It appears that in discovery-first mode, the config client is not updating its location information for the config server when that data changes in eureka (i.e. when a config server instance is removed or added in eureka). Is this currently expected behavior, or am I doing something wrong?
my bootstrap config:
spring:
cloud:
config:
enabled: true
failFast: true
discovery:
enabled: true
eureka:
client:
serviceUrl:
defaultZone: [[my eureka defaultZone]]
When I take down the first config server instance, the configserver healthcheck in the app begins to fail despite there being a second working node registered with eureka. The same occurs if I restart the config server node with a new password in the eureka metadataMap, in which case the client keeps trying to use the old password.
spring.cloud.config.discovery.enabled=true is currently built on a one time eureka lookup. It is not built with ribbon (ie a client-side load balancer).
Is this something that you might like to see a PR on?
Yes, we'd want to make sure it's well tested and easy to understand.
I'll give it a shot
DiscoveryClientConfigServiceBootstrapConfiguration heartbeat event triggers the refresh config. Uri. It holds EurekaClient is old, the EurekaClient should be a singleton, but the introduction of the config is created twice, causing it to refresh for the config address is still not available
any update?
any update?
+1 any update?
If there was an update it would have been posted here
any update?
any update ?
I think the analysis done here can be useful spring-cloud/spring-cloud-netflix#3675
During startup a DiscoveryClient is created on bootstrap context and then shut down. Then DiscoveryClient is restarted, but ConfigServerInstanceProvider continues to use the shutdown bean, getting no updates.
This makes discovery-first bootstrap working only if config server is already registered on Eureka when config client starts.
I have same issue with ReactiveEurekaDiscoveryClient. It is configured with DiscoveryClient which is used during bootstrap, then this client is shut down and never refresh, and loadbalancer never gets instances updates.
You can subscribe to the spring-cloud/spring-cloud-netflix#3675 questi
There's no need to comment on multiple issues. GitHub links them automatically.
I this issue fixed now ? If yes then which version we can follow ?
Most helpful comment
Is this something that you might like to see a PR on?