Skip to main content

Deployments

Deployments are the most common workload controller for running stateless applications. They make sure your application always runs the desired number of Pods - automatically handling creation, scaling, updates, and recovery.

Instead of managing Pods manually, Deployments let Kubernetes:

  • Run multiple identical Pods for reliability and load distribution
  • Scale automatically by adjusting replica counts
  • Recover failed Pods without manual intervention
  • Perform rolling updates without downtime
  • Rollback easily to previous versions when needed

Creating a Deployment​

Let's deploy the retail store UI using a deployment:

~/environment/eks-workshop/base-application/ui/deployment.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
  name: ui
  labels:
    app.kubernetes.io/created-by: eks-workshop
    app.kubernetes.io/type: app
spec:
  replicas: 1
  selector:
    matchLabels:
      app.kubernetes.io/name: ui
      app.kubernetes.io/instance: ui
      app.kubernetes.io/component: service
  template:
    metadata:
      annotations:
        prometheus.io/path: /actuator/prometheus
        prometheus.io/port: "8080"
        prometheus.io/scrape: "true"
      labels:
        app.kubernetes.io/name: ui
        app.kubernetes.io/instance: ui
        app.kubernetes.io/component: service
        app.kubernetes.io/created-by: eks-workshop
    spec:
      serviceAccountName: ui
      securityContext:
        fsGroup: 1000
      containers:
        - name: ui
          env:
            - name: JAVA_OPTS
              value: -XX:MaxRAMPercentage=75.0 -Djava.security.egd=file:/dev/urandom
            - name: METADATA_KUBERNETES_POD_NAME
              valueFrom:
                fieldRef:
                  fieldPath: metadata.name
            - name: METADATA_KUBERNETES_NAMESPACE
              valueFrom:
                fieldRef:
                  fieldPath: metadata.namespace
            - name: METADATA_KUBERNETES_NODE_NAME
              valueFrom:
                fieldRef:
                  fieldPath: spec.nodeName
          envFrom:
            - configMapRef:
                name: ui
          securityContext:
            capabilities:
              add:
                - NET_BIND_SERVICE
              drop:
                - ALL
            readOnlyRootFilesystem: true
            runAsNonRoot: true
            runAsUser: 1000
          image: "public.ecr.aws/aws-containers/retail-store-sample-ui:1.2.1"
          imagePullPolicy: IfNotPresent
          ports:
            - name: http
              containerPort: 8080
              protocol: TCP
          livenessProbe:
            httpGet:
              path: /actuator/health/liveness
              port: 8080
            initialDelaySeconds: 45
            periodSeconds: 20
          resources:
            limits:
              memory: 1.5Gi
            requests:
              cpu: 250m
              memory: 1.5Gi
          volumeMounts:
            - mountPath: /tmp
              name: tmp-volume
      volumes:
        - name: tmp-volume
          emptyDir:
            medium: Memory
A

kind: Deployment: Defines a Deployment controller

B

metadata.name: Name of the Deployment (ui)

C

spec.replicas: Desired number of pods (1 in this example)

D

spec.selector: Labels used to find managed Pods

E

spec.template: Pod template defining what each pod should looks like

The deployment ensures that the actual Pods always match this template.

Apply the Deployments:

~$kubectl apply -k ~/environment/eks-workshop/base-application/ui


Inspecting Deployment​


Check deployment status:


~$kubectl get deployment -n ui
NAME   READY   UP-TO-DATE   AVAILABLE   AGE
ui     1/1     1            1           30s


List the Pods created by the Deployment:


~$kubectl get pods -n ui
NAME                  READY   STATUS    RESTARTS   AGE
ui-6d5bb7b9c8-xyz12   1/1     Running   0          30s


Get detailed information:


~$kubectl describe deployment -n ui ui


Scaling Deployment​


Scale up to 5 replicas:


~$kubectl scale deployment -n ui ui --replicas=5
~$kubectl get pods -n ui
NAME                  READY   STATUS    RESTARTS   AGE
ui-6d5bb7b9c8-abc12   1/1     Running   0          2m
ui-6d5bb7b9c8-def34   1/1     Running   0          12s
ui-6d5bb7b9c8-ghi56   1/1     Running   0          12s
ui-6d5bb7b9c8-arx97   1/1     Running   0          10s
ui-6d5bb7b9c8-uiv85   1/1     Running   0          10s


info
Kubernetes automatically spreads these Pods across available worker nodes for high availability.


Scale back down to 3 replicas:


~$kubectl scale deployment -n ui ui --replicas=3
~$kubectl get pods -n ui
NAME                  READY   STATUS    RESTARTS   AGE
ui-6d5bb7b9c8-abc12   1/1     Running   0          2m
ui-6d5bb7b9c8-def34   1/1     Running   0          12s
ui-6d5bb7b9c8-ghi56   1/1     Running   0          12s


Rolling Updates and Rollbacks​


You can update a Deployment by changing the image version:


~$kubectl set image deployment/ui ui=public.ecr.aws/aws-containers/retail-store-sample-ui:v2 -n ui
~$kubectl get pods -n ui
NAME                  READY   STATUS         RESTARTS   AGE
ui-5989474687-5gcbt   1/1     Running        0          13m
ui-5989474687-dhk6q   1/1     Running        0          14s
ui-5989474687-dw8x8   1/1     Running        0          14s
ui-7c65b44b7c-znm9c   0/1     ErrImagePull   0          7s




You'll see a new pod created but with status ErrImagePull.




Now let's rollback the change


~$kubectl rollout undo deployment/ui -n ui
~$kubectl get pods -n ui
NAME                  READY   STATUS         RESTARTS   AGE
ui-5989474687-5gcbt   1/1     Running        0          13m
ui-5989474687-dhk6q   1/1     Running        0          14s
ui-5989474687-dw8x8   1/1     Running        0          14s


Rolling updates let you update your application gradually without downtime, while Kubernetes ensures new Pods match the desired state.
If something goes wrong — like an invalid image — you can rollback safely to the previous working version, keeping your application available and stable.


This demonstrates how Deployments simplify application updates, maintain availability, and reduce risk in production environments.


Key Points to Remember​


    

  • Deployments manage multiple identical pods automatically
    • 

  • Use deployments instead of creating pods directly in production
    • 

  • Scaling is as simple as changing the replica count
    • 

  • Pod names include the deployment name plus random suffixes
    • 

  • Deployments are perfect for stateless applications like web apps and APIs
    •