Reactive Data Access with RxJava ... and N1QL!

I gave this talk at Couchbase Connect 2015 in Santa Clara. Check out the couchbase website for the recording!

Transcript

1. REACTIVE  DATA  ACCESS  WITH  RXJAVA  
   …  INCLUDING  N1QL!
   Michael  Nitschinger,  Couchbase
   @daschl
   

   View Slide

2. ©2015  Couchbase  Inc.   2  
   §  New  challenges  
   §  React  to  user  load  
   §  React  to  failure  
   §  Be  responsive  under  load  and  failure  
   §  Need  new  solutions  
   §  Decoupled,  event-­‐driven  architectures  
   §  Optimal  resource  utilization  
   Why  Reactive?  
   2  
   

   View Slide

3. ©2015  Couchbase  Inc.   3  
   Why  Reactive?  
   3  
   

   View Slide

4. ©2015  Couchbase  Inc.   4  
   Detour:  Async  Java  Client  Stack  
   4  
   

   View Slide

5. ©2015  Couchbase  Inc.   5  
    
   Detour:  Smart  Batching  
   Source:  http://mechanical-­‐sympathy.blogspot.co.at/2011/10/smart-­‐batching.html  
   5  
   

   View Slide

6. ©2015  Couchbase  Inc.   6  
    
   Detour:  Smart  Batching  
   Source:  http://mechanical-­‐sympathy.blogspot.co.at/2011/10/smart-­‐batching.html  
   6  
   

   View Slide

7. RxJava  101  
   A  Gentle  Introduction  
   

   View Slide

8. ©2015  Couchbase  Inc.   8  
   §  Java  implementation  for  Reactive  Extensions    
   https://github.com/ReactiveX  
   §  A  library  to  compose  asynchronous  and  event-­‐driven  
   programs  through  observable  sequences.  
   RxJava:  Introduction  
   single   multiple  
   sync   T   Iterable  
   async   Future   Observable  
   8  
   

   View Slide

9. ©2015  Couchbase  Inc.   9  
   §  Observables  are  the  duals  of  Iterables  
   §  They  describe  both  Latency  and  Error  side  effects.  
   RxJava:  Introduction  
   event   Iterable  (pull)   Observable  (push)  
   data  retrieval   T  next()   onNext(T)  
   error  discovery   throws  Exception   onError(Exception)  
   completion   returns   onCompleted()  
   9  
   

   View Slide

10. ©2015  Couchbase  Inc.   10  
    Consuming  Observables  
    10  
    §  The  Observer  subscribes    
    and  receives  events.  
    §  A  cold  Observable  
    starts  when  subscribed.  
    §  onNext  can  be  called  
    0..N  times  
    

    View Slide

11. ©2015  Couchbase  Inc.   11  
    RxJava:  Creating  Observables  
    just
    11  
    

    View Slide

12. ©2015  Couchbase  Inc.   12  
    RxJava:  Creating  Observables  
    12  
    

    View Slide

13. ©2015  Couchbase  Inc.   13  
    RxJava:  Creating  Observables  
    13  
    

    View Slide

14. ©2015  Couchbase  Inc.   14  
    RxJava:  Creating  Observables  
    14  
    

    View Slide

15. ©2015  Couchbase  Inc.   15  
    RxJava:  Creating  Observables  
    15  
    

    View Slide

16. ©2015  Couchbase  Inc.   16  
    RxJava:  Creating  Observables  
    16  
    

    View Slide

17. ©2015  Couchbase  Inc.   17  
    RxJava:  Transforming  Observables  
    17  
    

    View Slide

18. ©2015  Couchbase  Inc.   18  
    RxJava:  Transforming  Observables  
    18  
    

    View Slide

19. ©2015  Couchbase  Inc.   19  
    RxJava:  Transforming  Observables  
    19  
    

    View Slide

20. ©2015  Couchbase  Inc.   20  
    RxJava:  Transforming  Observables  
    20  
    

    View Slide

21. ©2015  Couchbase  Inc.   21  
    RxJava:  Transforming  Observables  
    21  
    

    View Slide

22. ©2015  Couchbase  Inc.   22  
    RxJava:  Transforming  Observables  
    22  
    

    View Slide

23. ©2015  Couchbase  Inc.   23  
    RxJava:  Transforming  Observables  
    23  
    

    View Slide

24. ©2015  Couchbase  Inc.   24  
    RxJava:  Transforming  Observables  
    24  
    

