Enabling Microservice @ Orbitz - GOTO Chicago 2016
•
2 likes•461 views
In this talk we will discuss how we enabled decomposition of one of our 500+ system components into a continuously deployed microservice cluster. Our platform is comprised of Apache Mesos/Marathon, Docker, and a number of local services including Consul for service discovery, Logstash for diskless logging, and a custom metrics forwarder to Graphite. Building on this, we'll detail our CI pipeline using Jenkins workflows to build and publish microservices as Docker images, test and deploy via Marathon/Mesos, and automated change tickets. Finally, we'll discuss lessons learned from building our own enterprise PaaS and scaling it out to a large organization.
Recommended
Recommended
More Related Content
What's hot
What's hot (20)
Viewers also liked
Viewers also liked (6)
Similar to Enabling Microservice @ Orbitz - GOTO Chicago 2016
Similar to Enabling Microservice @ Orbitz - GOTO Chicago 2016 (20)
More from Steve Hoffman
More from Steve Hoffman (6)
Recently uploaded
Recently uploaded (20)
Enabling Microservice @ Orbitz - GOTO Chicago 2016
- 1. Enabling Microservices @ Orbitz Steve Hoffman Senior Principal Engineer @bacoboy Rick Fast Principal Engineer @tortiepoint
- 3. #GOTOChgo 2016 Agenda • Brief Architecture Overview/History • From Monolithic to Services to Microservices with Docker • Automated Pipelines • Questions
- 21. #GOTOChgo 2016 2012 • Multiple Brands • Websites • Webservices • Multiple Backends • 500+ apps / thousands of instances • Deployments Daily (sometimes more)
- 22. #GOTOChgo 2016 2015 • Multiple Brands • Websites • Webservices • Multiple Backends • 500+ apps / thousands of instances • Deployments Daily (sometimes more)
- 23. #GOTOChgo 2016 • Multiple Brands • Websites • Webservices • Multiple Backends • 500+ apps / thousands of instances • Deployments Daily (sometimes more) 2015
- 24. #GOTOChgo 2016 2016 • Multiple Brands • Websites • Webservices • Multiple Backends • 500+ apps / thousands of instances • Deployments Daily (sometimes more)
- 26. #GOTOChgo 2016 Different Provisioning Tools DEV OPS
- 27. #GOTOChgo 2016 Different Provisioning Tools DEV OPS Application Platform!=
- 28. #GOTOChgo 2016 A New Experiment • Microservices? • Decompose single “service” into the actual 40+ sub-services • Any change to sub-service was a deployment of many • Could it be a simple Spring Boot App in Docker (12 factor?) • Backward compatible with existing service infrastructure. • Code to Production w/o help from other Humans
- 29. #GOTOChgo 2016 The Docker Slave aka “The Rickbot”
- 30. #GOTOChgo 2016 The Docker Slave aka “The Rickbot” App App App
- 33. #GOTOChgo 2016 App App App Register & Lookup
- 46. #GOTOChgo 2016 Mesos Scheduler Marathon Scheduler Slave xNMaster x3 ZK x3
- 58. #GOTOChgo 2016 A Little More Background…
- 59. #GOTOChgo 2016
- 61. #GOTOChgo 2016 Editorial Module A Continuously Deployed Microservice www.orbitz.com (scroll down)
- 63. #GOTOChgo 2016 Orbitz Content Orchestration Service Content (Solr) Editorial Module Search Module Hotel Module Hotel Team Search Team Content Team
- 65. #GOTOChgo 2016 Orbitz Content Orchestration Service Content (Solr) Editorial Module Search Module Hotel Module Hotel Team Search Team Content Team
- 66. #GOTOChgo 2016 Pre-Continuous Delivery 1.2-BETA-20150401-113002 Test Discard Test Discard 1.2-BETA-20150402-093002 Build Build And so on…
- 67. #GOTOChgo 2016 Pre-Continuous Delivery 1.2-BETA-20150401-113002 Test Discard Test Discard 1.2-BETA-20150402-093002 Build Build Test Deploy 1.2 Build
- 68. #GOTOChgo 2016 Continuous Delivery with Jenkins, Docker, Ansible and Marathon
- 69. #GOTOChgo 2016 Yo rfast-mbp:git rfast$ yo microservice _-----_ | | .--------------------------. |--(o)--| | Welcome to the kickass | `---------´ | Microservice | ( _´U`_ ) | generator! | /___A___ '--------------------------' | ~ | __'.___.'__ ´ ` |° ´ Y ` ? Enter the name of your service. (E.G. "my-service")
- 75. #GOTOChgo 2016 if(…) { x ++; } Merge
- 76. #GOTOChgo 2016 Pipeline (Simplified) build unit test publish deploy dev acceptance tests deploy qa deploy staging open RFC deploy prod close RFC
- 77. #GOTOChgo 2016 Build if(…) { x ++; } 1.2 Merge
- 78. #GOTOChgo 2016 Build if(…) { x ++; } 1.2 Merge 1.2. editorial-module.jar ./gradlew build {BUILD_NUMBER}
- 79. #GOTOChgo 2016 Build 1.2 Merge editorial-module.jar if(…) { x ++; } 171.2.
