Here are the critical behavior/code changes that lead to the service not being rediscovered after a refresh:

1. Aggressive Service Discovery Cache Implementation
   In 2.2.1+, Spring Cloud Gateway introduced a stricter caching layer for service discovery clients (like Eureka or Consul) to cut down on repeated registry calls. However, this cache wasn't properly invalidated when you called the POST /actuator/gateway/refresh endpoint. Even after your service restarted and re-registered, the gateway stuck with stale cached instance data, so it couldn't see the new service instance.

2. Broken Refresh Logic for Dynamic Service Routes
   In 2.2.3, the RefreshEventListener (which handles gateway refresh events) was modified to optimize static route reloads—but it accidentally broke the reloading flow for service-discovery-based dynamic routes. When you triggered the refresh, only static routes were refreshed; the gateway didn't trigger a fresh pull of service instances from the registry.

3. Flawed Instance Update Event Handling
   2.2.2+ included tweaks to how the gateway listens for service instance change events from the registry. In edge cases like a service that stops and restarts quickly, the gateway fails to pick up the new instance registration event. And since the refresh endpoint doesn't force a full registry sync, the gateway never learns about the restarted service.

Workarounds If You Don't Want to Downgrade #

If you need to stay on Hoxton.SR5 for other dependencies, try these fixes:

• Disable the service discovery cache entirely by adding this to your configuration:

  spring.cloud.gateway.discovery.locator.cache.enabled=false
  

• Use the POST /actuator/gateway/reset endpoint instead of refresh—this endpoint clears all existing routes and triggers a full reload of both static and service-discovery-based routes (note: this will cause brief downtime for gateway routes).

内容的提问来源于stack exchange，提问作者Tonyukuk