Kubernetify All The Things!

Good tooling today exists to spin up containerized environments with Docker on a single host and tear down just as fast. The difficulty is when trying to deploy across two or more servers. Kubernetes is an open source project to manage a cluster of Linux containers as a single system and solves just that problem. This talk will outline all of the components of the Kubernetes system leave attendees with enough knowledge to spin up their own cluster and deploy at scale. #codemash

Transcript

1. KUBERNETIFY ALL THE THINGS @STEVESLOKA
2. OVERVIEW ▸ Overview ▸ Background of Kubernetes ▸ k8s Level Concepts ▸ DEMO!!! (On a Pi cluster! —————————————————> ▸ Even More Demo ▸ More Demo ———————————————> ▸ Summary
3. SOFTWARE ARCHITECT / ENGINEER ABOUT ME
4. KUBERNETES ▸ open-source platform for automating deployment, scaling, and operations of application containers ▸ Servers 2-N ▸ Open Sourced by Google 2014 ▸ github.com/kubernetes/kubernetes ▸ Written in Go ▸ #GIFEE (Google-like Infrastructure for Everyone Else)
5. GOOGLE EXPERIENCE ‣ Google has been using containers for over a decade ‣ 2 BILLON Per Week! (With a “B”) ‣ Everything at Google runs in a container
6. “HELMSMAN OF A SHIP” ▸ lean: lightweight, simple, accessible ▸ extensible: modular, pluggable, hookable, composable, toolable ▸ portable: public, private, hybrid, multi cloud ▸ self-healing: auto-placement, auto-restart, auto- replication
7. COMMUNITY KUBERNETES
8. ARCHITECTURE Master: Node(s): ETCD API SERVER CONTROLLER MANAGER KUBELET PROXY SCHEDULER CONTAINER ENGINE CADVISOR SOFTWARE DEFINED NETWORK ADDONS
9. PAM WEBO2 DWIGHT DB01
10. APPLICATION CONTAINERS KUBERNETES CLUSTER OF MACHINES
11. HIGH LEVEL CONCEPTS ▸node ▸pod ▸scheduler ▸labels ▸replication controller ▸service
12. NODE
13. NODE RESPONSIBILITIES 1.Runs containers ‣container engine (docker / rkt) ‣kubelet 2.Proxies service requests ‣kubeproxy
14. POD
15. REPRESENTS A LOGICAL APPLICATION CONTAINER A CONTAINER B ‣Shared ‣namespaces ‣volumes
16. POD EXAMPLE: REST API (CONTAINER) LOGGER (CONTAINER) Logs
17. SCHEDULER
18. SCHEDULES PODS TO RUN ON NODES ▸global ▸best fit chosen based on pod requirements ▸allows for over-commit ▸pluggable
19. LABELS
20. LABELS EXAMPLE: TYPE=FRONTEND
 ENV=PRODUCTION TYPE=FRONTEND
 ENV=STAGING TYPE=BACKEND
 ENV=PRODUCTION ENG=STEVE
 ENV=STAGING
21. TYPE=FRONTEND TYPE=FRONTEND
 ENV=PRODUCTION TYPE=FRONTEND
 ENV=STAGING TYPE=BACKEND
 ENV=PRODUCTION ENG=STEVE
 ENV=STAGING
22. ENV=PRODUCTION TYPE=BACKEND
 ENV=PRODUCTION ENG=STEVE
 ENV=STAGING TYPE=FRONTEND
 ENV=PRODUCTION TYPE=FRONTEND
 ENV=STAGING
23. ENG=STEVE TYPE=FRONTEND
 ENV=PRODUCTION TYPE=FRONTEND
 ENV=STAGING TYPE=BACKEND
 ENV=PRODUCTION ENG=STEVE
 ENV=STAGING
24. DESIRED STATE
25. CREATE PODS Desired: Actual: POD POD POD POD POD POD ACTION IS TO CREATE 3 PODS ——>
26. DELETE PODS Desired: Actual: POD POD POD POD ACTION IS TO DELETE 2 PODS ——>
27. HEALTH CHECK Desired: Actual: POD POD POD POD ACTION IS TO CREATE 2 PODS ——> POD POD
29. REPLICATION CONTROLLER
30. REPLICATION CONTROLLER ▸Pod template ▸Docker image(s) ▸Labels ▸CPU / Memory ▸Replicas ▸Volume Mounts ▸Define in JSON || YAML
31. REPLICATION CONTROLLER EXAMPLE
32. REPLICATION CONTROLLERS VOLUMES ▸ emptyDir ▸ hostPath ▸ gcePersistentDisk ▸ awsElasticBlockStore ▸ nfs ▸ persistentVolumeClaim ▸ flocker ▸ glusterfs ▸ rbd ▸ gitRepo ▸ secret ▸ iscsi
33. SERVICES
34. KUBERNETES SAMPLE NGINX NGINX LOAD BALANCER API API API API LOAD BALANCER NGINX DB LOAD BALANCER K8S SERVICE K8S SERVICE K8S SERVICE NGINX.NAMESPACE.SVC.K8S.LOCAL API.NAMESPACE.SVC.K8S.LOCAL DB.NAMESPACE.SVC.K8S.LOCAL
35. KUBERNETES PODS ARE MORTAL. THEY ARE BORN AND THEY DIE, AND THEY ARE NOT RESURRECTED.
36. SERVICE DISCOVERY FOR PODS ▸proxy runs on each node ▸virtual IP per service ▸dns address ▸[serviceName].[namespace].svc.k8s.local ▸dynamic “pods” based on label queries ▸pods are auto-injected with environment variables
39. SERVICE TYPES ▸NodePort: Allocated port (range 30000-32767) ▸LoadBalancer: Provision loadbalancer on cloud provider ▸External IP’s
40. NAMESPACES
41. PARTITION RESOURCES ▸A mechanism to partition resources created by users into a logically named group ▸Allows for work to be done in isolation ▸Each namespace is given its own: 1.resources (pods, services, replication controllers, etc.) 2.policies (who can or cannot perform actions in their namespace) 3.constraints (this namespace is allowed this much quota, etc.)
42. HORIZONTAL POD AUTOSCALER
43. AUTOSCALER ▸allows the number of pods in a replication controller or deployment to scale automatically based on observed CPU utilization
44. DAEMON SET
45. WHAT IS A DAEMON SET? ▸Ensure that all (or some) nodes run a copy of a pod ▸Examples: ▸running a cluster storage daemon (glusterd, ceph) ▸running a logs collection daemon on every node (fluentd or logstash) ▸running a node monitoring daemon on every node (Prometheus Node Exporter, collectd, New Relic agent, or Ganglia gmond)
46. JOBS
47. WHAT IS A JOB? ▸ Creates one or more pods and ensures that a specified number of them successfully terminate
49. THANK YOU! @stevesloka steve@stevesloka.com github.com/stevesloka github.com/upmc-enterprises enterprises.upmc.com