Tweet
Tweet

Slide 1

Slide 1 text

HOW TO DEV&OPS INTERNAL PAAS

Slide 2

Slide 2 text

TAICHI NAKASHIMA @deeeet @tcnksm

Slide 3

Slide 3 text

INTERNAL PAAS? = PaaS for Rakuten engineers

Slide 4

Slide 4 text

ONLY FOR TEST? = No. It receives production requests

Slide 5

Slide 5 text

WHY PAAS? = Fast app experimentation and iteration with PROD-grade

Slide 6

Slide 6 text

WHY PAAS? = You don’t need to prepare servers by yourself

Slide 7

Slide 7 text

WHY PAAS? = You don’t need to provision servers by yourself

Slide 8

Slide 8 text

WHY PAAS? = You don’t need to prepare DBs by yourself

Slide 9

Slide 9 text

WHY PAAS? = You can scale your app by *one command*

Slide 10

Slide 10 text

WHY PAAS? = You can focus on development, not deployment

Slide 11

Slide 11 text

WHY INTERNAL PAAS? = Easy to connect with other internal service

Slide 12

Slide 12 text

WHY INTERNAL PAAS? = Instant support when something happen

Slide 13

Slide 13 text

WHY INTERNAL PAAS? (From organizational point of view) = You can reduce duplicated tooling by different teams

Slide 14

Slide 14 text

HOW LARGE? How many request? servers? language?

Slide 15

Slide 15 text

16000 req/sec. All application requests

Slide 16

Slide 16 text

2500 instances 1400 (PROD) + 700 (STG) + 400 (DEV)

Slide 17

Slide 17 text

4300 VMs 2800 (PROD) + 1200 (STG) + 300 (DEV)

Slide 18

Slide 18 text

+300 VMs/mon. Growth forecasting

Slide 19

Slide 19 text

4 languages support Ruby, Node.js, Java, PHP

Slide 20

Slide 20 text

3 DB services Redis, MongoDB, Clustrix

Slide 21

Slide 21 text

100 Redis clusters 230 Instances

Slide 22

Slide 22 text

40 components Components (Roles) to run PaaS

Slide 23

Slide 23 text

320 chef recipes `ls cookbooks/*/recipes | wc -l`

Slide 24

Slide 24 text

8 Engineers Dev & Ops, From 7 Countries

Slide 25

Slide 25 text

HOW TO DEV&OPS INTERNAL PAAS

Slide 26

Slide 26 text

HOW TO DEV&OPS INTERNAL PAAS

Slide 27

Slide 27 text

No content

Slide 28

Slide 28 text

Router API Health Check Messaging DBs Apps

Slide 29

Slide 29 text

DEV FLOW RELEASE FLOW

Slide 30

Slide 30 text

DEV FLOW RELEASE FLOW

Slide 31

Slide 31 text

Create Ticket on JIRA Write code Write Chef cookbook Test on LAB Create PR (Git-Flow) Review

Slide 32

Slide 32 text

DEV FLOW RELEASE FLOW

Slide 33

Slide 33 text

Assign release manager Collect all JIRA tickets Write internal blog CanaryRelease Release

Slide 34

Slide 34 text

1 release for 1 week DEV (2day) , STG (2day) , PROD(3day)

Slide 35

Slide 35 text

HOW TO RELEASE? = Chef + Capistrano

Slide 36

Slide 36 text

RELEASE 1 SERVER

Slide 37

Slide 37 text

Service-out Run Chef solo Run Serverspec Service-in

Slide 38

Slide 38 text

Stop Load-Balancing Disable Health Check Stop monit Service-out Run Chef solo Run Serverspec Service-in Start monit Enable Health Check Start Load-Balancing

Slide 39

Slide 39 text

/etc/service-out /etc/service-in Service-out Run Chef solo Run Serverspec Service-in

Slide 40

Slide 40 text

Every server has same startup/stop scripts = workflow is same = automation is easy

