Manage microservices like a Chef

Transcript

1. Manage microservices
   like a Chef
   

   View Slide

2. Mathieu Acthernoene
   @zoontek
   

   View Slide

3. Microservices are
   awesome ✨
   

   View Slide

4. Microservices are awesome ✨
   Each service is an app implementing its own
   business logic
   service
   database
   SaaS 1 SaaS 2
   

   View Slide

5. Microservices are awesome ✨
   Separated by concern, easier to understand
   what's going on
   auth email sms …
   

   View Slide

6. Microservices are awesome ✨
   They can evolve apart, deploy at different
   frequencies
   service 1 service 2 service 3 service 4
   service 1 service 4
   service 4
   

   View Slide

7. Microservices are awesome ✨
   If one breaks, you don't have to reboot everything
   service 1 service 3 service 4
   

   View Slide

8. Microservices are awesome ✨
   Scales according to the service needs, not globally
   CPU
   intensive
   service
   RAM
   intensive
   service
   GPU
   intensive
   service
   

   View Slide

9. Microservices are awesome ✨
   Each service team is free to use the right tools
   service
   using
   couchDB
   service
   using
   JavaScript
   service
   using
   Golang
   

   View Slide

10. But
    

    View Slide

11. But
    

    View Slide

12. Just as you think you're
    making microservices
    

    View Slide

13. You might just not
    

    View Slide

14. You are not building
    microservices when… #
    

    View Slide

15. You are not building
    microservices when… #
    Changes needs to be synced
    service 1 service 1
    change
    service 2
    

    View Slide

16. Deploys needs to be synced
    You are not building
    microservices when… #
    service 1 service 1
    deploy
    service 2
    

    View Slide

17. The same developers work across a large
    number of them
    service 1
    Kévin
    Sylvie
    Amine
    You are not building
    microservices when… #
    service 2
    Kévin
    Luc
    Clarisse
    service 3
    Kévin
    Zhang
    Nicolas
    

    View Slide

18. They share a datastore
    You are not building
    microservices when… #
    service 1
    database
    service 2
    

    View Slide

19. They share a lot of the same code / models
    You are not building
    microservices when… #
    service 1
    type User {}
    service 2
    type User {}
    service 3
    type User {}
    

    View Slide

20. You just created a distributed
    monolith
    crypto
    REST
    logging
    REST
    email
    REST
    sms
    REST
    storage
    REST
    billing
    REST
    

    View Slide

21. View Slide

22. Managing microservices is
    like managing a restaurant
    

    View Slide

23. waiter
    REST
    sous-chef
    REST
    pastry chef
    REST
    wine waiter
    REST
    washer
    REST
    chef
    REST
    

    View Slide

24. View Slide

25. How does a restaurant
    manage services?
    

    View Slide

26. View Slide

27. The “simple” gateway solution
    crypto
    REST
    logging
    REST
    email
    REST
    sms
    REST
    storage
    REST
    billing
    REST
    gateway
    REST
    

    View Slide

28. One interlocutor
    passing messages to
    the right services
    

    View Slide

29. The REST / JSON gateway issues ⚠
    If a first API is RESTful, the twelfth iteration
    is RESTish
    service API v1
    ✨
    service API v3
    
    … service API v9
    
    …
    

    View Slide

30. Latency caused by serialization / deserialization
    service 1 service 2
    API gateway
    decode
    JSON
    understand
    message
    create
    message
    encode
    JSON
    The REST / JSON gateway issues ⚠
    

    View Slide

31. You embed your model schemas in every messages
    {
    "login": "octocat",
    "avatar_url": "https://github.com/images/error/octocat_happy.gif",
    "gravatar_id": "",
    "url": "https://api.github.com/users/octocat",
    "html_url": "https://github.com/octocat",
    …
    The REST / JSON gateway issues ⚠
    

    View Slide

32. As a service owner, you have to wait for a complete
    implementation of another one to start using it
    service 1
    is finished &
    deployed
    the API
    gateway
    implements
    service 1
    service 2 starts
    implementing
    its usage of
    service 1
    The REST / JSON gateway issues ⚠
    

    View Slide

33. As JSON types are basic (JavaScript interop…), a
    documentation is mandatory
    The REST / JSON gateway issues ⚠
    

    View Slide

34. How does a restaurant
    manage communication?
    

    View Slide

35. View Slide

36. People working in the
    kitchen don't
    necessarily have the
    same mother tongue
    

    View Slide

37. They don't force their
    coworkers to use their
    language
    

    View Slide

38. Protobuf ↔
    

    View Slide

39. Protobuf ↔
    Created by, and for Google in 2001
    

    View Slide

40. A binary format for typed exchanges
    Protobuf ↔
    

    View Slide

41. Protobuf ↔
    Generate clients for C++, Java, Python, Go,
    Ruby, C#, JS, ObjC, PHP, Dart…
    

    View Slide

42. Offers backward compatibility for free
    Protobuf ↔
    

    View Slide

43. Offers a JSON mapping
    Protobuf ↔
    

    View Slide

44. Tooling and linters are available
    Protobuf ↔
    

    View Slide

45. Allows RPC methods definition
    Protobuf ↔
    

    View Slide

46. How to write a .proto file
    

    View Slide

47. syntax = "proto3";
    

