Tech basics series : Containers , Microservices & Kubernetes - Part 3

 In the third part of our tech basics series on Containers, Microservices & Kubernetes, we will talk about Pods, ReplicaSets, and Replication controllers. If you are new here, do check out Part 1 and Part 2 of this blog series first!!

What are Pods?

Pods are the smallest object you can create in Kubernetes that encapsulates containers. Imagine a single node K8s clusters running a single pod. When the application needs to scale, you create additional pods of the same application. The pods can also be distributed across multiple nodes in a cluster. Usually, the relationship between pods and containers is 1:1, but it is not mandatory. There is a case of a side car  container as well, which could be helping the main application and included in the same pod. Every time a new pod of the application is created, both the main container and sidecar container are created together. They share the same network and storage and can connect to each other as localhost. 

Basic commands to manage pods

1. Command to run a container in K8s :

kubectl run <pod name> --image <image name>>

Image can be fetched from any container registry , for eg: Docker hub. Image name has to match the name in the registry

2. To view status of running pods:

kubectl get pods

3. To get additional details of the pods:

kubectl describe pod <name of the pod>

3. Command to view details of the node where the pod is deployed, its IP address etc:

kubectl get pods -o  wide

4. Sample yaml file for a pod: https://github.com/cloudyknots/Kubernetessample/blob/main/poddefinition.yaml

You can use the below command to create the pod

kubectl create -f pod-definition.yml

ReplicaSets & Replicas

If there is only one instance of the application running on a pod, the application will become unavailable if the pod fails. To prevent users from losing access, you can have more than one application instance running in multiple pods. This is enabled by the replication controller. Even in the case of a single pod, the replication controller can bring back the pod if it fails. The replication controller spans multiple nodes in a cluster. When the demand for application increases, the replication controller can deploy the pods across multiple nodes as well to scale the application. ReplicaSets is the new version of the replication controller and is used in current versions of Kubernetes

The details of the replica set are defined in the spec of the replication controller yaml file by adding a section called template, please see sample yaml file here to create a ReplicaSet. You can see that we have additionally specified the number of replicas and a selector in the yaml file. 

kubectl create -f replicaset.yaml

Kubectl get replicaset

Replicaset will monitor the pods and recreate them if they fails. For replicaset to do that we need to create labels for pods while creating the poids. Now we can use the "matchLabels" filters to identify which pods should be monitored by ReplicaSets. Even if you have created additional pods not included in the yaml file definition of the replicaset, it can monitor them based on the labels of the pods and ensure that the number of replicas of the pod(3, in this example) is always running.

If you want to scale your application, you can update the number of replicas in the definition file and run the following command

kubectl replace -f replicaset.yaml

Or you can use the kubectl scale command

or you can increase the replicas on the fly using the following command

kubectl scale --replicas=6 replicaset.yaml