    View Slide

25. ©2015  Couchbase  Inc.   25  
    RxJava:  Transforming  Observables  
    25  
    

    View Slide

26. ©2015  Couchbase  Inc.   26  
    RxJava:  Filtering  Observables  
    26  
    

    View Slide

27. ©2015  Couchbase  Inc.   27  
    RxJava:  Filtering  Observables  
    27  
    

    View Slide

28. ©2015  Couchbase  Inc.   28  
    RxJava:  Filtering  Observables  
    28  
    

    View Slide

29. ©2015  Couchbase  Inc.   29  
    RxJava:  Filtering  Observables  
    29  
    

    View Slide

30. Reactive  N1QL  
    Cranking  Queries  up  to  Eleven!  
    

    View Slide

31. ©2015  Couchbase  Inc.   31  
    § Simple  
    §  Executes  a  single  N1QL  statement,  no  options  available.  
    § Parameterized  
    §  Executes  a  parameterized  query  with  positional  or  named  params  
    § Prepared  
    §  Executed  a  previously  prepared  statement  
    Query  Types  
    31  
    

    View Slide

32. ©2015  Couchbase  Inc.   32  
    § (Async)  QueryResult  
    §  Observable  rows()  
    §  Observable  errors()  
    §  Observable  info()  
    §  Observable  signature()  
    §  Observable  finalSuccess()  
    §  boolean  parseSuccess()  
    §  String  requestId()  
    §  String  clientContextId()  
    Query  Response  
    32  
    

    View Slide

33. ©2015  Couchbase  Inc.   33  
    Simple  Query  
    33  
    

    View Slide

34. ©2015  Couchbase  Inc.   34  
    Parameterized  Query  
    34  
    § Named  Params  
    

    View Slide

35. ©2015  Couchbase  Inc.   35  
    Parametrized  Query  
    35  
    § Positional  Params  
    

    View Slide

36. ©2015  Couchbase  Inc.   36  
    Prepared  Query  
    36  
    

    View Slide

37. ©2015  Couchbase  Inc.   37  
    Index  Handling  
    37  
    

    View Slide

38. ©2015  Couchbase  Inc.   38  
    Index  Handling  
    38  
    

    View Slide

39. ©2015  Couchbase  Inc.   39  
    Conditional  Index  Creation  
    39  
    

    View Slide

40. Error  Handling  
    with  RxJava  
    

    View Slide

41. ©2015  Couchbase  Inc.   41  
    § Things  will  go  wrong,  so  better  plan  for  it  
    § Do  not  trust  integration  points  (including  the  SDK)  
    § Synchronous  retry  &  fallback  is  too  damn  hard  
     
    § RxJava  provides  (almost)  everything  you  need  to  
    §  fallback  
    §  retry  
    §  fail  fast  
    Preparing  to  Fail  
    41  
    

    View Slide

42. ©2015  Couchbase  Inc.   42  
    Timeouts  
    42  
    § The  network  is  unreliable  
    § Servers  fail  
    § The  SDK  contains  bugs  
    § Always  specify  timeouts  and  deal  with  them!  
    § The  synchronous  wrapper  defines  them  for  you  all  the  time.  
    

    View Slide

43. ©2015  Couchbase  Inc.   43  
    Timeouts:  Simple  
    43  
    

    View Slide

44. ©2015  Couchbase  Inc.   44  
    Timeouts:  Synchronous  API  
    44  
    

    View Slide

45. ©2015  Couchbase  Inc.   45  
    Timeouts:  Complex  Example  
    45  
    

    View Slide

46. ©2015  Couchbase  Inc.   46  
    Coordinated  Retry  
    46  
    §  Fail  fast  
    §  Don’t  let  your  system  get  stuck  
    §  Shed  load  with  backpressure  
    § Retry  
    §  immediately  won’t  help  
    §  either  linear  or  exponential  back-­‐off  necessary  
    §  have  a  strategy  if  retry  also  doesn’t  work  (Fallbacks!)    
    

    View Slide

47. ©2015  Couchbase  Inc.   47  
    Coordinated  Fallback  
    47  
    

    View Slide

48. ©2015  Couchbase  Inc.   48  
    Coordinated  Retry:  Builder  
    48  
    §  Declarative  API  instead  of  complicated  retryWhen  
    

    View Slide

49. Thank  you.  
    

    View Slide