Working with the Cluster Autoscaler as a Standalone Program

1. In a text editor, create a file called cluster-autoscaler.yaml with the following content:

   ---
   apiVersion: v1
   kind: ServiceAccount
   metadata:
     labels:
       k8s-addon: cluster-autoscaler.addons.k8s.io
       k8s-app: cluster-autoscaler
     name: cluster-autoscaler
     namespace: kube-system
   ---
   apiVersion: rbac.authorization.k8s.io/v1
   kind: ClusterRole
   metadata:
     name: cluster-autoscaler
     labels:
       k8s-addon: cluster-autoscaler.addons.k8s.io
       k8s-app: cluster-autoscaler
   rules:
     - apiGroups: [""]
       resources: ["events", "endpoints"]
       verbs: ["create", "patch"]
     - apiGroups: [""]
       resources: ["pods/eviction"]
       verbs: ["create"]
     - apiGroups: [""]
       resources: ["pods/status"]
       verbs: ["update"]
     - apiGroups: [""]
       resources: ["endpoints"]
       resourceNames: ["cluster-autoscaler"]
       verbs: ["get", "update"]
     - apiGroups: [""]
       resources: ["nodes"]
       verbs: ["watch", "list", "get", "patch", "update"]
     - apiGroups: [""]
       resources:
         - "pods"
         - "services"
         - "replicationcontrollers"
         - "persistentvolumeclaims"
         - "persistentvolumes"
       verbs: ["watch", "list", "get"]
     - apiGroups: ["extensions"]
       resources: ["replicasets", "daemonsets"]
       verbs: ["watch", "list", "get"]
     - apiGroups: ["policy"]
       resources: ["poddisruptionbudgets"]
       verbs: ["watch", "list"]
     - apiGroups: ["apps"]
       resources: ["statefulsets", "replicasets", "daemonsets"]
       verbs: ["watch", "list", "get"]
     - apiGroups: ["storage.k8s.io"]
       resources: ["storageclasses", "csinodes"]
       verbs: ["watch", "list", "get"]
     - apiGroups: ["batch", "extensions"]
       resources: ["jobs"]
       verbs: ["get", "list", "watch", "patch"]
     - apiGroups: ["coordination.k8s.io"]
       resources: ["leases"]
       verbs: ["create"]
     - apiGroups: ["coordination.k8s.io"]
       resourceNames: ["cluster-autoscaler"]
       resources: ["leases"]
       verbs: ["get", "update"]
     - apiGroups: [""]
       resources: ["namespaces"]
       verbs: ["watch", "list"]
     - apiGroups: ["storage.k8s.io"]
       resources: ["csidrivers", "csistoragecapacities"]
       verbs: ["watch", "list"]
   ---
   apiVersion: rbac.authorization.k8s.io/v1
   kind: Role
   metadata:
     name: cluster-autoscaler
     namespace: kube-system
     labels:
       k8s-addon: cluster-autoscaler.addons.k8s.io
       k8s-app: cluster-autoscaler
   rules:
     - apiGroups: [""]
       resources: ["configmaps"]
       verbs: ["create","list","watch"]
     - apiGroups: [""]
       resources: ["configmaps"]
       resourceNames: ["cluster-autoscaler-status", "cluster-autoscaler-priority-expander"]
       verbs: ["delete", "get", "update", "watch"]
   
   ---
   apiVersion: rbac.authorization.k8s.io/v1
   kind: ClusterRoleBinding
   metadata:
     name: cluster-autoscaler
     labels:
       k8s-addon: cluster-autoscaler.addons.k8s.io
       k8s-app: cluster-autoscaler
   roleRef:
     apiGroup: rbac.authorization.k8s.io
     kind: ClusterRole
     name: cluster-autoscaler
   subjects:
     - kind: ServiceAccount
       name: cluster-autoscaler
       namespace: kube-system
   
   ---
   apiVersion: rbac.authorization.k8s.io/v1
   kind: RoleBinding
   metadata:
     name: cluster-autoscaler
     namespace: kube-system
     labels:
       k8s-addon: cluster-autoscaler.addons.k8s.io
       k8s-app: cluster-autoscaler
   roleRef:
     apiGroup: rbac.authorization.k8s.io
     kind: Role
     name: cluster-autoscaler
   subjects:
     - kind: ServiceAccount
       name: cluster-autoscaler
       namespace: kube-system
   
