Effective Testing with API Simulation and (Micro)Service Virtualisation

Transcription

1 Effective Testing with API Simulation and (Micro)Service Virtualisation Andrew Morgan Daniel 1

2 tl;dr Testing distributed systems is hard Many of us building distributed systems (microservices) API simulation is a new (old) tool to help with testing 2

3 About Us Andrew Daniel Independent Consultant Microservices / Cloud / Continuous Delivery InfoQ Editor Consultant/CTO SpectoLabs Cloud, containers, and microservices O Reilly book on the way! 3

4

5 Agenda The problems with testing API driven architectures Introduction to API simulation Applying API simulation patterns Data management Conclusion 5

6 Agenda The problems with testing API driven architectures Introduction to API simulation Applying API simulation patterns Data management Conclusion 6

7 API-Driven Architecture Payment Provider Social Network API Call API Call Application 7

8 Microservices Company Flights Payment Auth Booking 8

9 A natural evolution... 9

10 Challenges 10

11 Airline Booking Application Microservice architecture Integration with third party services Integration with legacy internal services Heavy resource consumption A standard CI/CD pipeline 11

12 Resources Rewards Flights Payments Cost Booking Fraud 12

13 13

14 Licensing Third party Cost Producer Producer Consumer Consumer 14

15 Rate Limiting Producer Rate Limited Producer Consumer Consumer 15

16 Flakey Dependencies Producer Producer Consumer Consumer 16

17 Doesn't exist? Consumer 17

18 Non Deterministic Booking Service { } date : , , Consumer assertthat(booking.getdate()).isequalto(localdate.of(2013,1,8)); 18

19 Fault Injection 4xx, 5xx, timeouts Producer Producer Operationally challenging to trigger faults Consumer Consumer 19

20 Agenda The problems with testing API driven architectures Introduction to API simulation Applying API simulation patterns Data management Conclusion 20

21 API Simulation Emulation of real API Non-intrusive Lightweight - or should be 21

22 Open Source Tooling 22

23 Architecture Simulate Capture Mode Consumer Producer Intercept Replay recorded and store requests & responses 23

24 Creating Data 24

25 Creating Data 25

26 Forward Proxy Consumer Producer Proxy Setting 26

27 Reverse Proxy Consumer Producer Points to Hoverfly Reverse Proxy Setting 27

28 Not Mocking Service Under Test Mocking Http Mocks bleed into the architectural boundary of your application Dependant Service Virtual Service 28

29 Agenda The problems with testing API driven architectures Introduction to API simulation Applying API simulation patterns Data management Conclusion 29

30 Record and replay Consumer External Producer Intercept & Simulate 30

31 JUnit public HoverflyRule hoverflyrule = HoverflyRule.inCaptureMode( src/test/ resources/external.json public void shouldbeabletoretrieveabooking() throws URISyntaxException { // Given final int bookingid = 1; // When final Booking booking = bookingservicegateway.getbooking(bookingid); // Then assertthat(booking.getcustomer()).isequalto( Andrew Morgan ); assertthat(booking.getdate()).isequalto(localdate.of(2016,3,1)); } 31

32 JUnit Example { "data": [ { "request": { "path": "/api/bookings/1", "method": "GET", "destination": " "query": null, "body": "" }, "response": { "status": 200, "body": {\ customer\ :\ Andrew Morgan\", \ date\ : \ \ }}}, "headers": { "Content-Type": [ "application/json" ] } } } ] } 32

33 JUnit public HoverflyRule hoverflyrule = HoverflyRule.inSimulationMode(classpath( external.json public void shouldbeabletoretrieveabooking() throws URISyntaxException { // Given final int bookingid = 1; // When final Booking booking = bookingservicegateway.getbooking(bookingid); // Then assertthat(booking.getcustomer()).isequalto( Andrew Morgan ); assertthat(booking.getdate()).isequalto(localdate.of(2016,3,1)); } 33

