Tweet
Tweet

Slide 1

Slide 1 text

Raft Consensus for Rubyists

Slide 2

Slide 2 text

@vanstee Big Nerd Ranch

Slide 3

Slide 3 text

Coffee Snob Scenario

Slide 4

Slide 4 text

Matt Amy Kim Dan

Slide 5

Slide 5 text

Matt Amy Kim Dan

Slide 6

Slide 6 text

Matt Amy Kim Dan Want to meet for coffee?

Slide 7

Slide 7 text

Matt Amy Kim Dan Totally! I’m in. zzzzz…

Slide 8

Slide 8 text

Matt Amy Kim Dan Ok, let’s meet at Octane.

Slide 9

Slide 9 text

Matt Amy Kim Dan Sounds good. I’ll be there. …

Slide 10

Slide 10 text

Matt Amy Kim Dan Who wants coffee? *wakes up

Slide 11

Slide 11 text

Matt Amy Kim Dan Who wants coffee?

Slide 12

Slide 12 text

Matt Amy Kim Dan Oh, we’re going to Octane.

Slide 13

Slide 13 text

Delicious Coffee Consensus

Slide 14

Slide 14 text

What did we just do?

Slide 15

Slide 15 text

• Leader election • State replication • Partition tolerance

Slide 16

Slide 16 text

Now write an algorithm.

Slide 17

Slide 17 text

Humans > Computers > Congress

Slide 18

Slide 18 text

Why Ruby

Slide 19

Slide 19 text

Multiple db servers? Multiple app servers? Multiple clients?

Slide 20

Slide 20 text

Multiple db servers? Distributed Multiple app servers? Multiple clients?

Slide 21

Slide 21 text

Multiple db servers? Distributed Multiple app servers? Distributed Multiple clients?

Slide 22

Slide 22 text

Multiple db servers? Distributed Multiple app servers? Distributed Multiple clients? Distributed

Slide 23

Slide 23 text

If you build webapps, you build distributed systems.

Slide 24

Slide 24 text

Raft

Slide 25

Slide 25 text

@ongardie Diego Ongaro John Ousterhout and

Slide 26

Slide 26 text

con·sen·sus /kənˈsensəs/ ! Agreeing upon state across distributed processes even in the presence of failures.

Slide 27

Slide 27 text

When should I use it?

Slide 28

Slide 28 text

• Distributed locks • Configuration • Background jobs

Slide 29

Slide 29 text

Why not Paxos?

Slide 30

Slide 30 text

Laying the Groundwork

Slide 31

Slide 31 text

{…} Log State Machine Consensus Module

Slide 32

Slide 32 text

Leader Follower Follower Follower

Slide 33

Slide 33 text

Leader Follower Follower Follower

Slide 34

Slide 34 text

Leader Follower Follower Follower

Slide 35

Slide 35 text

Follower

Slide 36

Slide 36 text

Leader Election Log Replication Safety

Slide 37

Slide 37 text

Follower Candidate Leader

Slide 38

Slide 38 text

Follower Candidate Leader

Slide 39

Slide 39 text

Follower Candidate Leader Times out, Starts election

Slide 40

Slide 40 text

Follower Candidate Leader Times out, Starts election Wins election

Slide 41

Slide 41 text

Follower Candidate Leader Times out, Starts election Times out, Restarts election Wins election

Slide 42

Slide 42 text

Follower Candidate Leader Times out, Starts election Times out, Restarts election Wins election Discovers current or new leader, Steps down

Slide 43

Slide 43 text

Follower Candidate Leader Times out, Starts election Times out, Restarts election Wins election Discovers new leader, Steps down Discovers current or new leader, Steps down

Slide 44

Slide 44 text

RequestVote Used by candidates to ask for votes during an election. Log information included for comparison. ! AppendEntries Used by leaders to tell followers which entries to replicate and commit. Also used as a heartbeat to remain leader. Follower Candidate Vote for me please. Follower Here are some entries for your log. Oh, and btw I’m still the leader. Leader

Slide 45

Slide 45 text

Term Higher numbers are used to determine leaders and check log entries. The term is incremented each time an election is started. Any command with an old term is ignored. Term 1 Term 2 Term 3 Election Normal Operation Split vote

