Slide 1

Slide 1 text

Data Migration with Confidence Juanito Fatas RedDotRubyConf 2017

Slide 2

Slide 2 text

@JuanitoFatas Ramen Specialist

Slide 3

Slide 3 text

Spanish name From Taiwan Live in Tokyo

Slide 4

Slide 4 text

I became a salaryman #

Slide 5

Slide 5 text

Cookpad Global Cookpad Japan

Slide 6

Slide 6 text

Made in Japan $

Slide 7

Slide 7 text

Reserve for local jokes %

Slide 8

Slide 8 text

Data Migration?

Slide 9

Slide 9 text

Schema Migration
 Data Migration

Slide 10

Slide 10 text

Schema Migration Alter Schemas over time https://en.wikipedia.org/wiki/Schema_migration

Slide 11

Slide 11 text

Data Migration https://en.wikipedia.org/wiki/Data_migration Transfer data from to A System B System

Slide 12

Slide 12 text

Existing Data Migration

Slide 13

Slide 13 text

No content

Slide 14

Slide 14 text

Data Migration to Existing System

Slide 15

Slide 15 text

Data Migration with Confidence @JuanitoFatas Specialist of cookpad RedDotRubyConf 2017

Slide 16

Slide 16 text

Why Migration?

Slide 17

Slide 17 text

Rewrote for Clients

Slide 18

Slide 18 text

New Partner joins company

Slide 19

Slide 19 text

Data Migration

Slide 20

Slide 20 text

~ Get All Data to our system Simple Goal

Slide 21

Slide 21 text

~ Import data Modeling Migrate After migrate

Slide 22

Slide 22 text

~ Get the Data Provider API Data Dump

Slide 23

Slide 23 text

Provider API & Generic Migration Code ✅

Slide 24

Slide 24 text

Data Dump & Generic Migration Code

Slide 25

Slide 25 text

HOWTO Data Migration

Slide 26

Slide 26 text

~ Start with a rake task

Slide 27

Slide 27 text

~

Slide 28

Slide 28 text

~ lib/tasks/data_migration.rake

Slide 29

Slide 29 text

~ lib/data_migration.rb

Slide 30

Slide 30 text

~ Import dump to local

Slide 31

Slide 31 text

~ GBs-size file

Slide 32

Slide 32 text

monthly users 62 30M countries ~

Slide 33

Slide 33 text

~ Add delay to the SQL dump for production

Slide 34

Slide 34 text

~ Add sleep() before INSERT INTO

Slide 35

Slide 35 text

~ Editing huge file

Slide 36

Slide 36 text

~ +

Slide 37

Slide 37 text

~

Slide 38

Slide 38 text

~ Enumerable#lazy https://ruby-doc.org/core-2.4.1/Enumerable.html#method-i-lazy

Slide 39

Slide 39 text

~

Slide 40

Slide 40 text

~

Slide 41

Slide 41 text

~ set accordingly for staging & production

Slide 42

Slide 42 text

~ Modeling Database

Slide 43

Slide 43 text

~

Slide 44

Slide 44 text

~ With these 5 methods, you can model anything.

Slide 45

Slide 45 text

~ Map data to your current system

Slide 46

Slide 46 text

~ Sometimes as easy as

Slide 47

Slide 47 text

~ As HTML in the recipes table field ‘steps’ , Sometimes…

Slide 48

Slide 48 text

~

Slide 49

Slide 49 text

~

Slide 50

Slide 50 text

~ Setup Test Suite

Slide 51

Slide 51 text

~

Slide 52

Slide 52 text

~

Slide 53

Slide 53 text

~ - Why Tests? The migration code only used once

Slide 54

Slide 54 text

~ Better code through boring tests

Slide 55

Slide 55 text

~ TDD to Get Things Done

Slide 56

Slide 56 text

~ Modeling Tests Repeat

Slide 57

Slide 57 text

Migration

Slide 58

Slide 58 text

~ Use all methods that raises exception

Slide 59

Slide 59 text

~ Fail Fast to find all errors

Slide 60

Slide 60 text

~ Example Migrate Recipes

Slide 61

Slide 61 text

~

Slide 62

Slide 62 text

~

Slide 63

Slide 63 text

No content

Slide 64

Slide 64 text

~

Slide 65

Slide 65 text

~ Add more migrators to migrate

Slide 66

Slide 66 text

~ Data Integrity

Slide 67

Slide 67 text

~ Transaction

Slide 68

Slide 68 text

~ Idempotent Operation

Slide 69

Slide 69 text

~

Slide 70

Slide 70 text

~ Run Migration many times

Slide 71

Slide 71 text

~ Produce the Same Result

Slide 72

Slide 72 text

~ f(x) = f(x)

