Microservices Architecture
- 1. Microservices Lucian Neghina Big Data & Cloud Computing by Developer for Developers
- 2. Before Microservices Traditional SOA Looser coupling Elements must be coordinated each other Monolithic Tight coupling All elements must agree on each change Microservice Decoupled Elements can be changed or added without prior coordination with others
- 3. Microservice benefits Autonomy Autonomous team with isolated implementation Speed of change Build and deploy small independent services Scale Functional and Team Scaling Tech Diversity Allows try-out of new technologies Intelligible Lower cognitive load
- 6. Organized around business capabilities UI Server DBA Shipping Orders Catalog
- 8. Disadvantages of distributed computing ● Latency ● Partial failure ● Infrastructure
- 9. Designing Microservices ● Size Single Responsibility Principle ● Scope Agile / Independent development and deployment ● Capabilities Bounded Context in Domain-Driven-Design
- 11. Communication
- 12. Interaction Styles Synchronous Asynchronous Request / Response -- Notification Publish / Subscribe Request / Async response Publish / Async response One-to-One One-to-Many
- 13. Point-to-point style Inventory microserviceREST API Accounting microserviceREST API Checkout microserviceREST API Consumer Store microserviceREST API
- 14. Api-Gateway style Inventory microserviceREST API Accounting microserviceREST API Checkout microserviceREST API Consumer Store microserviceREST API API Gateway
- 15. Message-Broker style Inventory microservice Accounting microservice Checkout microservice Store microservice Consumer publisher subscriber publisher publisher subscriber subscriber
- 16. Service Discovery Server-side and Client-side
- 17. Why use Service Discovery ● Dynamically network location ● Autoscaling ● Failures ● Upgrades
- 18. Server-Side Discovery Inventory microservice Instance 1 REST API Registry Client Inventory microservice Instance 2 REST API Inventory microservice Instance 3 REST API Inventory stack Service Registry Registry Client Registry Client Register Consumer Service Request Router Request Discover service Service location
- 19. Client-Side Discovery Inventory microservice Instance 1 REST API Registry Client Inventory microservice Instance 2 REST API Inventory microservice Instance 3 REST API Inventory stack Service Registry Registry Client Registry Client Discover service Register Consumer Service Request Service location
- 21. Circuit Breaker
- 22. Hystrix Fallback
- 24. Problem of Distributed Data Checkout Context Order Accounting Context Custommer id customer_id satus total id credit ... CUSTOMERORDER
- 25. Event-Driven Architecture Checkout Context Order Accounting Context id customer_id satus total 909 101 OPEN 1234 id credit ... 202 5000 Place Order Order Create event Order Create event Message Broker Credit Reserved Credit Reserved Custommer CUSTOMER ORDER cust_id order_id amount 202 909 1234 RESERVED_CREDIT
- 26. Security Patterns and Best Practices
- 27. Security Model Availability is guarantee of reliable access to the information by authorized people. Integrity is the assurance that the information is trustworthy and accurate. Confidentiality is set of rules that limits access to information.
