DevOps Classroom Series – 10/Apr/2021
Labels and Annotations
- 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
- To Remove the label
kubectl label pod label-demo1 server-
Selecting k8s objects using label selectors
- Lets create two more pod specs Refer Here
- We can query the objects using label selector
kubectl get pods -l {label-selector}
- There are two types of label selectors
- equality based
- set-based
- Equality based selectors:
- 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
- for running game of life
- for running spring pet clinic
- for nop commerce
HybrisWorld 