Slide 73

Slide 73 text

~ Upsert Update or Insert

Slide 74

Slide 74 text

~ MySQL ON DUPLICATE KEY UPDATE PostgreSQL ON CONFLICT UPDATE PostgreSQL 9.5+ seamusabshere/upsert

Slide 75

Slide 75 text

~

Slide 76

Slide 76 text

~ Data Accuracy

Slide 77

Slide 77 text

~ Manually Check

Slide 78

Slide 78 text

~ Automated Check

Slide 79

Slide 79 text

~ Example Check users with most Recipes

Slide 80

Slide 80 text

~

Slide 81

Slide 81 text

~

Slide 82

Slide 82 text

~

Slide 83

Slide 83 text

~

Slide 84

Slide 84 text

~

Slide 85

Slide 85 text

~ To check more things Add more Checker objects

Slide 86

Slide 86 text

~ Use many small objects to compose

Slide 87

Slide 87 text

~ Objects Everywhere

Slide 88

Slide 88 text

For better object design

Slide 89

Slide 89 text

For better object design

Slide 90

Slide 90 text

Background Jobs

Slide 91

Slide 91 text

Workers = CPU cores

Slide 92

Slide 92 text

Designated Queues

Slide 93

Slide 93 text

No content

Slide 94

Slide 94 text

No content

Slide 95

Slide 95 text

Log Every Unexpected Error

Slide 96

Slide 96 text

No content

Slide 97

Slide 97 text

No content

Slide 98

Slide 98 text

For Better handling of Errors

Slide 99

Slide 99 text

Run against all data to be migrated

Slide 100

Slide 100 text

Fix every error you can before real migration

Slide 101

Slide 101 text

~ Tools

Slide 102

Slide 102 text

Retry mechanism

Slide 103

Slide 103 text

Foreign Key Constraints Locks

Slide 104

Slide 104 text

MySQL deadlock Results in

Slide 105

Slide 105 text

Automatic Retry # Rails 4: ActiveRecord::StatementInvalid

Slide 106

Slide 106 text

Make sure what should be Retry

Slide 107

Slide 107 text

retry_on discard_on ActiveJob::Exceptions

Slide 108

Slide 108 text

Automatic Retry

Slide 109

Slide 109 text

Examine & Retry

Slide 110

Slide 110 text

In Resque

Slide 111

Slide 111 text

No content

Slide 112

Slide 112 text

Status Reporting

Slide 113

Slide 113 text

No content

Slide 114

Slide 114 text

No content

Slide 115

Slide 115 text

Report every minute

Slide 116

Slide 116 text

Monitoring CPU Usage

Slide 117

Slide 117 text

No content

Slide 118

Slide 118 text

~ Performance

Slide 119

Slide 119 text

Performance is a Rabbit hole

Slide 120

Slide 120 text

Preload associations

Slide 121

Slide 121 text

Minimize scope of transaction

Slide 122

Slide 122 text

Transaction Isolation Levels https://dev.mysql.com/doc/refman/5.7/en/innodb-transaction-isolation-levels.html https://www.postgresql.org/docs/current/static/transaction-iso.html

Slide 123

Slide 123 text

Avoid unnecessary callbacks

Slide 124

Slide 124 text

No content

Slide 125

Slide 125 text

Example You can touch http://api.rubyonrails.org/classes/ActiveRecord/NoTouching/ClassMethods.html#method-i-no_touching after migration

