Kubernetes is an open-source system for automating deployment, scaling, and management of containerized applications. It coordinates activities across a cluster of machines by defining basic building blocks like pods (which contain containers), replication controllers (which ensure a specified number of pods are running), and services (which define logical groups of pods). Kubernetes provides tools for running applications locally on a single node as well as managing resources in the cluster, including creating, deleting, viewing, and updating resources from configuration files.

1. Kubernetes -
introduction
Tomasz Piekarczyk

2. Kubernetes (k8s):
is an open-source system for automating
deployment, scaling, and management of
containerized applications

3. Why would you care?
The pitch:
“Kuberentes makes it easier to organize and schedule your application across a fleet of
machines”
Specifically by:
• Horizontal scaling
• Load distribution
• Service discovery
• Health monitoring
• Deploying new versions, rollbacks
• Handling hardware failure

4. High level architecture
Master - coordinates all activities in the cluster
Nodes:
• virtual or physical machines
• actual workers
• runs processes:
• kubelet
• container runtime

5. Basic building blocks
Containers
• Define single running process*
• Eg docker container
Pods
• the way of running containers in kubernetes
• basic deployable and scaling unit
• defines one or more containers
• containers are co-located on a node
• flat network structure
Nodes:
• physical worker machines
• can run multiple pods
• pods running within single node don’t know
about each other

6. Running things locally
Minikube:
• single node cluster
• running in a VM
• supports linux, windows and macOS
• mature project
https://kubernetes.io/docs/setup/minikube/
https://kubernetes.io/docs/tasks/tools/install-minikube/
https://docs.docker.com/docker-for-windows/kubernetes/
https://docs.docker.com/docker-for-mac/kubernetes/
Docker with built-in kubernetes:
• single node cluster
• running in a VM
• windows and macOS
• drag & drop installation
• bound to specific kubernetes version
• relatively new
More options listed in https://kubernetes.io

7. Managing cluster resources
Create resource from file
• kubectl create -f resource_file.yml
Change existing (or create) resource based on file
• kubectl apply -f resource_file.yml
Delete existing resource
• kubectl delete resource_type resource_name
List resources of type
• kubectl get resource_type
Edit resource on the server
• kubectl edit resource_type resource_name

8. Debugging cluster resources
Execute command on the container
• kubectl exec [-it] pod_name process_to_run
Get container logs
• kubectl logs pod_name [-c container_name]
Forward port from a pod
• kubectl port-forward pod_name local_port:remote_port
Print detailed description of a resource
• kubectl describe resource_type resource_name

9. Defining a pod
Main definition components
• api
• resource type
• metadata
• spec

10. Grouping pods - labels
• Each pod can have multiple labels
• Labels can be changed on a live pod
• Basic mechanism to execute commands on
multiple pods

11. Pod health - liveness probe
Out of the box (before the probe)
• Pod defines a restart policy
• Crashing process can be restarted by kubelet
Process can be alive but unhealthy eg:
• deadlocked
• stuck in a loop
Liveness probe is a way of telling kubelet that a process in no longer
healthy and needs restarting
3 ways of querying:
• http
• tcp
• exec
Limitations of liveness probe

12. Replica Sets
Replication Controller
Ensures desired number of pods exist by:
• scaling up or down
• running new pods when nodes fail
Definition consists of:
• label selector
• replica count
• pod template

13. Demo

14. Service
Define logical groups of pods - label selectors
Types:
• ClusterIP (default)
• LoadBalancer
• NodePort
• ExternalName
Service is constant:
• IP doesn’t change (except headless)
• each service gets a dns entry

15. Pod health - readiness probe
Similar to liveness probe:
• Defined in a pod
• 3 types supported
• similar parameters
When probe fails pod is marked as not ready
Not ready pods are removed from services

16. Service - LoadBalancer
• Allows automatic provisioning of load
balancers
• Needs to be supported by cloud provider
• Exposes external IP to the cluster

17. DEMO

18. Deployments
• Higher level resource
• Deals with updating the application
• Allows graceful rollback
• Can pause or abort a bad release
• Manages internally multiple replica sets
• Changing pod template triggers update

19. Deployments
Revision history:
• rolling back release
• switching between releases
• number of revisions can be specified
Update strategy:
• RollingUpdate
• Recreate
Detecting bad releases with minReadySeconds
Specifying update process
• MaxSurge
• MaxUnavailable

20. DEMO

21. Passing configuration
Pod definition can embed:
• environment variables
• command line parameters
This couples pod definition with environment:
• separate definitions for prod/test
• changing configuration requires changing pod

22. ConfigMap
• Lives as a separate resource
• Supports:
• environment variables:
• individual
• entire environment
• configuration files /directories
• Live reloading of config if supported by pod
• Missing ConfigMap prevents container from starting

23. ConfigMap - basic usage
Defining and passing single
environment variable

24. ConfigMap - mounting files

25. Secrets
• Similar to configMap
• Stored in memory on the nodes
(includes mounting files)
• Ability to store binary data -
base64 encoded

26. Where to go from here
Managing persistent storage:
• persistent volumes
• persistent volume claims
• storage class
Advanced routing - ingress controllers
Multiple replicas with own storage - stateful sets
Even more resources - deamon sets, jobs, cronjobs
Package management - Helm
Managing Kubernetes through API
Service mesh - Istio

27. Thank
you