34 Advantages Isolation Flaky dependencies Rate limiting Deterministic Licensing 34

35 Operations Internal Virtual Producer Service Virtual Internal Service Producer Consumer Virtual Service External Producer Virtual Internal Service Producer 35

36 CLI Example $ hoverctl start proxy-port 8080 $ hoverctl import accounts-service-simulation.json $ hoverctl start proxy-port 8081 $ hoverctl import payments-service-simulation.json $ curl proxy localhost:8080 [{"name":"foo"}] $ curl -proxy localhost:8081 [{ amount":" 23.27"}] 36

37 37

38 Matching Matchers Method Destination Path Query GET /miss?token=foo GET /login?token=foo Request GET /login?token=foo 38

39 Why Looser Matching? Matcher Method Destination Path Query GET /login?token=foo Request GET /login?token=miss?token=foo??? 39

40 Field Omission Matcher Method Destination Path Query GET /login?token=foo Request GET /login?token=bar 40

41 Regex Matching Matcher Method Destination Path Path Regex GET name\\/(.*) /name/john Request GET /name/tim 41

42 JSON Matching Matcher Body Body JSON { foo : bar } Request { foo : bar } 42

43 JSON Matching Matcher Body JSON Body JSON loose=true { foo : bar, not : important } Request { foo : bar } 43

44 Additional Matchers XML (Strict and Loose) JSONPath / XPath Anything 44

45 Scoring Request GET /name/tim?token=bar Matchers Match? Score GET 2 GET /name/tim 3 GET /name/tim?token=bar 4 45

46 Debugging Request GET /name/tim?token=bar Matchers Score Match? Closest? MISS miss 1 MISS /name/tim 2 GET /name/tim miss 3 46

47 Templating Path Matcher Response Body /name\\/(.*) { name : {{ { name : john } Request.Path.[1] }} } Request Path Response Body /name/john { name : john } /name/peter { name : peter } { name : john } 47

48 The Problem of State Request Response GET basket {} POST basket { item : eggs } 200 GET basket {} 48

49 State Request Response Requires State Transitions State GET basket {} basket == empty POST basket { item : eggs } 200 basket = eggs GET basket { item : eggs } basket == eggs 49

50 SSL Bump Generate Fake Self Signed Certificate SSL SSL Again Consumer Producer Needs to Decrypt MITM Attack & Impersonate 50

51 Fault Injection Consumer External Producer Middleware 51

52 Fault Injection 52

53 Agenda The problems with testing API driven architectures Introduction to API simulation Applying API simulation patterns Data management Conclusion 53

54 Stale Data Consumer Virtual Real Service Service 54

55 Ownership Own Consumer Own Producer Strategy Integration Contract Testing 55

56 Integration Contract Testing Integration Testing or Contract Diff Diff? Scheduled Build Reproduce Simulations Diff No diff? 56

57 Ownership Own Consumer Own Producer Strategy Integration Contract Testing Simulation as First Class Citizen 57

58 First Class Simulations Parse Contract Run and Capture Service Tests Transform to simulation Publish Virtual Service Parse Documentati on 58

59 Ownership Own Consumer Own Producer Strategy Integration Contract Testing Virtual Service as First Class Citizen Consumer Driven Contract Testing 59

60 Consumer Driven Contracts Consumer Defines Contract Generate Producer Stubs Implement Consumer Generate Producer Contract Tests Implement Producer Run Consumer Contract Tests 60

61 Consumer Driven Contracts Change producer Run Consumer Contract Tests 61

62 Tools 62

63 Testing Pyramid End to end Consumer Driven Contract Testing Integration / Service Testing Unit Testing 63

64 Agenda The problems with testing API driven architectures Introduction to API simulation Applying API simulation patterns Data management Conclusion 64

65 Conclusion Reduced cost be replacing real resources Improved deployment time with API simulations Less dependance on end-to-end testing Advanced patterns for simplifying API simulations Improved feedback cycle 65

66 Questions? hoverfly.io 66