SlideShare a Scribd company logo

Building Bizweb Microservices with Docker

•
•
Khôi Nguyễn Minh
Khôi Nguyễn Minh

This document discusses using Docker to build and deploy microservices for a Bizweb application. It proposes using Docker Swarm mode for orchestration along with Netflix OSS components like Eureka, Zuul and Ribbon to provide service discovery, API gateway functionality and load balancing between microservices. Continuous integration is achieved using Jenkins to build Docker images from code commits, tag them and push to a private Docker registry. When code is committed, Jenkins triggers deployment of updated services to a Docker Swarm cluster for testing or production. This approach aims to reduce delivery time, simplify deployment and scaling of microservices.

Related slideshows

Toi uu hoa he thong 30 trieu nguoi dung
Toi uu hoa he thong 30 trieu nguoi dungToi uu hoa he thong 30 trieu nguoi dung
Toi uu hoa he thong 30 trieu nguoi dung
 
itlchn 20 - Kien truc he thong chung khoan - Phan 2
itlchn 20 - Kien truc he thong chung khoan - Phan 2itlchn 20 - Kien truc he thong chung khoan - Phan 2
itlchn 20 - Kien truc he thong chung khoan - Phan 2
 
Go micro framework to build microservices
Go micro framework to build microservicesGo micro framework to build microservices
Go micro framework to build microservices
 
1 of 25
Download to read offline
Building Bizweb Microservices with Docker Nguyễn Minh Khôi CTO of DKT Technology dkt.com.vn
Programming Languages Frameworks & Libraries Web ServersDatabases Message Queues |S3, EC2, Route53 Cloud Services Others Bizweb Tech Stack 2
Bizweb Microservices Components • Based on Spring Boot & Netflix OSS • Service Discovery: Eureka (Server), Ribbon (Client) • API Gateway: Zuul • Centralized Configuration: Spring Cloud Config • API Security: Spring Security & Spring Security OAuth • REST API: Spring Boot • Job Service: Kafka & Spring Boot 3
Bizweb Microservices Architecture 4
Eureka Service Discovery • Each service has unique serviced • Service uses Eureka Client to interact with Eureka Server: • Register: serviceId, host, port • Renew: using heartbeats to check status • Get Registry: return list host:port of services by serviceId 5
Zuul API Gateway • Single entry point for clients to call API • Edge service: not use to call between microservices • Using Ribbon to call microservices • Client Load Balancer • Caching • Fault tolerance • Zuul Filter: • Handle request in pipeline mode • PRE, ROUTING, POST, ERROR 6
Centralized Configuration 7 Problems: - Stored on each service, hard to control - Hard to update new configuration - Reload config at runtime
Centralized Configuration 8
Problems • Take times to deploy on new servers: • Install Java • Copy fat .jar file (~75-100MB) using FTP/SCP • Make script to run as a Linux service • Take times to update services: • 20 microservices + job services • Manual update on multiple hosts • Manual scale & choose server to deploy • Quite hard to monitor these microservices 9
Solved with Docker & Jenkins 10
Simple Dockerfile for all services FROM frolvlad/alpine-oraclejdk8:slim ADD lib lib ADD product.jar app.jar RUN sh -c 'touch /app.jar' ENTRYPOINT ["java","-Xmx128m","-Xms128m","- Djava.security.egd=file:/dev/./urandom","-jar","/app.jar"] 11