Slide 126

Slide 126 text

Process multiple records in one job

Slide 127

Slide 127 text

No content

Slide 128

Slide 128 text

Cache data in Memory

Slide 129

Slide 129 text

Cache data in Redis

Slide 130

Slide 130 text

Migrate Important things first

Slide 131

Slide 131 text

First 10000 users w/ most recipes

Slide 132

Slide 132 text

IO bound

Slide 133

Slide 133 text

Scale up the Database

Slide 134

Slide 134 text

Decrease the workers /0

Slide 135

Slide 135 text

Bulk Insert Bulk Upsert* * Only MySQL supports bulk upsert zdennis/activerecord-import

Slide 136

Slide 136 text

Every change to make it fast

Slide 137

Slide 137 text

Run the WHOLE migration again

Slide 138

Slide 138 text

Keep CPU usage max at 75% all

Slide 139

Slide 139 text

~ Post Migration

Slide 140

Slide 140 text

Update all necessities

Slide 141

Slide 141 text

Redirects

Slide 142

Slide 142 text

Redirect tables Cookpad Redirect programs Server redirects Provider

Slide 143

Slide 143 text

Redirection Service cookpad/mirin

Slide 144

Slide 144 text

~ Stories

Slide 145

Slide 145 text

Cases of Email

Slide 146

Slide 146 text

Remove duplicate emails before migration

Slide 147

Slide 147 text

Remove invalid emails before migration

Slide 148

Slide 148 text

downcase all the emails

Slide 149

Slide 149 text

~ Get Site Dump

Slide 150

Slide 150 text

~ 100GB generated on EC2* EC2 has bandwidth limits

Slide 151

Slide 151 text

~ scp takes days ONLY if nothing failed within days

Slide 152

Slide 152 text

~ delivers encrypted disk

Slide 153

Slide 153 text

Migrate Millions of records

Slide 154

Slide 154 text

AR + transaction bulk in/upsert activerecord-import load data in file

Slide 155

Slide 155 text

Weeks Month-ish

Slide 156

Slide 156 text

Run low priority job to migrate them

Slide 157

Slide 157 text

When migrated User signed in

Slide 158

Slide 158 text

Migrate their data in high priority

Slide 159

Slide 159 text

No content

Slide 160

Slide 160 text

Migrate 100K photos

Slide 161

Slide 161 text

How our image work

Slide 162

Slide 162 text

Design so it produces the same hash 4

Slide 163

Slide 163 text

Set the designated hash during migration instead of upload, generate hash

Slide 164

Slide 164 text

Benchmark how long to finish all

Slide 165

Slide 165 text

X days?

Slide 166

Slide 166 text

Migrate them X days before in low priority

Slide 167

Slide 167 text

99% photos won’t change

Slide 168

Slide 168 text

Migrate users password to secure auth

Slide 169

Slide 169 text

Figure out what algorithm(s) was(ere) used

Slide 170

Slide 170 text

When migrated user signed in

Slide 171

Slide 171 text

System’s password auth will fail

Slide 172

Slide 172 text

Fallback to Legacy Auth

Slide 173

Slide 173 text

No content

Slide 174

Slide 174 text

When password matched from legacy auth

Slide 175

Slide 175 text

Set his password through secure password scheme

Slide 176

Slide 176 text

No content

Slide 177

Slide 177 text

Migration ~ The Future

Slide 178

Slide 178 text

Migration done in Ghost Table fashion

Slide 179

Slide 179 text

Data Dump & Generic Migration Code Only need to modeling database

Slide 180

Slide 180 text

~ Takeaways

Slide 181

Slide 181 text

Rails provides sharp tools thanks to rails core team

Slide 182

Slide 182 text

Use Small objects to make your code more readable & maintainable

Slide 183

Slide 183 text

Abstraction is the God of Programming! “ ” — Matthew Mongeau @halogenandtoast

Slide 184

Slide 184 text

Schedule >> Fast

Slide 185

Slide 185 text

Schedule >> Fast Integrity >> schedule

Slide 186

Slide 186 text

Data Migration sounds hard

Slide 187

Slide 187 text

Keep it Simple Made it Easy

Slide 188

Slide 188 text

Do the Simplest Things “ ” — Winston Teo Yon Wei @winstonyw

Slide 189

Slide 189 text

Enjoy ☕ Thank you!