Is there a notice function inside eureka while applications shutting down

[WilliamChen-luckbob]

I'm new in java, in recent days I'm trying to develop a graceful shut down script for my springboot applications, but this confused me a lot...

I have this little question while observing what will happen when I shut down one of my service application.

here is the structure of my testing applications:

1. Eureka server service (E)
2. one consumer serivce (A)
3. two service provider service (B1 and B2) A calls B by feign client and returns which B is invoked(by using returning string 'service B at [it's port] is invoked').

I have found that no matter when I tried to shut down one of service B in a graceful way by spring boot acurator, there is a strange thing: I though this shut down instance will still be stay in eureka's instance list and also, in those applications' local cache because A and E may not have chance to reach the fetch or refresh time for new instance list after one of B was down.

So, in this case, I thought there might be a loadbalancer exception thown out because eureka client in A do not fetch a new instance list.

But what I observed was totally different from what I thought.

It seems to have a sneaky method to cancel the shutdown application's instance from eureka and notice all the applications which will call B for a downstream service immediately. All request will be send to B2 after a little time delay if I shut down B1 by sending /actuator/shutdown to B1.

Parent of the pom is:

    <parent>
        <groupId>org.springframework.boot</groupId>
        <artifactId>spring-boot-starter-parent</artifactId>
        <version>2.3.2.RELEASE</version>
        <relativePath/> <!-- lookup parent from repository -->
    </parent>

Here are my Eureka configuration: eureka server :

eureka:
  server:
    eviction-interval-timer-in-ms: 30000
  client:
    service-url:
      defaultZone: http://${my-config.eureka.instance.hostname}:8000/eureka/
    register-with-eureka: false

client A and B:

eureka:
  instance:
    lease-renewal-interval-in-seconds: 30
    lease-expiration-duration-in-seconds: 90
  client:
    service-url:
      defaultZone: http://${my-config.eureka.instance.hostname}:8000/eureka/
    registry-fetch-interval-seconds: 30

Well, after plenty of jmeter test, it seems that something was doing a quick refresh of alive instance ignoring the eureka server. local cache will refresh automatically when B was shut down( even shut down by kill -9 , nothing wrong will happen if there are at least one B stiill alive.)

[troshko111]

Sorry it's not entirely clear to me what you're trying to achieve? What's the observed behavior vs the expected behavior in your test setup?