Kubernetes: Manageable Microservices

Talk given at Javascript Iceland 24 September 2015.

Introduction to kubernetes

Transcript

1. Manageable Microservices Kubernetes Javascript Iceland. Obviously. 24 Sept 2015
2. Alex Couper Developer @ Azazo @couperalex Confessions: My JS ~= my Icelandic Vim skills of a CEO Hello!
3. Contents 1. Containers 2. Kubernetes 3. Demo 4. Deployments
5. Containers in 2 mins • Define your application • Package all of the dependencies (from OS upwards) • Define what ports are exposed • Single, agreed format to handle in deployment for all your services. • (Plus some significant cost benefits)
6. # # Elasticsearch Dockerfile # https://github.com/dockerfile/elasticsearch # # Pull base image. FROM dockerfile/java:oracle-java8 # Install Elasticsearch. RUN cd / && \ wget https://elasticsearch.org/download/elasticsearch.tar.gz \ && tar xvzf elasticsearch.tar.gz # Define default command. CMD ["/elasticsearch/bin/elasticsearch"] # Expose ports. # - 9200: HTTP # - 9300: transport EXPOSE 9200 EXPOSE 9300 An example docker image
7. The Monolith to Microservices The Monolith Component 1 Component 2 Component 3 Component 2
8. Problems you soon hit • How can I make sure I’ve got the right things deployed? • How can containers find each other when each layer is being switched out all the time? (discovery) • Was all this mess of containers worth it? • GIVE ME BACK MY MONOLITH!
9. Kubernetes
10. Kubernetes • One of many ways to solve the orchestration/discovery problem. • ~an open source equivalent of what Google uses to run all its software. • “Manage a cluster of linux containers as a single system”
11. Basic Architecture For more detail see http://kubernetes.io/v1.0/docs/design/architecture.html Master Node 1 Node 2 Node 3 Node 4 Kubernetes cluster
12. Basic Architecture For more detail see http://kubernetes.io/v1.0/docs/design/architecture.html Master Node 1 Node 2 Node 3 Node 4 Kubernetes cluster API Calls Containers run here
13. Key components • Pods • Replication controllers • Services
14. Pods • 1 or more container • Within a pod: ◦ Share network/port space ◦ Share access to disk specified • Each container has a “liveness probe” • Changeable labels Container 1 Container 2 Disk
15. Replication Controllers • Responsible for checking that the state matches what you’ve dictated. ◦ Create containers if they are missing ◦ Remove containers if there are too many • Used for deployments - rolling updates, canary deploys.
16. Services • How can our containers talk to each other? How do they know where to find each other? • A service targets a set of pods by label • Its name is available via DNS to all nodes in the cluster.
17. Scheduler Kubernetes picks which nodes to run your pods on for you - based on resources and labels. If you have a db that needs ssd, you can say the machine needs ssd - and it’ll only run that pod on such a machine.
18. Demo: KubeShark • Apache • Django App • Postgres https://github.com/alexcouper/kubeshark
19. Demo time
20. Rolling Update App 0.4 App 0.4 RC-App Service: App
21. Rolling Update App 0.4 App 0.4 RC-App Service: App RC-App-2
22. Rolling Update App 0.4 App 0.4 RC-App Service: App App 0.6 RC-App-2
23. Rolling Update App 0.4 RC-App Service: App App 0.6 RC-App-2
24. Rolling Update App 0.4 RC-App Service: App App 0.6 App 0.6 RC-App-2
25. Rolling Update Service: App App 0.6 App 0.6 RC-App
26. Rolling Update Demo
27. Credits/References • Mandy Waite https://twitter.com/tekgrrl • Landing Page • Excellent article on microservice trade-offs by Martin Fowler • Kubernetes documentation • Google Container Engine documentation • https://github.com/alexcouper/kubeshark Experienced/interested in this stuff? Let’s chat! Slides will be published on speakerdeck. @couperalex
28. The End