A rolling update replaces application Pods incrementally, upgrading them one at a time to minimize user downtime. Kubernetes Deployments coordinate the update while maintaining the requested replica count. section 2.3.1 section 2.3.2
A Deployment uses RollingUpdate by default. Its configuration can limit unavailable replicas and allow additional replicas during the update; readiness and liveness probes provide health signals for container availability and process status. section 2.3.1 section 2.3.2
A rolling update is triggered when the Deployment Pod template changes:
- Kubernetes creates or updates ReplicaSets under the Deployment and progressively replaces Pods using the new specification. section 2.3.3
- Changing a setting inside spec.template.spec, such as an environment variable, starts the rolling update mechanism. section 2.3.3
- The alternative Recreate strategy stops all existing Pods before creating replacements, unlike the incremental RollingUpdate strategy. section 2.3.2
Updating the Conference Frontend
Scale the Frontend Deployment to two replicas with kubectl scale --replicas=2 deployments/conference-frontend-deployment, then enable FEATURE_DEBUG_ENABLED with kubectl set env deployment/conference-frontend-deployment FEATURE_DEBUG_ENABLED=true. Kubernetes creates a new ReplicaSet and progressively updates the Frontend Pods; refreshing the application can show responses from different replicas. section 2.3.3
Listing 2.3 Describing a deployment to see its details
> kubectl describe deploy conference-frontend-deployment
Name: conference-frontend-deployment
Namespace: default
CreationTimestamp: Tue, 27 Jun 2023 08:21:21 +0100
Labels: app.kubernetes.io/managed-by=Helm
Annotations: deployment.kubernetes.io/revision: 1
meta.helm.sh/release-name: conference
meta.helm.sh/release-namespace: default
Selector: app=frontend
Replicas: 1 desired | 1 updated | 1 total | 1 available #1
StrategyType: RollingUpdate
MinReadySeconds: 0
RollingUpdateStrategy: 25% max unavailable, 25% max surge
Pod Template:
Labels: app=frontend
Containers:
frontend:
Image: salaboy/frontend-go... #2
Port: 8080/TCP
Host Port: 0/TCP
...
...
Environment: #3
AGENDA_SERVICE_URL: agenda-service.default.svc.cluster.local
C4P_SERVICE_URL: c4p-service.default.svc.cluster.local
NOTIFICATIONS_SERVICE_URL: notifications-service.default.svc.cluster.local
KAFKA_URL: conference-kafka.default.svc.cluster.local
POD_NODENAME: (v1:spec.nodeName)
POD_NAME: (v1:metadata.name)
POD_NAMESPACE: (v1:metadata.namespace)
POD_IP: (v1:status.podIP)
POD_SERVICE_ACCOUNT: (v1:spec.serviceAccountName)
Mounts: <none>
Volumes: <none>
Conditions:
Type Status Reason
---- ------ ------
Available True MinimumReplicasAvailable
Progressing True NewReplicaSetAvailable
OldReplicaSets: <none>
NewReplicaSet: conference-frontend-deployment-<ID> (1/1 replicas created)
Events: #4
Type Reason Age From Message
---- ------ ---- ---- -------
Normal ScalingReplicaSet 48m deployment-controller Scaled up replica set conference-frontend-deployment-59d988899 to 1
Inspect StrategyType, replica counts, max unavailable, max surge, and the Pod template with kubectl describe deploy. section 2.3.2
