1. 1
Microservices.
Test smarter, not harder.
Beth Skurrie (DiUS)
@bethesque
#VoxxedAus2019

2. A big problem

3. Problem:
Long time to market

4. Microservices!

5. Solved a problem!
New problem...

6. WHAT WE ARE KNOWN FOR
6
Microservices
Solved problems New problems
● E2E integration tests
○ Slow tests
○ Easy to break
○ Hard to fix
○ Scale badly
○ Lots of set up
○ Flakey tests ignored
○ Takes dev time away
from features
● Feature time to market
A STORY @bethesque

7. 7
@bethesque
To E2E test, or not to E2E test?
A STORY

8. 8
@bethesque
What if there was another way?

9. 9
C P
mock
ANOTHER WAY @bethesque
Consumer tests Provider tests
Integration test

10. WHAT WE ARE KNOWN FOR
10
Test symmetry
Solved problems New problems
● Hard to keep both sides in
sync
● Fast feedback
● Few dependencies
● No dedicated environment
● Reliable
● Easy to debug
ANOTHER WAY @bethesque

11. 11
C P
mock
ANOTHER WAY @bethesque

12. 12
C P
mock
ANOTHER WAY @bethesque

13. PPCC
mock
contract
ANOTHER WAY @bethesque
Consumer tests Provider tests
Integration test

14. 14
Message Contract
OTHER TOOLS
C P
contract
@bethesque

15. WHAT WE ARE KNOWN FOR
15
Contracts
Solved problems New problems
● ???● Keeping tests in sync
ANOTHER WAY @bethesque

16. How long between writing the
code and finding the bug?

17. Local
tests
Isolated CI
tests
Deploy E2E
tests
5 mins 15 mins 45 mins 1 hour +
Find
bug
here!
Not
here!
@bethesque

18. Find integration bugs
early and quickly
Contracts tests
ANOTHER WAY @bethesque

19. Know before you
commit
Contracts tests
ANOTHER WAY @bethesque

20. Make changes with
speed and confidence
Contracts tests
ANOTHER WAY @bethesque

21. Deploy independently
Contracts tests
ANOTHER WAY @bethesque

22. Better API design
(Consumer)
Contracts tests
ANOTHER WAY @bethesque

23. Are not functional tests
Contracts tests
ANOTHER WAY @bethesque

24. Are not an API
specification
Contracts tests
ANOTHER WAY @bethesque

25. Are not good for “public”
APIs (many, unknown
consumers)
(Consumer)
Contracts tests
ANOTHER WAY @bethesque

26. Are not a silver bullet!
Contracts tests
ANOTHER WAY @bethesque

27. 27
@bethesque
Business value
Code correctness

28. 28
@bethesque
The Testing Pyramid of Hell

29. 29
@bethesque

30. WHAT WE ARE KNOWN FOR
30
Speed up your releases
Do less Do more
● Contract testing
● Aggregated logging
● Metrics
● Semantic monitoring
● Alerting
● Correlation IDs
● Integration testing
ANOTHER WAY @bethesque

31. My contract
testing journey

32. 32
@bethesque
● Open source
● Multiple languages
○ JVM
○ .NET
○ Javascript
○ Python
○ Go
○ + more
pact.io
● HTTP contracts
● Message contracts

33. 33
A B
mock
C P
contract
A CONTRACT TESTING JOURNEY @bethesque

34. 34
A CONTRACT TESTING JOURNEY @bethesque

35. 35
@bethesque
Too many microservices?
You need another microservice!
A CONTRACT TESTING JOURNEY

36. 36
Pact Broker
A CONTRACT TESTING JOURNEY @bethesque

37. 37
@bethesque
Problem 1: Contract exchange
Solution: Pact Broker
A CONTRACT TESTING JOURNEY

38. 38
@bethesque
WHEN
the provider receives
<some request>
THEN
it will return
<some response>
A CONTRACT TESTING JOURNEY

39. 39
@bethesque
WHEN
the provider receives
a GET request for /alligators/Mary
THEN
it will return
a 200 OK response
with a JSON body {“name”: “Mary”}
A CONTRACT TESTING JOURNEY

40. 40
@bethesque
WHEN
the provider receives
a GET request for /alligators/Mary
THEN
it will return
a 404 Not Found
response
a 200 OK
response
A CONTRACT TESTING JOURNEY

