Kafka Will Get The Message Across, Guaranteed.

Transcript

1. 1.

   KAFKA WILL GET THE MESSAGE ACROSS. GUARANTEED. SymfonyCon 2016 Berlin,

   Germany
2. 2.

   David Zuelke

3. 3.

   None
4. 4.

   dz@heroku.com

5. 5.

   @dzuelke

6. 6.

   KAFKA

7. 7.

   LinkedIn

8. 8.

   APACHE KAFKA

9. 9.

   "uh oh, another Apache project?!"

10. 10.

    None
11. 11.

    KEEP CALM AND LOOK AT THE WEBSITE

12. 12.

    None
13. 13.

    "Basically it is a massively scalable pub/sub message queue. architected

    as a distributed transaction log."
14. 14.

    "so it's a queue?"

15. 15.

    it's not a queue

16. 16.

    queues are not multi-subscriber :(

17. 17.

    "so it's a pubsub thing?"

18. 18.

    it's not a pubsub thing

19. 19.

    pubsub broadcasts to all subscribers :(

20. 20.

    it's a log

21. 21.

    None
22. 22.

    not that kind of log

23. 23.

    WAL

24. 24.

    Write-Ahead Log

25. 25.

    WRITE-AHEAD LOG

26. 26.

    None
27. 27.

    1 foo 2 bar 3 baz 4 hi

28. 28.

    1 create document: "foo", data: "…" 2 update document: "foo",

    data: "…" 3 create document: "bar", data: "…" 4 remove document: "foo"
29. 29.

    None
30. 30.

    never corrupts

31. 31.

    sequential I/O

32. 32.

    None
33. 33.

    sequential I/O

34. 34.

    every message will be read at least once, no random

    access
35. 35.

    FileChannel.transferTo (shovels data straight from e.g. disk cache to network

    interface, no copying via RAM)
36. 36.

    "HI, I AM KAFKA" "Buckle up while we process (m|b|tr)illions

    of messages/s."
37. 37.

    TOPICS

38. 38.

    streams of records

39. 39.

    1 2 3 4 5 6 7 …

40. 40.

    1 2 3 4 5 6 7 8 … producer

    writes consumer reads
41. 41.

    can have many subscribers

42. 42.

    1 2 3 4 5 6 7 8 … producer

    writes consumerB reads consumerA reads
43. 43.

    can be partitioned

44. 44.

    P0 1 2 3 4 5 6 7 … P1

    1 2 3 4 … P2 1 2 3 4 5 6 7 8 … P3 1 2 3 4 5 6 …
45. 45.

    partitions let you scale storage!

46. 46.

    partitions let you scale consuming!

47. 47.

    None
48. 48.

    all records are retained, whether consumed or not, up to

    a configurable limit
49. 49.

    PRODUCERS

50. 50.

    byte[]

51. 51.

    (typically JSON, XML, Avro, Thrift, Protobufs)

52. 52.

    (typically not funny GIFs)

53. 53.

    can choose explicit partition, or a key (which is used

    for auto-partitioning)
54. 54.

    https://github.com/edenhill/librdkafka & https://arnaud-lb.github.io/php-rdkafka/

55. 55.

    BASIC PRODUCER $rk = new RdKafka\Producer(); $rk->addBrokers("127.0.0.1"); $topic = $rk->newTopic("test");

    $topic->produce(RD_KAFKA_PARTITION_UA, 0, "Unassigned partition, let Kafka choose"); $topic->produce(RD_KAFKA_PARTITION_UA, 0, "Yay consistent hashing", $user->getId()); $topic->produce(1, 0, "This will always be sent to partition 1");
56. 56.

    CONSUMERS

57. 57.

    cheap

58. 58.

    only metadata stored per consumer: offset

59. 59.

    guaranteed to always have messages in right order (within a

    partition)
60. 60.

    can themselves produce new messages!

61. 61.

    None
62. 62.

    BASIC CONSUMER $conf = new RdKafka\Conf(); $conf->set('group.id', 'myConsumerGroup'); $rk =

    new RdKafka\Consumer($conf); $rk->addBrokers("127.0.0.1"); $topicConf = new RdKafka\TopicConf(); $topicConf->set('auto.commit.interval.ms', 100); $topic = $rk->newTopic("test", $topicConf); $topic->consumeStart(0, RD_KAFKA_OFFSET_STORED); while (true) { $msg = $topic->consume(0, 120*10000); do_something($msg); }
63. 63.

    AT-MOST ONCE DELIVERY $conf = new RdKafka\Conf(); $conf->set('group.id', 'myConsumerGroup'); $rk

    = new RdKafka\Consumer($conf); $rk->addBrokers("127.0.0.1"); $topicConf = new RdKafka\TopicConf(); $topicConf->set('auto.commit.enable', false); $topic = $rk->newTopic("test", $topicConf); $topic->consumeStart(0, RD_KAFKA_OFFSET_STORED); while (true) { $msg = $topic->consume(0, 120*10000); $topic->offsetStore($msg->partition, $msg->offset); do_something($msg); }
64. 64.

    AT-LEAST ONCE DELIVERY $conf = new RdKafka\Conf(); $conf->set('group.id', 'myConsumerGroup'); $rk

    = new RdKafka\Consumer($conf); $rk->addBrokers("127.0.0.1"); $topicConf = new RdKafka\TopicConf(); $topicConf->set('auto.commit.enable', false); $topic = $rk->newTopic("test", $topicConf); $topic->consumeStart(0, RD_KAFKA_OFFSET_STORED); while (true) { $msg = $topic->consume(0, 120*10000); do_something($msg); $topic->offsetStore($msg->partition, $msg->offset); }
65. 65.

    EXACTLY-ONCE DELIVERY

66. 66.

    you cannot have exactly-once delivery

67. 67.

    THE BYZANTINE GENERALS "together we can beat the monsters. let's

    both attack at 07:00?" "confirm, we attack at 07:00" ☠
68. 68.

    USE CASES

69. 69.

    • LinkedIn • Yahoo • Twitter • Netflix • Square

    • Spotify • Pinterest • Uber • Goldman Sachs • Tumblr • PayPal • Airbnb • Mozilla • Cisco • Etsy • Foursquare • Shopify • CloudFlare
70. 70.

    ingest Twitter firehose and turn it into a pointless demo

    ;)
71. 71.

    None
72. 72.

    messaging, of course

73. 73.

    track user activity

74. 74.

    record runtime metrics

75. 75.

    IoT

76. 76.

    replicate information between data centers

77. 77.

    billing!

78. 78.

    "shock absorber" between systems to avoid overload of DBs, APIs,

    etc.
79. 79.

    in PHP: mostly producing messages; better languages exist for consuming

80. 80.

    The End

81. 81.

    THANK YOU FOR LISTENING! Questions? Ask me: @dzuelke & dz@heroku.com