DevOps Classroomnotes 30/Apr/2023

Storage Solutions in K8s

Stateful applications store data locally. In Containers the data created locally will be lost once you delete it. So to solve this in docker we have used volumes. Volumes have a lifecycle which has no relation to container lifecycle (refer docker containers, image layers, volumes)
IN k8s we are running docker containers, k8s is an orchestration solutions.
Lets see what are options for storage provisioning in k8s Refer Here
The most widely used storage types
- Volumes
- Persistent Volumes
  - Storage Classes
  - Persistent Volume Claims
Volumes Refer Here

Volumes

Volumes can be mounted to containers and they have lifetime equivalent to Pods.
The types of Volumes Refer Here
The types
- storage on cloud
  - ebs
  - azure disk
  - efs
  - azure file
  - gcs
- empty dir
- hostPath
Lets create a manifest with mysql-pod with volume

---
apiVersion: v1
kind: Pod
metadata:
  name: mysql-vol
  labels:
    app: mysql
    layer: db
spec:
  containers:
    - name: mysql
      image: mysql:8
      ports:
        - containerPort: 3306
      volumeMounts:
        - name: test-volume
          mountPath: /var/lib/mysql
  volumes:
    - name: test-volume
      emptyDir:
        sizeLimit: 100Mi

Persistent Volumes

These volumes will have a lifetime different than Pod i.e. they exist even after pod is dead.
Types of PVC Refer Here
Persistent Volumes in Azure Refer Here
Storage classes in Azure Refer Here
Storage classes in AWS Refer Here

Lets create a mysql-pod with dynamic volume provisioing

Get storage classes
Lets create a Persistent Volume Claim and a mysql pod claiming pvc
Refer Here for the initial manifests written by us
Now lets modify the spec to add mysql related environment variables
Refer Here for the mysql env values
INsert some data into mysql
now delete pod and recreate and check for the data created
For static provisioning of volume Refer Here
If you want to scale Pods we can use
- Replica Sets
- Deployments

StatefulSets

Statefulset is like deployment with replicas. But each pod gets its own volume.
Stateful Set is for stateful applications
When we create replicas in Stateful Set we get predictable names
Refer Here for official docs
We can access individual pod, by creating headless service and by using ...svc.cluster.local
Lets experiment with stateful sets and create nginx pods
Refer Here for the manifests
Naming is predictable in stateful sets
Now create a different html page for every pod
Send a curl request from any alpine pod <pod-name>.<headless.svc-name>.<namespace>.svc.cluster.local