41. 41
@bethesque
GIVEN
<the provider is in a certain state>
WHEN
the provider receives
<some request>
THEN
it will return
<some response>
A CONTRACT TESTING JOURNEY

42. 42
@bethesque
Problem 2: Data setup, code coverage
Solution: Provider states
A CONTRACT TESTING JOURNEY

43. A contract testing
journey
● Automate the contract exchange
● You still need to think about test
data
A CONTRACT TESTING JOURNEY @bethesque

44. 44
@bethesque
Problem 3: brittle tests
A CONTRACT TESTING JOURNEY

45. 45
@bethesque
Problem 3: brittle tests
Solution: flexible matching
A CONTRACT TESTING JOURNEY

46. A contract testing
journey
● Automate the contract exchange
● You still need to think about test data
● Contracts should focus on the
messages, not the technology
● Contracts should be as flexible as
possible - but no more
A CONTRACT TESTING JOURNEY @bethesque

47. 47
@bethesque
Problem 4: Dealing with
contract changes
A CONTRACT TESTING JOURNEY

48. 48
@bethesque
The other service needs to know when a
contract has changed
Contracts are not a substitute for good
communication between teams
A CONTRACT TESTING JOURNEY

49. 49
@bethesque
Pact Broker webhooks
A CONTRACT TESTING JOURNEY

50. 50
@bethesque
Contracts are STILL not a substitute for
good communication between teams
A CONTRACT TESTING JOURNEY

51. ● Automate the contract exchange
● You still need to think about test data
● Contracts should focus on the
messages, not the technology
● Contracts should be as flexible as
possible - but no more
● The provider needs to know when a
contract has changed
● Remember: contracts are not a
substitute for good communication
between teams
A contract testing
journey
A CONTRACT TESTING JOURNEY @bethesque

52. 52
A CONTRACT TESTING JOURNEY @bethesque

53. 53
@bethesque
Problem 5: Communicating
verification results back to
consumer
A CONTRACT TESTING JOURNEY

54. 54
@bethesque
Pact Broker verifications
A CONTRACT TESTING JOURNEY

55. 55
@bethesqueA CONTRACT TESTING JOURNEY

56. 56
@bethesqueA CONTRACT TESTING JOURNEY

57. 57
@bethesque
If you can’t deploy your services
independently,
you don’t have microservices.
You have a distributed monolith.
A CONTRACT TESTING JOURNEY
Contract
tests
E2E
tests

58. 58
@bethesque
If you can’t deploy your services
independently,
you don’t have microservices.
You have a distributed monolith.
A CONTRACT TESTING JOURNEY

59. 59
@bethesque
If you can’t deploy your services
independently,
you don’t have microservices.
You have a distributed monolith.
A CONTRACT TESTING JOURNEY
Warning!
Do not build a
distributed monolith

60. 60
Consumer
version
Provider
version
Verification
result
11 54 success
12 54 failure
12 55 success
13 56 success
13 57 failure
“The Matrix”
A CONTRACT TESTING JOURNEY @bethesque

61. 61
Consumer
version
Provider
version
Verification
result
11 ✓ 54 PROD. success
12 54 failure
12 55 success
13 56 success
13 57 unknown
Can I deploy?
A CONTRACT TESTING JOURNEY @bethesque

62. 62
A CONTRACT TESTING JOURNEY @bethesque

63. 63
A CONTRACT TESTING JOURNEY @bethesque

64. 64
@bethesque
What about Swagger/OAS?
OTHER TOOLS

65. 65
Provider contracts
OTHER TOOLS @bethesque

66. 66
Consumer contracts
OTHER TOOLS @bethesque

67. 67
A B
mock/write
C P
pact swagger
verify
verify
Pact+Swagger
OTHER TOOLS @bethesque

68. ● OAS support
● Improved Broker workflow
What are the next
problems to
solve for
Pact?
ANOTHER WAY @bethesque

69. WHAT WE ARE KNOWN FOR
69
Contracts
Solved problems New problems
● ???● Shipping code faster
@bethesque

70. 70
pact.io
pactflow.io
slack.pact.io
@pact_up
Microservices.
Test smarter, not harder
Beth Skurrie (DiUS)
@bethesque
#VoxxedAus2019