Continuous Delivery with Spinnaker and Kubernetes
Download as PPTX, PDF1 like404 views
This document provides an overview of Continuous Delivery with Spinnaker and Kubernetes. It discusses SDL Language Cloud's migration from a legacy Jenkins deployment system to using Docker, Kubernetes, Nexus, and Spinnaker. Key points include: - SDL is building a language platform using microservices deployed to private datacenters and public clouds - The legacy Jenkins system was cumbersome so they adopted Spinnaker for deployment management and Kubernetes for container orchestration - Spinnaker supports delivery pipelines, deployment strategies, protection mechanisms, and integrations for CI/CD and multiple clouds - A live demo shows how SDL configured a Spinnaker pipeline to deploy Language Cloud services to Kubernetes
Continuous Delivery with Spinnaker and Kubernetes
- 1. Continuous Delivery with Spinnaker and Kubernetes by Adrian Ivan
- 2. About myself • Cluj Napoca, Romania • Software Architect at SDL • Language Cloud Platform Team
- 3. Agenda • SDL Language Cloud • Language Cloud Legacy Deployment • Spinnaker • Language Cloud Service delivery pipelines • Closing thoughts • Questions
- 4. SDL Language Cloud • SDL Language Tech - SaaS Offering • Monetize Language capabilities in the Cloud • Technical Strategy: • Build a Platform • API-s • Solutions meshes
- 5. SDL Language Cloud - Architecture • MicroServices • Java and .Net – approx. 40 MicroServices (Q1 2018) • Spring Boot/Cloud + Netflix OSS stack • Eureka/Zuul/Ribbon/Hystrix • Production since 2015 • Chef - Configuration management • Deployment to • Private Datacenter • Public Cloud – Estimated Q4 2018
- 6. Continuous Delivery • MicroServices: It is all about Speed • Automation • ZERO Downtime • Confidence: • Testing: Thoughtworks Testing Model for Microservices • Ability to Recover
- 7. Legacy Deployment - Jenkins • Pipeline • Delivery artifact - RPM • Jenkins orchestration- Multiple stages • Chef recipes • Automated and manual promotions • Shortcomings • Deployment not a First Class citizen • Home grown solutions via scripts • No Authorization • Works, but cumbersome
- 8. Deployment: New Stack • Docker Containers – Delivery artifact • Nexus3 – Artifact repository • Kubernetes – Container orchestration • Spinnaker – Deployment management
- 9. Spinnaker “Spinnaker is an open source, multi-cloud continuous delivery platform that helps you release software changes with high velocity and confidence” spinnaker.io • Initiated by Netflix (replacement for Asgard) • Backed-up by major players: Google, Microsoft, Pivotal, etc • Becoming the de-facto tool for deploying in the Cloud • Opinionated based on experience and lessons learned
- 10. Spinnaker Features • Delivery Pipelines • Triggers + Multiple stages • Predefined Stage types • CI Systems integrations: Jenkins, Travis • Cloud Integrations(Cloud Drivers): • AWS, GCE, GKE, Cloud Foundry, Azure, OpenStack, Oracle, DC/OS, Netflix Titus • K8s • Enables Global Strategy: Multi-Cloud/Multi-region • Deployment Strategies: • Rolling, Blue/Green - Red/Black, Canary
- 11. Spinnaker Features • Protection: • Deployment Time Restrictions • Notifications: Slack, HipChat, Email • Authentication and Authorization • LDAP, OAuth2, SAML 2.0
- 12. Spinnaker vs/and Jenkins • Thoughtworks Tech Radar: • Jenkins - NOV 2016 HOLD • https://www.thoughtworks.com/radar/tools/jenkins-as-a-deployment- pipeline • Spinnaker - NOV 2017 ASSESS • https://www.thoughtworks.com/radar/tools/spinnaker • Leverages Jenkins CI capabilities
- 13. Spinnaker Architecture - Spring Boot microservices - HTML5 UI - S3/Redis/Cassandra storage - Install: HA or Single machine - Configuration: Halyard - General One-off settings - Accounts - Kubernetes - Jenkins integration
- 14. Language Cloud Service Pipeline • Live Walkthrough
- 15. Spinnaker Concepts and Kubernetes Spinnaker Kubernetes Account = Deployment Target + Credentials Cluster + Credentials Server-Group = Deployment artifact + configuration Replica Set or Deployment Instance = Runtime Pod Cluster = Logical grouping of Server-Groups Application = Logical grouping of Cluster Load Balancer = Balance Traffic in Server Groups Service Security Group = Network Firewall
- 16. Pipeline Definition Management • UI • Edit as JSON • Templates • Best practice: Pipeline Version control (GIT) + Spinnaker API
- 17. Spinnaker - Issues • Subset of options not available in the UI Use ”Edit as Json” • UI – does not reflect actual User permissions • Halyard – some config files require manual post edits • Rollback does not work with Kubernetes Deployment (work around: change direct in K8s)
- 18. Spinnaker New Features • Kayenta – Automated Analysis for Canary Deployment • Kubernetes Driver V2 (alpha)
- 19. DevOps - Observations • Spinnaker = DevOps enabler • Common tools and terminology • Self Service for Devs for managing Applications
- 20. Resources • www.spinnaker.io • https://www.spinnaker.io/reference/api/ • https://blog.spinnaker.io • blog.armory.io • https://martinfowler.com/articles/microservice-testing
- 21. Thank you
Editor's Notes
- #8: LC Jenkins - https://cluj-jenkins02.global.sdl.corp:8443/view/LC-AGW-PIPE/
- #14: CloudDriver – talks to Cloud Providers Orca – the orchestration engine Front50 – persistence Igor – Ci integrationsl pipeline triggers Fiat – Authorization
- #15: - Deployment stage: Time restrictions, Spel, - Artifact version travels through the pipeline LC Scripts: live, readiness, presto Health – Spring Boot health
- #16: - Account: configured with Halyard hal config - Server-Group - We use Deployment