Docker Swarm Mode • Built-In Orchestration • Easy to start docker swarm init docker services create --name product product:1.2.0 • Secure by default docker swarm join --token [manager_token|worker_token] • Easy to scale docker service scale product=10 • Rolling updates docker service update --update-delay 1m --update-parallelism 2 --image product:1.2.1 product 12
Swarm Manager & Worker 13
Swarm Manager & Worker 14 • Maintaining cluster state • Scheduling services • Use RAFT consensus to elect new leader
Swarm Manager & Worker 15 • Pull tasks from Managers • Receive tasks in batch (100ms or 100 tasks) • Manager node can be worker, too
Combine Netflix OSS with Docker Swarm • Zuul for API Gateway • Eureka for Service Discovery • Ribbon Client for direct call between microservices • Docker Swarm: • Manage microservices • Deploy, scale, update microservices 16
Docker Swarm Deployment • 3 manager nodes on 3 different physical machines • Workers on Physical & Virtual Machines • Using overlay network: • Communicate with Eureka Server • Direct call between microservices • Label for services & environment: • INFRASTRUCTURE: running Zuul, Eureka, Config, OAuth Service • SERVICE: running microservices • JOB: running Job services • DEV: running all containers of dev environment 17
Docker Swarm Cluster 18 swarm manager (primary) config server eureka server zuul server swarm manager (secondary) config server eureka server zuul server swarm manager (secondary) config server eureka server zuul server worker-01 microservice-02 microservice-xx microservice-01 worker-xx microservice-02 microservice-xx microservice-01 worker-job job-02 job-xx job-01 oauth server oauth server oauth server Bizweb Overlay Network server-01 redis INFRASTRUCTUREINFRASTRUCTUREINFRASTRUCTURE SERVICE SERVICE JOB worker-dev job-yy microservices-xx DEV Docker Swarm Cluster server-02 redis server-03 redis Redis Cluster
CI with Jenkins & Docker • Environment Prerequisites: • Gitlab 8.12 (support webhooks) • Jenkins 2 • Docker Registry 2.0 • Run on Docker 19 Data Center Zone Office Dev Zone VPN, IP Restriction local dev server
20 1. Developer pushes to Gitlab repo 2. Gitlab webhook triggers Jenkins 3. Jenkins pull code from Gitlab: - source code - Dockerfile - Jenkinsfile 4. Compile source code & build Docker image, auto tag image 5. Push image to Private Docker Registry 6. Run custom script based on branch committed
CI with Jenkins & Docker • Using Spotify docker-maven-plugin: • Save space by caching java libraries image layer (~75MB) • Reduce network traffic & deploy time (only 200-700KB transferred) • Docker image auto tag: {git_commit_short_code}-{branch} -> 4b4a71ef-dev • Custom script based on branch committed: • dev: trigger another Jenkins Server to update service • live: manual update 21
Summary • Docker is great way to package microservices • Building HA system is easier with Docker Swarm Mode • CI with Jenkins will reduce delivery time, increase productivity 22
References • http://www.slideshare.net/juminchoi/bizweb-microservices- architecture • https://docs.docker.com/engine/swarm/ • https://www.docker.com/use-cases/cicd 23
Contact • Nguyễn Minh Khôi – DKT Technology • Email: khoinm@dkt.com.vn • Facebook: https://fb.com/khoinguyen84 24
Thank you! Q&A

