[Kubernetes 배우기] Section 5. Kubernetes Concepts - Pods, ReplicaSets, Deployments

Kubernetes for the Absolute Beginners - Hands-on | Udemy

 

Pods

The Kubernetes pod definition file contains 4 top-level fields:

• apiVersion
  • the version of Kubernetes version used
• kind
  • the kind of service that is created
    • ex) Pod, Service, ReplicaSet, Deployment
• metadata
  • data of the object such as name and labels
    • under label property, we can set whatever key-value pair we want
• spec
  • where containers are defined along with their image and name

kubectl apply -f [name of the file].yaml: creates pod according to the YAML file in the argument

 

Replication Controllers and ReplicaSets

 

Replication Controller: allows running of multiple pods of the same image for high availability

• ensures a specified number of pods are running at a time
• can span across multiple nodes to scale the application
• in the spec section of the YAML file
  • template: The YAML file of the pod is copied
  • replicas: the number of replicas is specified

ReplicaSets: newer technology that is replacing replication controllers

• apiVersion needs to be specified as apps/v1
• spec section is almost the same except that there is one more field to it
  • selector: monitors pods with specified labels unlike replication controllers
    • so if pods of the same labels are already running when ReplicaSets are created, it does not create a new one until one fails

Scaling ReplicaSet:

1. change the replica filed of the YAML file and run kubectl replace -f [filename]
2. use kubectl scale --replicas=[number of replicas] -f [filename] command

 

Deployment

Rolling Updates: Upgrading instances one by one to maintain availability

• two ReplicaSets

Deployment: wraps around ReplicaSet to update the instance seamlessly and control changes

• The definition file is almost the same as ReplicaSet except that the section for kind is changed to Deployment.

kubectl get all: lists all the objects created in the Kubernetes such as pods, ReplicaSets, deployments

 

Rollout is triggered when a new deployment is created.

• the revision number increases when there is a newer version
  • this allows us to keep track of the revision history

kubectl rollout status [deployment name]: see the current status of the rollout

kubectl rollout history [deployment name]: see the history of the deployment

 

Deployment strategy:

1. (Bad) Destroying all the instances and creating updated new instance
   1. (Good) Taking down one instance and bringing up a new instance for availability (Rolling Update)

 

kubectl apply -f [YAML file name]: updates the current running instance to a newer version

kubectl rollout undo [deployment name]: goes back to older format by destroying the current instance