Deploy MongoDB Replica Set on K8s

So, I set a new k8s cluster on DigitalOcean recently, and I want to share the way I deploy my MongoDB Replica Set on the k8s cluster. Let’s start it.

Storage Class

I use Longhorn instead of the default do-block-storage provided by DigitalOcean as the default. I set up the longhorn using the helm chart. Since I use Rancher to manage my cluster, I install Longhorn from the App menu and choose Longhorn with the default value. super simple!

Cluster IP Service

To create a statefulset, we will need to provide the serviceName so, let’s create a ClusterIP service for our MongoDB. The mongodb-clusterip.yaml file looks like this.

apiVersion: v1 kind: Service metadata: annotations: field.cattle.io/description: MongoDB Cluster IP name: mongodb-clusterip namespace: default spec: ports: - name: mongo port: 27017 protocol: TCP targetPort: 27017 selector: app: mongodb type: ClusterIP 

Then execute the kubectl apply command
kubectl apply -f mongodb-clusterip.yaml

Secret

We will save the mongodb root user and mongodb root password in the Secret, so we need to create a new Secret for our MongoDB.

The mongodb-secret.yaml file looks like this

apiVersion: v1 data: password: base6StringVersionOfpassword user: base6StringVersionOfuser kind: Secret metadata: name: mongodb-secret namespace: default 

Then execute the kubectl apply command
kubectl apply -f mongodb-secret.yaml

StatefulSet

Now the last thing that we need to execute is the statefulset for our MongoDB. The mongodb-statefulset.yaml file looks like this

apiVersion: apps/v1 kind: StatefulSet metadata: name: mongodb namespace: default spec: replicas: 2 selector: matchLabels: app: mongodb serviceName: mongodb-clusterip template: metadata: namespace: default labels: app: mongodb spec: containers: - args: - '--dbpath' - /data/db command: - mongod - '--bind_ip_all' - '--replSet' - rs0 env: - name: MONGO_INITDB_ROOT_USERNAME valueFrom: secretKeyRef: key: user name: mongodb-secret - name: MONGO_INITDB_ROOT_PASSWORD valueFrom: secretKeyRef: key: password name: mongodb-secret image: mongo:6.0.11 imagePullPolicy: IfNotPresent name: mongodb-c volumeMounts: - mountPath: /data/db name: mongodb-pvc volumeClaimTemplates: - apiVersion: v1 kind: PersistentVolumeClaim metadata: name: mongodb-pvc spec: accessModes: - ReadWriteOnce resources: requests: storage: 8Gi storageClassName: longhorn 

Then execute the kubectl apply command
kubectl apply -f mongodb-statefulset.yaml

you can change the replica number with your own. I use 2 here.

After a few minutes, you should see the statefulset created. You can see it in your Rancher or using the kubectl get command.

Let’s make sure that the Mongodb is running using kubectl exec command below.

kubectl --namespace=default exec -ti mongodb-0 -- mongosh
Then you should get the mongosh prompt opened. For the second replica, you can do the same by changing the mongodb-0 with mongodb-1 instead.

Configure Replica Set

To configure MongoDB replica set. First we need to open the mongodb-0 (first pod of the MongoDb) using above command.

Then we use the rs.initiate() command as below

rs.initiate( { _id: "rs0", members: [ { _id: 0, host: "mongodb-0.mongodb-clusterip.default.svc.cluster.local:27017" }, { _id: 1, host: "mongodb-1.mongodb-clusterip.default.svc.cluster.local:27017" } ] } ) 

if you notice that the pattern is <pod>.<service>.<namespace>.svc.cluster.local:port

Then try to add some data to the test db using the query
db.test.insertOne({ testFromMaster: "true" })

Now, exit from the master replica and open the mongosh in the secondary replica using the same command

kubectl --namespace=default exec -ti mongodb-1 -- mongosh
You will notice that the prompt is displaying the secondary. To start reading the data changes from the primary replica, we need to run the command

db.getMongo().setReadPref("primaryPreferred")

Then we can test to query the test data that we create before using the query

db.test.find({})

Now, you should see the secondary also has the test data.

That’s it. hopefully, you found this useful. Cheers.