Slide 41

Slide 41 text

RELEASE X SERVERS

Slide 42

Slide 42 text

cap service-in cap service-out cap setup-role Service-out X servers Run Chef solo X servers Run Serverspec X servers Service-in X servers

Slide 43

Slide 43 text

Role A Role B Role C Operation 170.20.20.21.RoleA 170.20.20.22.RoleA 170.20.20.23.RoleA 170.20.20.24.RoleA 170.20.20.25.RoleA 170.20.20.26.RoleA 170.20.20.27.RoleA VMLIST

Slide 44

Slide 44 text

cap service-out 170.20.20.21.RoleA 170.20.20.22.RoleA 170.20.20.23.RoleA 170.20.20.24.RoleA 170.20.20.25.RoleA 170.20.20.26.RoleA 170.20.20.27.RoleA VMLIST Operation Role A Role B Role C Parallel execution

Slide 45

Slide 45 text

cap setup-role Operation Parallel execution 170.20.20.21.RoleA 170.20.20.22.RoleA 170.20.20.23.RoleA 170.20.20.24.RoleA 170.20.20.25.RoleA 170.20.20.26.RoleA 170.20.20.27.RoleA VMLIST Role A Role B Role C

Slide 46

Slide 46 text

cap service-in Role A Role B Role C Operation 170.20.20.21.RoleA 170.20.20.22.RoleA 170.20.20.23.RoleA 170.20.20.24.RoleA 170.20.20.25.RoleA 170.20.20.26.RoleA 170.20.20.27.RoleA VMLIST Parallel execution

Slide 47

Slide 47 text

cap service-out Operation Parallel execution 170.20.20.31.RoleB 170.20.20.32.RoleB 170.20.20.33.RoleB 170.20.20.34.RoleB 170.20.20.35.RoleB 170.20.20.36.RoleB 170.20.20.37.RoleB VMLIST Role A Role B Role C

Slide 48

Slide 48 text

cap service-out 170.20.20.21.RoleA VMLIST Operation Role A Role B Role C Start from Canary