More Related Content

Building Bizweb Microservices with Docker

  • 1. Building Bizweb Microservices with Docker Nguyễn Minh Khôi CTO of DKT Technology dkt.com.vn
  • 2. Programming Languages Frameworks & Libraries Web ServersDatabases Message Queues |S3, EC2, Route53 Cloud Services Others Bizweb Tech Stack 2
  • 3. Bizweb Microservices Components • Based on Spring Boot & Netflix OSS • Service Discovery: Eureka (Server), Ribbon (Client) • API Gateway: Zuul • Centralized Configuration: Spring Cloud Config • API Security: Spring Security & Spring Security OAuth • REST API: Spring Boot • Job Service: Kafka & Spring Boot 3
  • 5. Eureka Service Discovery • Each service has unique serviced • Service uses Eureka Client to interact with Eureka Server: • Register: serviceId, host, port • Renew: using heartbeats to check status • Get Registry: return list host:port of services by serviceId 5
  • 6. Zuul API Gateway • Single entry point for clients to call API • Edge service: not use to call between microservices • Using Ribbon to call microservices • Client Load Balancer • Caching • Fault tolerance • Zuul Filter: • Handle request in pipeline mode • PRE, ROUTING, POST, ERROR 6
  • 7. Centralized Configuration 7 Problems: - Stored on each service, hard to control - Hard to update new configuration - Reload config at runtime
  • 9. Problems • Take times to deploy on new servers: • Install Java • Copy fat .jar file (~75-100MB) using FTP/SCP • Make script to run as a Linux service • Take times to update services: • 20 microservices + job services • Manual update on multiple hosts • Manual scale & choose server to deploy • Quite hard to monitor these microservices 9
  • 10. Solved with Docker & Jenkins 10
  • 11. Simple Dockerfile for all services FROM frolvlad/alpine-oraclejdk8:slim ADD lib lib ADD product.jar app.jar RUN sh -c 'touch /app.jar' ENTRYPOINT ["java","-Xmx128m","-Xms128m","- Djava.security.egd=file:/dev/./urandom","-jar","/app.jar"] 11
  • 12. Docker Swarm Mode • Built-In Orchestration • Easy to start docker swarm init docker services create --name product product:1.2.0 • Secure by default docker swarm join --token [manager_token|worker_token] • Easy to scale docker service scale product=10 • Rolling updates docker service update --update-delay 1m --update-parallelism 2 --image product:1.2.1 product 12
  • 13. Swarm Manager & Worker 13
  • 14. Swarm Manager & Worker 14 • Maintaining cluster state • Scheduling services • Use RAFT consensus to elect new leader
  • 15. Swarm Manager & Worker 15 • Pull tasks from Managers • Receive tasks in batch (100ms or 100 tasks) • Manager node can be worker, too
  • 16. Combine Netflix OSS with Docker Swarm • Zuul for API Gateway • Eureka for Service Discovery • Ribbon Client for direct call between microservices • Docker Swarm: • Manage microservices • Deploy, scale, update microservices 16
  • 17. Docker Swarm Deployment • 3 manager nodes on 3 different physical machines • Workers on Physical & Virtual Machines • Using overlay network: • Communicate with Eureka Server • Direct call between microservices • Label for services & environment: • INFRASTRUCTURE: running Zuul, Eureka, Config, OAuth Service • SERVICE: running microservices • JOB: running Job services • DEV: running all containers of dev environment 17
  • 18. Docker Swarm Cluster 18 swarm manager (primary) config server eureka server zuul server swarm manager (secondary) config server eureka server zuul server swarm manager (secondary) config server eureka server zuul server worker-01 microservice-02 microservice-xx microservice-01 worker-xx microservice-02 microservice-xx microservice-01 worker-job job-02 job-xx job-01 oauth server oauth server oauth server Bizweb Overlay Network server-01 redis INFRASTRUCTUREINFRASTRUCTUREINFRASTRUCTURE SERVICE SERVICE JOB worker-dev job-yy microservices-xx DEV Docker Swarm Cluster server-02 redis server-03 redis Redis Cluster
  • 19. CI with Jenkins & Docker • Environment Prerequisites: • Gitlab 8.12 (support webhooks) • Jenkins 2 • Docker Registry 2.0 • Run on Docker 19 Data Center Zone Office Dev Zone VPN, IP Restriction local dev server
  • 20. 20 1. Developer pushes to Gitlab repo 2. Gitlab webhook triggers Jenkins 3. Jenkins pull code from Gitlab: - source code - Dockerfile - Jenkinsfile 4. Compile source code & build Docker image, auto tag image 5. Push image to Private Docker Registry 6. Run custom script based on branch committed
  • 21. CI with Jenkins & Docker • Using Spotify docker-maven-plugin: • Save space by caching java libraries image layer (~75MB) • Reduce network traffic & deploy time (only 200-700KB transferred) • Docker image auto tag: {git_commit_short_code}-{branch} -> 4b4a71ef-dev • Custom script based on branch committed: • dev: trigger another Jenkins Server to update service • live: manual update 21
  • 22. Summary • Docker is great way to package microservices • Building HA system is easier with Docker Swarm Mode • CI with Jenkins will reduce delivery time, increase productivity 22
  • 24. Contact • Nguyễn Minh Khôi – DKT Technology • Email: khoinm@dkt.com.vn • Facebook: https://fb.com/khoinguyen84 24