- 28. API Gateway / Perimeter security ● Requests are authenticated and authorized by the gateway ● The public LB cannot send request to apps directly ● Apps trust all traffic they receive by assumption Pros ● Network setup can virtually guarantee assumptions ● Apps have stateless security Cons ● Does noting for internal threats
- 29. Basic + Central Auth DB ● All apps get to do authentication and authorization themselves ● Credentials are verified against a central DB ● Basic credentials are passed along in every request Pros ● Stateless - authenticate every time ● Central user store Cons ● Auth DB is hit every request ● DB lookup logic needs to be implemented everywhere ● User’s credentials can unlock all functionality
- 30. Sessions everywhere ● Same as before but each app gets to maintain a session with the client device Pros ● Auth DB is hit once per session Cons ● Hard to manage all the sessions ● No single sign on ● DB lookup logic needs to be implemented everywhere ● User’s credentials can unlock all functionality
- 31. API Tokens ● Username and password is exchanged for a token at a centralized auth server ● Apps validate the token for each request by hitting the auth server Pros ● App don’t see user credentials Cons ● Auth server bottleneck ● Token provides all or nothing access
- 32. Microservices Security Concerns ● Central user store bottleneck ● Single Sign On ● Statelessness ● User credentials ● Fine grained authorization ● Interoperability with non browser
- 33. OAuth2 and OpenID Connect Inventory microserviceREST API Accounting microserviceREST API Checkout microserviceREST API Consumer Store microserviceREST API API Gateway Authorization Server User Store JWT processing at each microservice Access Token Access Token JWT JWT
- 35. Testing Pyramid End-to-End Tests Contract Tests Component Tests Integration Tests Unit Tests TimeRequiredtoExecute NumberofTests FrequencyofExecution Verisimilitude
- 36. Unit Testing Make sure that the each component behave as expect. ● Domain ● Gateways ● Resources ● Persistence
- 37. Integration Testing Verifying the communications and interactions between components ● Persistence ● Gateways
- 38. Component Testing Testing the service in isolation ● In Memory DB ● Stub HTTP Client ● Simulate external service
- 39. Contract Testing Verifying the agreements between producer and consumers services ● Input & Output data structure ● Calls attributes
- 40. End-to-End Testing Testing the behaviour of the entire system ● Business flow ● Proxies ● Load Balancers
- 42. Multiple Service Instances per Host Pros ● Efficient resource utilization ● Fast deployment Cons ● Poor isolation ● Difficult to limit resource utilization ● Risk to dependency version conflicts ● Poor encapsulation of implementation technology
- 43. Service Instance per Virtual Machine Pros ● Great isolation ● Great manageability ● VM encapsulates implementation technology Cons ● Less efficient resource utilization ● Slow deployment
- 44. Service Instance per Container Pros ● Great isolation ● Great manageability ● Container encapsulates implementation technology ● Efficient resource utilization ● Fast deployment Cons ● Immature infrastructure for deploying containers
- 46. Rolling Update Instance 1 V1 Instance 2 V1 Instance 3 V1 Old Group New Group Router
- 47. Rolling Update Instance 1 V1 Instance 2 V1 Instance 3 V1 Old Group Instance 1 V2 New Group Router
- 48. Rolling Update Instance 1 V1 Instance 2 V1 Instance 3 V1 Old Group Instance 1 V2 New Group Router
- 49. Rolling Update Instance 2 V1 Instance 3 V1 Old Group Instance 1 V2 Instance 2 V2 New Group Router
- 50. Rolling Update Instance 3 V1 Old Group Instance 1 V2 Instance 2 V2 Instance 3 V2 New Group Router
- 51. Rolling Update Old Group Instance 1 V2 Instance 2 V2 Instance 3 V2 New Group Router
- 52. Canary Release Old Version Instance 1 V2 Instance 2 V2 Instance 3 V2 New Version Instance 1 V1 Instance 2 V1 Instance 3 V1 All users - 100% 0% Router
- 53. Canary Release Old Version Instance 1 V2 Instance 2 V2 Instance 3 V2 New Version Instance 1 V1 Instance 2 V1 Instance 3 V1 Most users - 95% Some users - 5% Router
- 54. Canary Release Old Version Instance 1 V2 Instance 2 V2 Instance 3 V2 New Version Instance 1 V1 Instance 2 V1 Instance 3 V1 0% All users - 100% Router
- 55. Blue-Green Old Group Instance 1 V1 Instance 2 V1 Instance 3 V1 DNS service-1.domain.com Router
- 56. Blue-Green Old Group Instance 1 V2 Instance 2 V2 Instance 3 V2 New Group Instance 1 V1 Instance 2 V1 Instance 3 V1 DNS service-1.domain.com Router Router
- 57. Blue-Green Old Group Instance 1 V2 Instance 2 V2 Instance 3 V2 New Group Instance 1 V1 Instance 2 V1 Instance 3 V1 DNS service-1.domain.com Router Router
- 59. What to monitor? ● Hosts CPU, Memory, I/O, Network, File system ● Containers CPU, Memory, I/O, Restarts, Bottleneck ● Applications Latency, Traffic, Errors
- 60. Metrics & Alerts Service Instance JMX Prometheus Metrics database Container Docker Engine Service Instance JMX Container Grafana Visualize metrics cAdvisor Collect Containers metrics VM Node Exporter Collect OS metrics Scrape metrics AlertManager Send notifications Email Slack Pushover Query Push