Implementing Micro Services Tasks (service discovery, load balancing etc.) - Infrastructure (Docker) vs applicative (Spring Cloud)
6 likes613 views
The document compares microservices frameworks Spring Cloud and Docker, detailing their capabilities in areas like service discovery, configuration management, load balancing, and health check mechanisms. Spring Cloud offers a range of tools for building distributed systems, while Docker provides lightweight containers for deploying applications across various environments. The discussion emphasizes the importance of integrating both systems for efficient microservices architecture management and highlights alternatives like Lightbend's Lagom.
![Configuration Management
• Swarm provides only Secrets.
• K8S provides both Secrets and ConfigMaps.
• Spring provides Config Server.
• E.g.:
@SpringBootApplication
@EnableConfigServer
public class ConfigServer {
public static void main(String[] args) {
SpringApplication.run(ConfigServer.class, args);
}
}](https://image.slidesharecdn.com/tsbzubxnraetv0wkyquo-signature-0c4627150387d9d2343e62af62fa4e43ea9b4863a1d5bab07972c51726576dae-poli-170628141131/85/Implementing-Micro-Services-Tasks-service-discovery-load-balancing-etc-Infrastructure-Docker-vs-applicative-Spring-Cloud-9-320.jpg)
![Service Discovery
• Starting a Eureka service is easy:
@SpringBootApplication
@EnableEurekaServer
public class Application {
public static void main(String[] args) {
new SpringApplicationBuilder(Application.class).web(true).run(args);
}
}](https://image.slidesharecdn.com/tsbzubxnraetv0wkyquo-signature-0c4627150387d9d2343e62af62fa4e43ea9b4863a1d5bab07972c51726576dae-poli-170628141131/85/Implementing-Micro-Services-Tasks-service-discovery-load-balancing-etc-Infrastructure-Docker-vs-applicative-Spring-Cloud-12-320.jpg)
Implementing Micro Services Tasks (service discovery, load balancing etc.) - Infrastructure (Docker) vs applicative (Spring Cloud)
- 1. Micro-Services Stack Spring Cloud vs. Docker
- 2. Micro-Services • Micro-services are the current architectural trend. • We need to deal with: • Discovery. • Configuration. • Load-balancing. • Security. • High-Availability. • Monitoring. • Scheduling.
- 3. The Infrastructure • Many of these concerns can be dealt by either the application framework (e.g., Spring Cloud) or the infrastructure (Docker Swarm/K8S). • Let’s review both and compare them…
- 4. Docker • Docker provides: • Decent isolation. • Evolution. • Distribution (e.g., Swarm, Kubernetes). • Fast start time. • Shared resources.
- 5. Docker • Docker is a lightweight container. • Useful for deploying and running an application/ service/ micro- service with its environment. • You can run your packaged application on production server, staging server, development machine or even a laptop.
- 6. Spring Cloud Spring Cloud provides tools for developers to quickly build some of the common patterns in distributed systems (e.g., configuration management, service discovery, circuit breakers, etc.) http://projects.spring.io/spring-cloud/
- 7. Spring Cloud • Based on Spring Boot. • An umbrella project for many other e.g., Spring Cloud Config, Spring Cloud Netflix, Spring Cloud Security, Spring Cloud AWS and more. • Allows relatively easy implementation of distributed architecture (micro services) design patterns and usecases.
- 8. Spring Cloud Netflix • Netflix OSS integration: • Eureka – Service Discovery. • Hystrix – Circuit Breaker. • Zuul – Routing. • Ribbon – client-side load balancing.
- 9. Configuration Management • Swarm provides only Secrets. • K8S provides both Secrets and ConfigMaps. • Spring provides Config Server. • E.g.: @SpringBootApplication @EnableConfigServer public class ConfigServer { public static void main(String[] args) { SpringApplication.run(ConfigServer.class, args); } }
- 10. Config Server • Default backend is Git repository. • Great for upgrades and auditing. • Can also use the Vault project for storing secrets. • Spring Cloud clients can easily benefit from the Config Server. • There is also Archaius by Netflix which provides dynamic configuration changes to be pushed to the clients.