    View Slide

48. package storage;
    

    View Slide

49. import "google/protobuf/timestamp.proto";
    import "google/api/annotations.proto";
    

    View Slide

50. option java_package = "com.scaleway.storage";
    

    View Slide

51. message Bucket {
    enum Attributes {
    LAZY = 0;
    DELUXE = 1;
    }
    string id = 1;
    google.protobuf.Timestamp created_at = 2;
    repeated Attributes attributes = 3;
    int32 old_field = 4 [deprecated=true];
    }
    

    View Slide

52. service Storage {
    rpc ListBuckets(ListBucketsRequest) returns (stream Bucket) {
    option (google.api.http) = { get: "/buckets" };
    }
    rpc GetBucket(GetBucketRequest) returns (Bucket) {
    option (google.api.http) = { get: "/buckets/{id}" };
    }
    }
    message ListBucketsRequest { string user_id = 1; }
    message GetBucketRequest { string id = 1; }
    

    View Slide

53. syntax = "proto3";
    package storage;
    option java_package = "com.scaleway.storage";
    service Storage {
    rpc ListBuckets(ListBucketsRequest) returns (stream Bucket) {
    option (google.api.http) = { get: "/buckets" };
    }
    rpc GetBucket(GetBucketRequest) returns (Bucket) {
    option (google.api.http) = { get: "/buckets/{id}" };
    }
    }
    message Bucket {
    enum Attributes {
    LAZY = 0;
    DELUXE = 1;
    }
    string id = 1;
    google.protobuf.Timestamp created_at = 2;
    repeated Attributes attributes = 3;
    int32 old_field = 4 [deprecated=true];
    }
    message ListBucketsRequest { string user_id = 1; }
    message GetBucketRequest { string id = 1; }
    

    View Slide

54. View Slide

55. I get Protobuf, but what is gRPC ?
    HTTP/2 clients for Protobuf messages transport
    

    View Slide

56. Unaries, uni- and bi-directional streams
    I get Protobuf, but what is gRPC ?
    

    View Slide

57. Blocking & Non-Blocking messages
    I get Protobuf, but what is gRPC ?
    

    View Slide

58. Authentification (per call, and TLS for trusted
    services)
    I get Protobuf, but what is gRPC ?
    

    View Slide

59. Cancellation and timeout handling
    I get Protobuf, but what is gRPC ?
    

    View Slide

60. Smart retries with backoff
    I get Protobuf, but what is gRPC ?
    

    View Slide

61. Error handling with standardized status codes
    I get Protobuf, but what is gRPC ?
    

    View Slide

62. Lameducking for graceful shutdowns
    I get Protobuf, but what is gRPC ?
    

    View Slide

63. A huge ecosystem
    

    View Slide

64. A huge ecosystem
    Used by Google, Lyft, Uber, Square, Netflix, CoreOS,
    Docker, Cockroachdb, Cisco, Juniper Networks…
    

    View Slide

65. A huge ecosystem
    Supported by nginx, kubernetes, envoy, etcd, istio…
    

    View Slide

66. A huge ecosystem
    A lot of middlewares available
    

    View Slide

67. A huge ecosystem
    A HTTP+JSON interface available
    

    View Slide

68. A new service is created
    protos
    monorepo
    crypto
    gRPC
    logging
    gRPC
    email
    gRPC
    sms
    gRPC
    storage
    gRPC
    billing
    gRPC
    service
    discovery
    ML
    gRPC
    

    View Slide

69. And for the front-end?
    

    View Slide

70. How does a restaurant
    expose services to a client?
    

    View Slide

71. View Slide

72. GraphQL
    

    View Slide

73. A request example
    (developer.github.com/v4/explorer)
    

    View Slide

74. dashboard
    CLI
    API gateway
    mobile app
    GraphQL
    adapter
    protos
    monorepo
    crypto
    gRPC
    gRPC
    logging
    gRPC
    email
    gRPC
    sms
    gRPC
    storage
    gRPC
    billing
    gRPC
    service
    discovery
    gRPC REST
    

    View Slide

75. View Slide

76. Keep in mind that there's
    no silver bullet
    

    View Slide

77. This technologies allow
    you to create a factory of
    products / services
    

    View Slide

78. Thank you!
    Questions?
    

    View Slide