Deploy a Kubernetes cluster with autoscaling

In this section, we will deploy a kubernetes cluster with the Cluster Autoscaler feature enabled. With autoscaling, Magnum automatically adds worker nodes when your workloads need more resources, and removes them again when they are no longer required, within a minimum and maximum node count that you define.

Creating an autoscaling cluster using the command line

The steps below use the OpenStack command line to enable autoscaling on the default-worker node group at cluster creation, by passing the auto_scaling_enabled, min_node_count and max_node_count labels.

1. Create a Kubernetes cluster with autoscaling enabled

   openstack coe cluster create --cluster-template \
   kubernetes-v1.35.4-ardc-syd-1-calico-v2 --master-count 1 --node-count 1 \
   --labels auto_scaling_enabled=true,min_node_count=1,max_node_count=3 \
   --keypair <project key> mycluster
   Request to create cluster cae2a382-26a5-438d-bbdc-e694249bbede accepted
   

2. Verify status of autoscaling label

   openstack coe cluster show mycluster -c labels_added
   

   +--------------+--------------------------------------------------------------------------------+
   | Field        | Value                                                                          |
   +--------------+--------------------------------------------------------------------------------+
   | labels_added | {'auto_scaling_enabled': 'true', 'min_node_count': '1', 'max_node_count': '3'} |
   +--------------+--------------------------------------------------------------------------------+
   

3. Verify that the cluster has been created successfully

   openstack coe cluster list
   

   +--------------------------------------+-----------+---------+------------+--------------+--------------------+---------------+
   | uuid                                 | name      | keypair | node_count | master_count | status             | health_status |
   +--------------------------------------+-----------+---------+------------+--------------+--------------------+---------------+
   | cae2a382-26a5-438d-bbdc-e694249bbede | mycluster | xxxxxx  |          1 |            1 | UPDATE_COMPLETE    | HEALTHY       |
   +--------------------------------------+-----------+---------+------------+--------------+--------------------+---------------+
   

Creating an autoscaling cluster using the Nectar Dashboard

You can also create the cluster using the Nectar Dashboard, as described in Creating a Cluster. The dashboard cluster-creation dialog does not enable autoscaling on the default-worker node group. To use autoscaling on any cluster, add a node group with Enable Autoscaling ticked, as described in Scaling. If you need the default-worker node group itself to autoscale, use the command line method above.

Deploying an application and testing autoscaling

1. Download the kubeconfig file and verify the cluster nodes using kubectl in accordance with the Administering a Cluster in Kubernetes tutorial

2. Deploy an application - nginx.yaml

   apiVersion: apps/v1
   kind: Deployment
   metadata:
     name: nginx-deployment
   spec:
     selector:
       matchLabels:
         app: nginx
     replicas: 3
     template:
       metadata:
         labels:
           app: nginx
       spec:
         containers:
         - name: nginx
           image: nginx:1.14.2
           ports:
           - containerPort: 80
   

   kubectl apply -f nginx.yaml
   

3. Deploy application and validate the deployment

   kubectl get deployments
   

   NAME               READY   UP-TO-DATE   AVAILABLE   AGE
   nginx-deployment   3/3     3            3           8m44s
   

   kubectl get pods
   

   NAME                               READY   STATUS    RESTARTS   AGE
   nginx-deployment-d556bf558-7phtc   1/1     Running   0          9m11s
   nginx-deployment-d556bf558-98d68   1/1     Running   0          9m11s
   nginx-deployment-d556bf558-shqt2   1/1     Running   0          9m11s
   kubectl get nodes
   NAME                                                STATUS   ROLES           AGE   VERSION
   mycluster-l76uten6iljx-control-plane-xlmq4          Ready    control-plane   42m   v1.35.4
   mycluster-l76uten6iljx-default-worker-4dqrj-cwmjq   Ready    <none>          39m   v1.35.4
   

4. Scaling up the nginx application deployment

   The number of NGINX web server replicas deployed must be carefully adjusted based on the configuration of the Kubernetes worker nodes. For example, in a cluster where each worker node is configured with 4 vCPUs and 8 GB of memory (m3.medium), a larger number of replicas may be required to saturate available resources and trigger the provisioning of additional nodes.

   kubectl scale deployment --replicas=120 nginx-deployment
   

   deployment.apps/nginx-deployment scaled
   

   kubectl get deployments
   

   NAME               READY     UP-TO-DATE   AVAILABLE   AGE
   nginx-deployment   101/120   120          101         14m
   

   kubectl get deployments
   

   NAME               READY     UP-TO-DATE   AVAILABLE   AGE
   nginx-deployment   120/120   120          120         20m
   

   os coe cluster list
   +--------------------------------------+-----------+---------+------------+--------------+-----------------+---------------+
   | uuid                                 | name      | keypair | node_count | master_count | status          | health_status |
   +--------------------------------------+-----------+---------+------------+--------------+-----------------+---------------+
   | cae2a382-26a5-438d-bbdc-e694249bbede | mycluster | xxxxxx  |          2 |            1 | CREATE_COMPLETE | HEALTHY       |
   +--------------------------------------+-----------+---------+------------+--------------+-----------------+---------------+
   

   kubectl get nodes
   

   NAME                                                STATUS   ROLES           AGE   VERSION
   mycluster-l76uten6iljx-control-plane-xlmq4          Ready    control-plane   51m   v1.35.4
   mycluster-l76uten6iljx-default-worker-4dqrj-br8p9   Ready    <none>          66s   v1.35.4
   mycluster-l76uten6iljx-default-worker-4dqrj-cwmjq   Ready    <none>          48m   v1.35.4
   

5. Scaling down the nginx application deployment

   kubectl scale deployment --replicas=2 nginx-deployment
   

   deployment.apps/nginx-deployment scaled
   

   kubectl get deployments
   

   NAME               READY   UP-TO-DATE   AVAILABLE   AGE
   nginx-deployment   2/2     2            2           37m
   

   kubectl get nodes
   

   NAME                                                STATUS   ROLES           AGE   VERSION
   mycluster-l76uten6iljx-control-plane-xlmq4          Ready    control-plane   100m   v1.35.4
   mycluster-l76uten6iljx-default-worker-4dqrj-cwmjq   Ready    <none>          97m    v1.35.4
   os coe cluster list
   +--------------------------------------+-----------+---------+------------+--------------+-----------------+---------------+
   | uuid                                 | name      | keypair | node_count | master_count | status          | health_status |
   +--------------------------------------+-----------+---------+------------+--------------+-----------------+---------------+
   | cae2a382-26a5-438d-bbdc-e694249bbede | mycluster | xxxxxx  |          1 |            1 | CREATE_COMPLETE | HEALTHY       |
   +--------------------------------------+-----------+---------+------------+--------------+-----------------+---------------+
   

Note: Please allow several minutes for the scale-down process to complete, as the typical duration ranges between 7 and 10 minutes.

More information

For more information, refer to:

• Cluster Autoscaler