Labels are the metadata that contains identifiable information to the kubernetes objects. These are basically key-value pairs attached to the objects such as pods
Each key must be unique for an object
Labels would appear in metadata section and the yaml would appear as
metadata:
labels:
key1: value1
key2: value2
Constraints for Labels:
Label prefix: This is optional and this must be a DNS subdomain. This cannot be longer thatn 253 characters & cannot contain spaces. The label prefix is always followed by a forwar slash Eg: directdevops.blog/. If no prefix is used the label key is assumed to be private to the user. label prefixes specific to k8s core system kubernetes.io/ and k8s.io/
Label name: The label name is required and can be upto 63 characters long.
Value of the key in k8s is label-prefix/label-name or label-name
Scenario: Lets create a ngnix pod with labels organized by team/project
Refer Here for the k8s manifest and apply the yaml file
Add labels to k8s object after creation
To change the labels
kubectl label --overwrite pod label-demo1 server=nginx-webserver
In this selector we have three kinds of operators = == !=
Lets try to get all the pods with the label environment is equal to developer
SET-BASED Selectors
In this we have 3 kinds of operators in notin exists
Lets try some examples
Annotations
Using labels we can add metadata which can be later used to filter/select objects
Annotations on the other hand have fewer constraints, however we cannot filter or select objects by annotations
Annotations are generally used by tools or users to get subjective information regarding k8s object.
Lets look at one annotation example Refer Here for the manifest
Like what we have done in labels we can add/modify/delete annotations from kubectl
kubectl annotate pod <podname> <annotate_key>=<annotate_value>
kubectl annotate --overwrite pod <podname> <annotate_key>=<annotate_value>
kubectl annotate pod <podname> <annotate_key>-
Kubernetes Controllers
Pod is the workload in the k8s cluster
When we deploy our application in production,
we might need more than one replica of pod
We can balance load across multiple pods so that one pod is not overloaded
k8s supports different controllers that you can use for replications and maintaining state. The controllers we have are
ReplicaSets
Deployments
DaemonSets
StatefulSets
Jobs
A controller is an object in k8s that ensures application runs in the desired state for its entire runtime
ReplicaSet
A ReplicaSet is a k8s controller that keeps a certain number of Pods running at any given time
ReplicaSet acts a supervisor for multiple Pods across different nodes in a k8s cluster. A ReplicaSet will terminate or start new Pods to match the configuration specified in Replicaset Template.
Create a sample replica set Refer Here for the yaml and apply
lets delete one pod manually and see what happens
Now lets try to delete the rs
Now lets create a pod with app: httpd label Refer Here for yaml and apply this
Now lets apply the replica set which needs 2 replicas
replica set didnt create 2 pods rather it created one pod as there was one pod which matched the label selctor mentioned in replica set specification
Manually Scaling the replicas
change the spec
executing the command
Note: command to create token and display the join command in k8s custer
kubeadm token create --print-join-command --ttl=0
kubeadm token list
Exercise: Try to create replicaset with 3 replicas