Slide 49

Slide 49 text

HOW TO DEV&OPS INTERNAL PAAS

Slide 50

Slide 50 text

LOGGING MONITORING ALERT HANDLING SUPPORT IAAS

Slide 51

Slide 51 text

LOGGING MONITORING ALERT HANDLING SUPPORT IAAS

Slide 52

Slide 52 text

700GB/day logs All logs produced in PaaS

Slide 53

Slide 53 text

LOGGING IN PAAS? = Application logs + Component logs

Slide 54

Slide 54 text

APPLICATION LOG ? = PaaS should provide user the way to debug

Slide 55

Slide 55 text

Instant logs Midterm logs Longterm logs Real time 1-2 weeks - 6 month

Slide 56

Slide 56 text

Router API Health Check Messaging DBs Apps Instant log

Slide 57

Slide 57 text

Log Server Apps Object Storage Instant log Midterm log Longterm log

Slide 58

Slide 58 text

Log Server Apps Instant log Midterm log Hadoop (BigData team) Analytics

Slide 59

Slide 59 text

Log Server Apps Instant log Midterm log Splunk  Dashboard

Slide 60

Slide 60 text

COMPONENT LOG ? = Log which we use for debug PaaS itself

Slide 61

Slide 61 text

Log Server Object Storage

Slide 62

Slide 62 text

Log Server Object Storage We can debug CF here

Slide 63

Slide 63 text

Log Server Object Storage GlusterFS LeoFS

Slide 64

Slide 64 text

Log Server Object Storage GlusterFS

Slide 65

Slide 65 text

LOGGING METRICS ALERT HANDLING SUPPORT IAAS

Slide 66

Slide 66 text

OpenTSDB, Pandra FMS

Slide 67

Slide 67 text

LOGGING METRICS ALERT HANDLING SUPPORT IAAS

Slide 68

Slide 68 text

1 week, 24H charge Primary & Sub admin

Slide 69

Slide 69 text

✉

Slide 70

Slide 70 text

2500 ✉/day MAX. Need to fix…

Slide 71

Slide 71 text

LOGGING METRICS ALERT HANDLING SUPPORT IAAS

Slide 72

Slide 72 text

JIRA, HipChat Instant support is one of *good* point of Internal PaaS

Slide 73

Slide 73 text

LOGGING METRICS ALERT HANDLING SUPPORT IAAS

Slide 74

Slide 74 text

IAAS Operating PaaS also means operating IaaS

Slide 75

Slide 75 text

vSphere

Slide 76

Slide 76 text

HOW TO BOOT SERVERS? = Internal tool like terraform

Slide 77

Slide 77 text

Role A vSphere Operation rvc create -c rvc.yml 170.20.21.RoleA RoleA: cpu: 2 mem: 8192 rvc.yml 170.20.20.21.RoleA 170.20.20.22.RoleA 170.20.20.23.RoleA VMLIST

Slide 78

Slide 78 text

Role A vSphere Operation VMLIST rvc create -c rvc.yml 170.20.21.RoleA RoleA: cpu: 2 mem: 8192 rvc.yml 170.20.20.21.RoleA 170.20.20.22.RoleA 170.20.20.23.RoleA VMLIST

Slide 79

Slide 79 text

Role A vSphere Operation rvc create -c rvc.yml 170.20.21.RoleA RoleA: cpu: 2 mem: 8192 rvc.yml 170.20.20.21.RoleA 170.20.20.22.RoleA 170.20.20.23.RoleA VMLIST

Slide 80

Slide 80 text

Role A vSphere Operation rvc create -c rvc.yml 170.20.22.RoleA RoleA: cpu: 2 mem: 8192 rvc.yml 170.20.20.21.RoleA 170.20.20.22.RoleA 170.20.20.23.RoleA VMLIST

Slide 81

Slide 81 text

Role A vSphere Operation rvc create -c rvc.yml 170.20.23.RoleA RoleA: cpu: 2 mem: 8192 rvc.yml 170.20.20.21.RoleA 170.20.20.22.RoleA 170.20.20.23.RoleA VMLIST

Slide 82

Slide 82 text

cap setup-role Role A Operation vSphere 170.20.20.21.RoleA 170.20.20.22.RoleA 170.20.20.23.RoleA VMLIST

Slide 83

Slide 83 text

cap setup-role Role A Operation vSphere 170.20.20.21.RoleA 170.20.20.22.RoleA 170.20.20.23.RoleA VMLIST

Slide 84

Slide 84 text

Easy to boot & setup servers = If there is *physical resource*

Slide 85

Slide 85 text

FUTURE? = We are moving to *version 2*

Slide 86

Slide 86 text

BE GOPHER CloudFoundry moves from Ruby to Golang

Slide 87

Slide 87 text

NO FORK Everything goes to upstream

Slide 88

Slide 88 text

BE OPEN Building tool as OSS

Slide 89

Slide 89 text

✉

Slide 90

Slide 90 text

NO MORE TOO MUCH ✉ Planing to use Pagerduty + Riemann

Slide 91

Slide 91 text

Log Server Object Storage GlusterFS LeoFS

Slide 92

Slide 92 text

Object Storage LeoFS Kafka

Slide 93

Slide 93 text

MORE FLEXIBLE LOG STACK Planning to use Apache Kafka

Slide 94

Slide 94 text

NEW METRICS STACK Planning to use InfluxDB + Grafana

Slide 95

Slide 95 text

CONTAINER Planning to support Docker

Slide 96

Slide 96 text

MORE HA Planning to have a ChaosMonkey

Slide 97

Slide 97 text

NEW IAAS Migrating to OpenStack

Slide 98

Slide 98 text

NEW IAAS Planning to Hybrid Cloud

Slide 99

Slide 99 text

WE HAVE MANY CHALLENGES

Slide 100

Slide 100 text

WE ARE HIRING http://corp.rakuten.co.jp/careers/experienced/

Slide 101

Slide 101 text

@deeeet