Slide 46

Slide 46 text

Follower Follower Follower Follower

Slide 47

Slide 47 text

Candidate Follower Follower Follower Votes Times out, starts election

Slide 48

Slide 48 text

Candidate Follower Follower Follower Votes Votes for self

Slide 49

Slide 49 text

Candidate Follower Follower Follower Votes Sends out RequestVotes

Slide 50

Slide 50 text

Candidate Follower Follower Votes Responds with success

Slide 51

Slide 51 text

Leader Follower Votes Yay! I won! Follower

Slide 52

Slide 52 text

Leader Election Log Replication Safety

Slide 53

Slide 53 text

Log Entries Explained

Slide 54

Slide 54 text

SET X = 1 SET Y = 2 SET Z = 3 SET X = 4 SET X = 5 1 1 1 2 2 SET X = 1 SET Y = 2 SET Z = 3 SET X = 4 1 1 1 2 Leader Follower 1 2 3 4 5 6

Slide 55

Slide 55 text

{ :entries => [{ 4 => 'SET X = 4' }], :term => 1, :prev_log_term => 1, :prev_log_index => 3, :leader_commit => 3, :leader_id => '192.168.1.101/7238' } APPEND ENTRIES EXAMPLE

Slide 56

Slide 56 text

SET X = 1 SET Y = 2 SET Z = 3 1 1 1 Before 1 1 1 1 After SET X = 1 SET Y = 2 SET Z = 3 SET X = 4 1 2 3 4 5 6 1 2 3 4 5 6

Slide 57

Slide 57 text

Log entries are always committed in the same order and are never uncommitted.

Slide 58

Slide 58 text

Happy Log Entry

Slide 59

Slide 59 text

Leader Follower Follower

Slide 60

Slide 60 text

Leader Follower Follower

Slide 61

Slide 61 text

Leader Follower Follower

Slide 62

Slide 62 text

Leader Follower Follower

Slide 63

Slide 63 text

Leader Follower Follower

Slide 64

Slide 64 text

Sad Log Entry

Slide 65

Slide 65 text

Leader Follower Follower

Slide 66

Slide 66 text

Leader Follower Follower

Slide 67

Slide 67 text

Leader Follower Follower

Slide 68

Slide 68 text

Leader Follower Follower

Slide 69

Slide 69 text

Leader Follower Leader

Slide 70

Slide 70 text

Leader Follower Leader

Slide 71

Slide 71 text

Leader Follower Leader

Slide 72

Slide 72 text

Leader Follower Leader

Slide 73

Slide 73 text

Leader Follower Leader

Slide 74

Slide 74 text

Follower Follower Follower Leader

Slide 75

Slide 75 text

Facepalm Log Entry

Slide 76

Slide 76 text

Leader Follower Follower

Slide 77

Slide 77 text

Leader Follower Follower

Slide 78

Slide 78 text

Leader Follower Follower

Slide 79

Slide 79 text

Leader Follower Follower

Slide 80

Slide 80 text

Leader Follower Follower

Slide 81

Slide 81 text

Leader Follower Follower

Slide 82

Slide 82 text

Leader Follower Leader

Slide 83

Slide 83 text

How do we guard against losing log entries?

Slide 84

Slide 84 text

Leader Election Log Replication Safety

Slide 85

Slide 85 text

Only cast votes for nodes with logs that contain at least as many entries as your own. 1

Slide 86

Slide 86 text

New leaders must commit a log entry form their new term before committing old entries. 2

Slide 87

Slide 87 text

Bonus Round

Slide 88

Slide 88 text

• Cluster changes • Log compaction • Client specifics

Slide 89

Slide 89 text

Why Ruby

Slide 90

Slide 90 text

No content

Slide 91

Slide 91 text

Ruby is great for expressing complex problems.

Slide 92

Slide 92 text

Then why doesn’t the academic community love it?

Slide 93

Slide 93 text

• Community • Understandability • Learning Material

Slide 94

Slide 94 text

• Read Papers • Go To Conferences • Talk To People

Slide 95

Slide 95 text

Raft Paper Raft Implementations Raft Website ThinkDistributed Ricon Google Scholar

Slide 96

Slide 96 text

Question Time Make sure I repeat your questions.