Background:
We have a set of about 15 Spring Boot applications as microservices. They all run as Docker containers, and run as clusters of one or more instances. We also use Spring Cloud Netflix components such as Eureka and discover the running application instances from the client using Feign/Ribbon.
Question:
As part of the POC exercises, we tested with Redis and Hazelcast for caching and Spring Boot configuration storage. Everything works using Spring Boot, Spring Cloud and Redis/Hazelcast Java client libraries. However, when we deploy Hazelcast in a multi-node peer-to-peer cluster, Hazelcast seems to require a "known" IP address/hostname and the accessible port to be available in the Java client's configuration (with or without Spring). Typically, when Hazelcast is deployed in a multi-instance cluster on ephemeral VM instances (for example, EC2), the IP address and the port information is not available. So we thought of two possible solutions:
Find a way to run Hazelcast as a Spring Boot application, and register it with Eureka as a Discovery Client. That way other Spring Boot applications can use Eureka to discover an instance of Hazelcast dynamically
Find a way in Hazelcast so that it can publish it's IP address and port information dynamically to a central Key/Value store
If anyone has played around with Hazelcast to be able to do either/both of the possible solutions, it would be great if you can share more information on that. If there is a third approach that'd work better, I will be eager to know that as well.
Related
I'm trying to see how to configure Client-Server with my Spring Boot application using Hazelcast on Kubernete, since we want to have the capability of sharing the cache between different Spring Boot applications (I'm already able to setup the Embedded distributed cache with Kubernetes - which is not what we need).
In case of Spring Boot single application(not on Kubernetes), its kind of easy where i will spin up a Server lets say with 'localhost' and also spin up the client connecting to localhost. Also i can have multiple instances(members) of Server which will form a Hazelcast Cluster.
However in case of Kubernetes, I know we need to have 2 different Spring Boot applications, one will act as a Server and others will be client accessing the cache, but want to know how the client would connect to the Server. Because in case of Spring we Autowire the HazelcastInstance, so how would i connect to the Server which is running in its own Kubernetes Pod ( and container).
There are a few deployment guides for Kubernetes here, and worked examples here & here.
If your server pods are joining, then you pretty much have it. A client uses the same discovery mechanism.
Add the hazelcast-kubernetes plugin to the client's pod, and set the configuration properties with the same values as you use on the server for namespace, dns, etc.
Thanks Neil. As you indicated its the same way i currently configured in embedded caching on Kubernetes. I'am using the Service-Name and namespace to discover and connect to the Server members from HazelcastClient instance.
This is from Client Spring Boot application's HazelcastConfiguration:
#Bean
public HazelcastInstance hazelcastInstance() {
final ClientConfig config = new ClientConfig();
config.setClusterName("cluster-name");
if (enableOnK8s) {
config.getNetworkConfig().getKubernetesConfig().setEnabled(true)
.setProperty("namespace", namespaceValue)
.setProperty("service-name", serviceName);
}
return HazelcastClient.newHazelcastClient(config);
}
And on Hazelcast Server Spring Boot application, configuration stays same as Embedded Hazelcast configuration.
I am setting up the Spring-boot microservices with the cluster bi-direction Pivotal cloud cache.
I have set up the bi-directional cluster in Pivotal Cloud, I have a list of locators with ports.
I have already some online docs.
https://github.com/pivotal-cf/PCC-Sample-App-PizzaStore
But couldn't understand the on which configuration the spring boot app will know to connect.
I am looking for some tutorial or some reference where I can have spring boot app linked up with the PCC(gemfire)
The way you configure a app running in PCF (Pivotal Cloud Foundry) to talk to a PCC (Pivotal Cloud Cache) service instance is by binding the app to that service instance. You can bind it either by running the cf bind command or by adding the service name in the app`s manifest.yml, something like the below
path: build/libs/cloudcache-pizza-store-1.0.0-SNAPSHOT.jar
services:
- dev-service-instance
I hope you are using Spring Boot for Apache Geode & Pivotal GemFire (SBDG) in your app, if not I recommend you to use it as it makes connecting to PCC service instance extremely easy. SBDG has the logic to extract credentials, hostname:ports needed to connect to a service instance.
You as a app developer just need to
Create the service instance.
Bind your app to the service instance.
The boilerplate code for configuring credentials, hostnames, ips are handled by SBDG.
When you deploy an application in Cloud Foundry, (or Pivotal Cloud), you need to bind it to one or more services. Service details are then automatically exposed to the app via the VCAP_SERVICES environment variable. In the case of PCC this will include the name and port of the locator. By adding the spring-geode-starter (or spring-gemfire-starter) jar to the application it will automatically process the VCAP_SERVICES value and extract the necessary endpoint information in order to connect to the cluster.
Furthermore, if security is enabled on your PCC instance, you will also need to have created a service key. As with the locator details, the necessary credentials will be exposed via VCAP_SERVICES and the starter jar will automatically process and configure them.
Was wondering the recommended way of starting a spring boot app if the Spring cloud config server is temporarily down or unavailable. What would be the approach? I know of the retry configurations, but I am wondering if there is a way to have a 'replica' config server and use that as a failover (or something along those lines).
Sure, why not?
After all, spring-cloud-config server exposes rest API and all the interaction with spring boot microservices is done over HTTP.
From this point of view, you can scale out the spring cloud config server by providing more than one instance of it all are up-and-running and mapping them to one virtual IP.
If you're running in some kind of orchestrated environment (like kubernetes) it is a very easy thing to do.
I am developing a microservice, using Spring Boot, that exposes REST Endpoint. Because of scalability, I have to run multiple instances of this services on a different port. What will be the configurations for the applications so that it can register with eureka and requests are load balanced? I am using Spring cloud config, Eureka server and zuul.
Attaching following entries in the client properties file will do the trick. This is for Spring cloud config dalston
eureka.instance.instanceId=${spring.application.name}:${spring.application.instance_id:${random.value}}
I guess you meant to register with Eureka instead of Config server.
To register multiple instances that might be running in the same host but listening on a different port you would need to set eureka.instance.metadataMap.instanceId to a unique value maybe using:
eureka.instance.metadataMap.instanceId=${spring.application.name}:${random.int}
Is Spring cloud config server an example of sidecar application for microservices?
Do you mean if the Spring Cloud Config Server itself is what the Spring Cloud documentation labels as Sidecar? Then no, as far as I know it is just a plain, regular Spring Boot app.
A Sidecar as referred to in Polyglot support with Sidecar is a Spring Boot application that acts as a bridge between your service infrastructure and a service that is not written in a JVM language. Apps written in Python, Go, Ruby, C#, NodeJS, Erlang or really any other language that can bind something to a port come to mind.
The benefits of the Sidecar are, that your Non-JVM apps
service discovery become automatically discoverable through Eureka, which means that JVM services can resolve the host:port/<service-id> of the Non-JVM apps as well as the other way around,
monitoring are monitorable through the same health-endpoints-infrastructure that is available in Spring Boot (Actuator), i.e. by manually providing the health endpoint in the Non-JVM app Eureka knows when the Non-JVM service is down
routing/proxying query the services by either manually looking up their hosts/ports or proxying these requests through Zuul, which in turn resolves their current addresses through Eureka
balancing be load balanced by Ribbon and
configuration may consume configuration properties provided via Spring Cloud Config.
I hope this answer addresses your question, if not (or someone finds it to be inaccurate or misleading) just let me know and I delete it to make room for something more suitable. ;-)