- 11. Service Discovery • Swarm and Kubernetes provide Service components which serve both as load-balancers and service discovery (DNS). • Spring Cloud provides Eureka & Consul for discovery and Ribbon for load balancing.
- 12. Service Discovery • Starting a Eureka service is easy: @SpringBootApplication @EnableEurekaServer public class Application { public static void main(String[] args) { new SpringApplicationBuilder(Application.class).web(true).run(args); } }
- 13. Service Discovery • On the client side (the micro-services) you just enable Eureka client: @EnableEurekaClient (and provide configuration). • Out-of-the-box support for health indicator (/health) and status url (/info). • You can also use DiscoveryClient to search for other micro- services.
- 14. Load Balancing • Spring Cloud provides Ribbon and Feign to get a client-side load balancer for your services. • K8S and Swarm provide strong support for server-side load balancing! • Based on health-check.
- 15. Health Check • Sometimes something can go wrong (and it will). • E.g., • VM crash (memory, disk or power). • Application is hung (GC, bug). • JVM crashes (OOME, Stack overflow). • A service is unusable (can’t connect to the DB). • First we need to know about it and then (maybe) take an action.
- 16. Possible Actions • The most common action we’ll want to perform is some kind of service restart. • Sometimes we’ll just want to exclude the service from the load- balancer (temporary problem). • Rarely, we’ll want to invoke some specific API of the service.
- 17. Basic Mechanism • Docker is bundled with a basic and simple mechanism, the entrypoint. • A special process in the container, that when exit, stops the container. • An easy way for errors like OOME! • However, usually it’s not enough.
- 18. Health Check (Docker) • Docker provides the HEALTHCHECK instruction for the dockerfile. • You can provide a command that will be invoked inside the container and by its exit code (0 for success) will determine if the container is healthy or not. • You can control the interval, timeout and retries of the check.
- 19. Swarm • Swarm services only load-balance to healthy containers (need at least a single successful health-check). • An unhealthy container will be restarted (another instance). • There is no out-of-box support yet for other actions.
- 20. Kubernetes • Kubernetes ignores the HEALTHCHECK instruction. • Instead, each Pod can be configured with two kinds of probes: • Liveness – a command or HTTP request to check if the container is alive (dead containers will be restarted). • Readiness – a command or HTTP request to check if the container is ready (unready containers will not belong to the load-balancer).
- 21. Health-check in Spring Cloud • Can use Spring Actuator (from Spring Boot) for built-in “/health” URL. • Combining both Spring Boot and Docker support can be very useful!
- 22. Hystrix • Provides a circuit breaker pattern implementation. • What is a circuit breaker?
- 23. Circuit Breaker • Let’s say we fail to connect to some service (request will timeout). • Usually the problem will take some time to fix. • In the meanwhile our service might crash (especially when we’re under load and there’s no pushback). • Circuit breaker to the rescue…
- 24. Circuit Breaker • Can be in 3 states: • Closed – normal work. • Open – will not forward calls to the failing service. Works with a fallback. • Half-Open – a single invocation will continue to the failing service to check the status.
- 25. Hystrix • Supports: • Fallbacks. • Dashboard.
- 26. Discussion • With Docker, you have a language-independent solution! • With Spring Cloud you only get JVM-based support. • Docker-based solutions make deployment extremely easy.
- 27. Discussion • Spring Cloud make client-side solutions (e.g., Hystrix & Ribbon) extremely powerful. • Combining both (Spring Cloud on Swarm/K8S) works great.
- 28. Discussion • Current trends lead into K8S/Swarm cluster management. • For managing: • Cluster Resilience. • Scale. • Server-side Load Balancing. • Secrets. • Jobs.
- 29. Discussion • So, where does it leave use regarding Spring services? • We can still benefit from: • Hystrix. • Config Server. • Internal application framework.
- 30. Alternatives • Spring Cloud is not the only possible application framework solution. • Lightbend’s Lagom can also be considered. • Combines: • Akka + Play. • Cassandra.
- 31. Akka Cluster • Can provide: • Client-side load-balancing. • Scaling. • Persistence. • Sharding. • Cluster-wide singleton. • Event bus.