Surrounded by Microservices

The slides for the 'Surrounded by Microservices' talk , Euruko 2019, Rotterdam.

Transcript

1. Surrounded by Microservices
2. The DVD Rental Company
3. Video on demand
4. Original content
5. Big Shift?
6. Old industry !
7. Remember these?
8. Problem space !
9. Studio apps
10. Starting with a monolith
11. Specialization !
12. Starting fresh !
13. Ruby On Rails !
14. The Active Record Pattern ! Domain Objects " Business Rules # Validations $ Persistence
15. Distributed data !
16. Various data sources ! REST API " JSON API # GRPC $ GraphQL % Local Database
17. "Data that lives in our database today might live in a service tomorrow"
18. Architecture that clearly separates the business logic from any kind of implementation details and protocols
19. Hexagonal Architecture
21. Core concepts
22. Entities
23. The repositories
24. The data sources
25. Interactors
26. Implementing business logic
27. Entities module Workflows class ProductionEntity < BaseEntity attribute :id, AttrType::Strict::Integer attribute :title, AttrType::Strict::String attribute :logline, AttrType::Strict::String.optional end end
28. dry-struct
29. Entities class ProductionEntity < BaseEntity attribute :id, AttrType::Strict::Integer attribute :title, AttrType::Strict::String attribute :logline, AttrType::Strict::String.optional end
30. Repositories (Custom DSL) class ProductionRepo < BaseRepo entity_class ProductionEntity def initialize(data_source: MovieProduction) super end def find_by_id(id) wrap(data_source.find_by_id(id)) end def search_by_title(title) wrap(data_source.search_by_title(title)) end end
31. Data Sources class MovieProduction < ApplicationRecord def self.search_by_title(title) where('title LIKE ?', "#{sanitize_sql_like(title)}%") } end end
32. Data Sources class MovieProductionAPI def self.search_by_title(title) SwaggerClient::MovieProductionsApi.advanced_search( search_param: title, headers: { ... }, ... ) end end
33. Data Source Independent # using the default data source repo = ProductionRepo.new # swapping for a REST API data source repo = ProductionRepo.new(data_source: MovieProductionAPI)
34. Interactor gem
35. Interactors module Workflows class OnboardProduction include NetflixInteractor def self.call(production_id:, vertical_id:, repo: ProductionRepo.new) super end def call # validate inputs # is it okay to onboard a production? # cover edge cases # notify others about a production being onboarded end end end
36. Hanami
37. Implementing data sources
38. API Clients
39. Handwritten API Clients
40. Autogenerated API Clients
41. Autogenerated API Clients Swagger Code Generator GRPC JSON API Spec
42. Dealing with errors
43. The network is unreliable
44. Graceful failure
45. Logging downstream requests
46. Metrics
47. Error reporting
48. Errors Must Be Actionable
49. Avoid Alarm Fatigue
50. Setting up thresholds
51. Going straight to the data source
52. Does it scale?
53. Eventual consistency
54. Consistency vs availability
55. "Services should respond in a timely manner"
56. What's in the future?
57. Delaying decisions
58. Project paradox
59. The purpose of a good architecture is to delay decisions. Why? Because when we delay a decision, we have more information when it comes time to make it. - Uncle Bob
60. Postgres? Cassandra? Redis? Elasticsearch?
61. Testing Strategy
62. The test suite must be reliable and fast
63. Rails Testing
64. Testing business logic
65. Dependency Injection
66. Testing with verified doubles describe Workflows::OnboardProduction do let(:production_id) { SecureRandom.uuid } let(:vertical_id) { SecureRandom.uuid } let(:response) do described_class.call(production_id: production_id, vertical_id: vertical_id, repo: repo) end describe 'when a production already exists' do let(:repo) { instance_double(Workflow::ProductionRepo, find_by_id: existing_production) } it 'the response is a failure' do expect(response).to be_a_failure end end ... end
67. Integration specs data sources & end to end
68. A fast and reliable spec suite Finished in 1 minute 2.35 seconds (files took 2.93 seconds to load) 2015 examples, 0 failures
69. Conclusion
70. Separate business logic from implementation details
71. Delay decisions
72. Damir Svrtan Twitter: @DamirSvrtan