- 80. #GOTOChgo 2016 Build 1.2 Merge editorial-module.jar FROM orbitz/java-8 COPY build/editorial-module.jar /opt/orbitz WORKDIR /opt/orbitz ENTRYPOINT [“java”,”-jar”,”editorial-module.jar”] if(…) { x ++; } 171.2.
- 81. #GOTOChgo 2016 Build 1.2 Merge editorial-module.jar orbitz/editorial-module:1.2.17 if(…) { x ++; } 171.2.
- 83. #GOTOChgo 2016 Deploy - hosts: dev - tasks: - name: find previous images docker_facts: image=orbitz/{{application}} register: previous - name: deploy new image docker: image=orbitz/{{application}}:{{version}} … - name: wait for service wait_for: port={{port}} … - name: check health endpoint uri: url="http://{{fqdn}}:{{port}}/health" … - name: kill old image docker: {{previous}} git pull playbook.yml (abridged)
- 84. #GOTOChgo 2016 - hosts: dev host-001 host-002 host-003 host-004
- 85. #GOTOChgo 2016 - hosts: dev host-001 host-002 host-003 host-004
- 86. #GOTOChgo 2016 - hosts: dev host-001 host-002 host-003 host-004
- 87. #GOTOChgo 2016 Downsides • How to handle failure with YAML files? • What happens when a VM is moved (destroyed/created)? • What if I need to add capacity?
- 88. #GOTOChgo 2016 Deploy (Second Attempt) - hosts: dev - tasks: - name: find previous images docker_facts: image=orbitz/{{application}} register: previous - name: deploy new image docker: image=orbitz/{{application}}:{{version}} … - name: wait for service wait_for: port={{port}} … - name: check health endpoint uri: url="http://{{fqdn}}:{{port}}/health" … - name: kill old image docker: {{previous}} git pull playbook.yml (abridged)
- 89. #GOTOChgo 2016 Deploy (Second Attempt) - hosts: dev - tasks: - name: find previous images docker_facts: image=orbitz/{{application}} register: previous - name: deploy new image docker: image=orbitz/{{application}}:{{version}} … - name: wait for service wait_for: port={{port}} … - name: check health endpoint uri: url="http://{{fqdn}}:{{port}}/health" … - name: kill old image git pull playbook.yml (abridged) - hosts: localhost - tasks: - name: marathon deploy marathon: image=orbitz/{{application}}:{{version}} instances=3
- 90. #GOTOChgo 2016 Deploy (Second Attempt) git pull playbook.yml (abridged) - hosts: localhost - tasks: - name: marathon deploy marathon: image=orbitz/{{application}}:{{version}} instances=3
- 91. #GOTOChgo 2016 Deploy (Second Attempt) New host setup
- 92. #GOTOChgo 2016 Deploy (Second Attempt) Mesos Agent New host setup 1.2.16
- 93. #GOTOChgo 2016 - tasks: marathon: … playbook.yml 1.2.16 1.2.16 1.2.16
- 94. #GOTOChgo 2016 - tasks: marathon: … 1.2.16 1.2.16 1.2.16
- 95. #GOTOChgo 2016 - tasks: marathon: … 1.2.16 1.2.16 1.2.16
- 96. #GOTOChgo 2016 - tasks: marathon: … 1.2.16 1.2.16 1.2.16 PUT /apps/editorial-module { “image”: “orbitz/editorial-module:1.2.17” … }
- 97. #GOTOChgo 2016 PUT /apps/editorial-module { “image”: “orbitz/editorial-module:1.2.17” … } - tasks: marathon: … 1.2.16 1.2.16 1.2.16 app = GET /v2/apps/editorial-module if not app then deploy_id = POST /v2/apps { “image”: “orbitz/editorial-module:1.2.17”, “id”: “editorial-module” } else deploy_id = PUT /v2/apps/editorial-module { “image”: “orbitz/editorial-module:1.2.17” } end if while GET /v2/deployments contains deploy_id // still deploying end // deploy complete
- 98. #GOTOChgo 2016 - tasks: marathon: … 1.2.16 1.2.16 1.2.16 “env”: { “SOME_DATABASE_URL”: “jdbc://whatever”, “SOME_DATABASE_USER”: “rickfast”, “SOME_OTHER_VALUE”: “whatever” } PUT /apps/editorial-module { “image”: “orbitz/editorial-module:1.2.17” … }
- 99. #GOTOChgo 2016 - tasks: marathon: … PUT /apps/editorial-module { “image”: “orbitz/editorial-module:1.2.17” … } 1.2.16 1.2.16 1.2.16
