Don't fight MongoDB

A brief introduction to MongoDB

Transcript

1. MongoDB «Don’t fight MongoDB» – @MirkoBonadei
2. Let’s start with some theory….
3. Brewer’sTheorem
4. Consistency Availability Partitioning
5. Consistency Availability Partitioning
6. «all nodes see the same data at the same time» Consistency
7. Consistency Availability Partitioning
8. «a guarantee that every request receives a response about whether it was successful or failed» Availability
9. Consistency Availability Partitioning
10. «the system continues to operate despite arbitrary message loss or failure of part of the system» Partitioning
11. You can choose only 2 out of 3
12. Which are the choices of MongoDB?
13. For Replica Sets MongoDB can be AP or CP
14. MongoDB buys PartitionTolerance For Replica Sets
15. Write Concerns AP CP
16. Write Concerns AP CP Unacknowledged (w=0)
17. Write Concerns AP CP Unacknowledged (w=0) Acknowledged (w=1) Default
18. Write Concerns AP CP Unacknowledged (w=0) Acknowledged (w=1) Default Journaled (w=1 & journal: true)
19. Write Concerns AP CP Unacknowledged (w=0) Acknowledged (w=1) Default Journaled (w=1 & journal: true) Replica Acknowledged (w > 1)
20. BUT…
21. Read Preferences CP Secondary Secondary Preferred Primary Preferred Default Primary Nearest
22. BE CAREFUL WITH REPLICA SETS
23. OpLog Primary OpLog Secondary OpLog
24. NETWORK PARTITION
25. OpLog Old Primary OpLog New Primary OpLog New Writes
26. NETWORK HEALING
27. OpLog New Primary OpLog Old Primary OpLog New Writes Rollback Synch
28. BASIC MongoDB
29. SCHEMA-LESS
30. MENTAL MODEL MATTERS (the Diaspora case study)
31. SLOW QUERIES …
32. Dear MongoDB, can you explain me why?
33. db.transfer_in.find({ "created_at": { $gt: ISODate("2013-11-25T11:00:00Z") , $lt: ISODate("2013-11-25T14:00:00Z") }, "business_model" : "subscription_activation", "operator" : "ES/Movistar", "seller_id": "4798119" }).explain();
34. { "cursor" : "BtreeCursor seller_id_1_transaction_id_1", "isMultiKey" : false, "n" : 44, "nscannedObjects" : 3235639, "nscanned" : 3235639, "nscannedObjectsAllPlans" : 6471278, "nscannedAllPlans" : 6471278, "scanAndOrder" : false, "indexOnly" : false, "nYields" : 16226, "nChunkSkips" : 0, "millis" : 38027, "indexBounds" : { ... }, "server" : " ….:27017" }
35. 3235639 scanned docs to find only 44 good documents!
36. Fight Full Scans!
37. db.transfer_in.ensureIndex( { created_at: 1, seller_id: 1, business_model: 1, operator: 1, status: 1 }, { background: true, name: “created_seller_bm_operator_status” } )
38. db.transfer_in.find({ "created_at": { $gt: ISODate("2013-11-25T11:00:00Z") , $lt: ISODate("2013-11-25T14:00:00Z") }, "business_model" : "subscription_activation", "operator" : "ES/Movistar", "seller_id": "4798119" }).explain();
39. { "cursor" : "BtreeCursor created_seller_businessmodel_operator_status", "isMultiKey" : false, "n" : 44, "nscannedObjects" : 44, "nscanned" : 13542, "nscannedObjectsAllPlans" : 27094, "nscannedAllPlans" : 40592, "scanAndOrder" : false, "indexOnly" : false, "nYields" : 178, "millis" : 219, "indexBounds" : { ... }, "server" : " ….:27017« }
40. Index Key Points
41. Order matters Background Ram eaters
42. Order matters Background Ram eaters
43. Order matters Background Ram eaters
44. Order matters Background Ram eaters
45. Aggregation Framework
46. Pipelines Friendly «Safe»
47. Pipelines Friendly «Safe»
48. Pipelines Friendly «Safe»
49. Pipelines Friendly «Safe»
50. Map/Reduce
51. Powerful Parallel Testable
52. Powerful Parallel Testable
53. Powerful Parallel Testable
54. Powerful Parallel Testable
55. REMEMBER THE FIRST RULE®
56. Resources • http://jandiandme.blogspot.de/2013/06/mongodb-and-cap-theorem.html • http://www.infoq.com/articles/cap-twelve-years-later-how-the-rules-have-changed • http://docs.mongodb.org/manual/ • http://edgystuff.tumblr.com/post/54709368492/how-to-speed-up-mongodb-map-reduce-by-20x • http://vimeo.com/21515035 (Gabriele Lana @ NoSQL Day 2011 – NoSQL Present, Past and Future) • http://vimeo.com/21586779 (Gabriele Lana @ NoSQL Day 2011 – MongoDB with Style) • http://www.sarahmei.com/blog/2013/11/11/why-you-should-never-use-mongodb/ (Sarah Mei wrong way of using MongoDB) • http://www.manning.com/banker/ (MongoDB in Action) • http://aphyr.com/posts/284-call-me-maybe-mongodb (Call me maybe: MongoDB – CP Analysis) • https://blog.serverdensity.com/mongodb-schema-design-pitfalls/ (MongoDB schema design pitfalls) • https://blog.serverdensity.com/does-everyone-hate-mongodb/ (Does everyone hate MongoDB) • http://www.allthingsdistributed.com/2007/12/eventually_consistent.html (AWS CTO) • https://github.com/gabrielelana/mongodb-with-style/tree/master/gol (GOL in Mongo by Gabriele Lana)