- 100. #GOTOChgo 2016 - tasks: marathon: … 1.2.16 1.2.16 1.2.16
- 101. #GOTOChgo 2016 - tasks: marathon: … 1.2.16 1.2.16 1.2.16 1.2.17 1.2.17 1.2.17
- 102. #GOTOChgo 2016 - tasks: marathon: … /health 1.2.16 1.2.16 1.2.16 1.2.17 1.2.17 1.2.17
- 103. #GOTOChgo 2016 - tasks: marathon: … /health 200 OK 200 OK 200 OK 1.2.16 1.2.16 1.2.16 1.2.17 1.2.17 1.2.17
- 104. #GOTOChgo 2016 - tasks: marathon: … 1.2.17 1.2.17 1.2.17
- 105. #GOTOChgo 2016 And off to the next environment… 1.2.17 1.2.17 1.2.17
- 109. #GOTOChgo 2016 1.2.17 1.2.17 1.2.17
- 110. #GOTOChgo 2016 /health 200 OK 1.2.17 1.2.17 1.2.17
- 111. #GOTOChgo 2016 1.2.17 1.2.17 1.2.17
- 116. #GOTOChgo 2016 Build Unit Test Deploy Dev Deploy Prod Deploy Staging Acceptance Test Code Review & Push Production Pre-Production Open RFC Close RFC
- 117. #GOTOChgo 2016 Pipeline as code! node { git ‘https://stash.orbitz.com/editorial-module.git‘ sh ‘./gradlew clean build’ def image = docker.build “editorial-module:${env.BUILD_TAG}" image() }
- 118. #GOTOChgo 2016 Pipeline as code! node { git ‘http://stash.orbitz.com/deploy-playbook.git' sh ‘./ansible-playbook … editorial-module:${env.BUILD_TAG} test’ } node { git ‘https://stash.orbitz.com/editorial-module.git‘ sh ‘./gradlew clean build’ def image = docker.build “editorial-module:${env.BUILD_TAG}" image.push() }
- 119. #GOTOChgo 2016 Pipeline as code! pipeline.json { “steps”: [ { “type”: “gradle-build” }, { “type”: “marathon-deploy”, “environment”: “test” } ] }
- 120. #GOTOChgo 2016 Pipeline as code! def pipeline = new JsonSluper().parse(‘pipeline.json’) pipeline.steps.each { step -> if (step.type == ‘gradle-build’) { node(‘java’) { git ‘https://stash.orbitz.com/${service}.git’ sh ‘./gradlew clean build’ def image = docker.build “${service}:${env.BUILD_TAG}” image.push() } } … }
- 121. #GOTOChgo 2016 Jenkins CI/CD - on Mesos? • Existing Solution • Dedicated Jenkins Slaves • Hand created • Snapshotted & Rolled Back to “Clean” state after each Job • Hard to Manage Build Environment for 300+ apps across many OSes, Java versions, Ruby versions, perl versions, python versions, protocol buffer compiler versions, etc…
- 122. #GOTOChgo 2016 Master Slave Commit/Push or Pull Request/Merge Before
- 123. #GOTOChgo 2016 Master Slave Poll or Push Trigger Commit/Push or Pull Request/Merge Clone & Build Push Artifacts Before
- 126. #GOTOChgo 2016 Master Master Commit/Push or Pull Request/Merge After Slave env:jenkins
- 127. #GOTOChgo 2016 Master Master Slave Poll or Push Trigger Commit/Push or Pull Request/Merge Clone & Build Push Artifacts After Slave env:jenkins
- 130. #GOTOChgo 2016 Master Master Slave Slave Deploy App Clone & Run Slave env:jenkins env:ci docker pull
- 132. #GOTOChgo 2016 Prod Network Pre-Prod Network env:ci env:qa env:staging env:jenkins Master env:prod
- 133. #GOTOChgo 2016 DC: East DC: West env:ci dc:west env:qa dc:west env:staging dc:west env:jenkins dc:west Master env:prod dc:east env:prod dc:west
- 134. #GOTOChgo 2016 Learnings • Create a shared platform for docker deployments using shared and app-specific “localhost” helpers — this was ours, yours SHOULD look different — adapt to change, don’t fight it. • Take people out of the release process • Docker & Packer AMIs - repeatable applications/servers • Chef - repeatable infrastructure & configuration — Environment aware • Jenkins - repeatable releases • Delineate configuration concerns: • Known at Compile time — Bake into Docker image • Known at Boot time — Pass from Playbook/Launcher/Jenkins - parameter to Docker run/AMI • Changes Anytime — Externalize (consul K/V, etcd, zookeeper)
- 135. #GOTOChgo 2016