   ---
   apiVersion: apps/v1
   kind: Deployment
   metadata:
     name: cluster-autoscaler
     namespace: kube-system
     labels:
       app: cluster-autoscaler
   spec:
     replicas: 3
     selector:
       matchLabels:
         app: cluster-autoscaler
     template:
       metadata:
         labels:
           app: cluster-autoscaler
         annotations:
           prometheus.io/scrape: 'true'
           prometheus.io/port: '8085'
       spec:
         serviceAccountName: cluster-autoscaler
         containers:
           - image: iad.ocir.io/oracle/oci-cluster-autoscaler:{{ image tag }}
             name: cluster-autoscaler
             resources:
               limits:
                 cpu: 100m
                 memory: 300Mi
               requests:
                 cpu: 100m
                 memory: 300Mi
             command:
               - ./cluster-autoscaler
               - --v=4
               - --stderrthreshold=info
               - --cloud-provider=oci
               - --max-node-provision-time=25m
               - --nodes=1:5:{{ node pool ocid 1 }}
               - --nodes=1:5:{{ node pool ocid 2 }}
               - --scale-down-delay-after-add=10m
               - --scale-down-unneeded-time=10m
               - --unremovable-node-recheck-timeout=5m
               - --balance-similar-node-groups
               - --balancing-ignore-label=displayName
               - --balancing-ignore-label=hostname
               - --balancing-ignore-label=internal_addr
               - --balancing-ignore-label=oci.oraclecloud.com/fault-domain
             imagePullPolicy: "Always"
             env:
             - name: OKE_USE_INSTANCE_PRINCIPAL
               value: "true"
             - name: OCI_SDK_APPEND_USER_AGENT
               value: "oci-oke-cluster-autoscaler"

2. In the cluster-autoscaler.yaml file you created, confirm that the --cloud-provider parameter is set correctly for the version of Kubernetes running on the cluster. By default, the parameter assumes the cluster is running Kubernetes version 1.27 (or 1.23 or earlier) and is set to oci. If the cluster is running Kubernetes version 1.26, 1.25, or 1.24, change the value of the --cloud-provider parameter to oci-oke:

   1. In the cluster-autoscaler.yaml file, locate the following line:

      - --cloud-provider=oci

   2. If the cluster is running Kubernetes version 1.26, 1.25, or 1.24, change the value of the --cloud-provider parameter to oci-oke :

      - --cloud-provider=oci-oke

   3. Save the cluster-autoscaler.yaml file.
3. In the cluster-autoscaler.yaml file you created, change the image path of the Kubernetes Cluster Autoscaler image to download from Oracle Cloud Infrastructure Registry. Images are available in a number of regions. For the best performance, choose the region closest to the one where the cluster is deployed:

   1. In the cluster-autoscaler.yaml file, locate the following template line:

      - image: iad.ocir.io/oracle/oci-cluster-autoscaler:{{ image tag }}

   2. Change the image path to one of the following, according to the location and Kubernetes version of the cluster in which to run the Kubernetes Cluster Autoscaler:

      Image Location	Kubernetes Version	Image Path
      Germany Central (Frankfurt)	Kubernetes 1.26	fra.ocir.io/oracle/oci-cluster-autoscaler:1.26.2-11
      Germany Central (Frankfurt)	Kubernetes 1.27	fra.ocir.io/oracle/oci-cluster-autoscaler:1.27.2-9
      Germany Central (Frankfurt)	Kubernetes 1.28	fra.ocir.io/oracle/oci-cluster-autoscaler:1.28.2-3
      Germany Central (Frankfurt)	Kubernetes 1.29	fra.ocir.io/oracle/oci-cluster-autoscaler:1.29.0-10
      UK South (London)	Kubernetes 1.26	lhr.ocir.io/oracle/oci-cluster-autoscaler:1.26.2-11
      UK South (London)	Kubernetes 1.27	lhr.ocir.io/oracle/oci-cluster-autoscaler:1.27.2-9
      UK South (London)	Kubernetes 1.28	lhr.ocir.io/oracle/oci-cluster-autoscaler:1.28.2-3
      UK South (London)	Kubernetes 1.29	lhr.ocir.io/oracle/oci-cluster-autoscaler:1.29.0-10
      US East (Ashburn)	Kubernetes 1.26	iad.ocir.io/oracle/oci-cluster-autoscaler:1.26.2-11
      US East (Ashburn)	Kubernetes 1.27	iad.ocir.io/oracle/oci-cluster-autoscaler:1.27.2-9
      US East (Ashburn)	Kubernetes 1.28	iad.ocir.io/oracle/oci-cluster-autoscaler:1.28.2-3
      US East (Ashburn)	Kubernetes 1.29	iad.ocir.io/oracle/oci-cluster-autoscaler:1.29.0-10
      US West (Phoenix)	Kubernetes 1.26	phx.ocir.io/oracle/oci-cluster-autoscaler:1.26.2-11
      US West (Phoenix)	Kubernetes 1.27	phx.ocir.io/oracle/oci-cluster-autoscaler:1.27.2-9
      US West (Phoenix)	Kubernetes 1.28	phx.ocir.io/oracle/oci-cluster-autoscaler:1.28.2-3
      US West (Phoenix)	Kubernetes 1.29	phx.ocir.io/oracle/oci-cluster-autoscaler:1.29.0-10

      For example, if you want to run the Kubernetes Cluster Autoscaler in a Kubernetes 1.29 cluster located in the UK South region, specify the following image:

      - image: lhr.ocir.io/oracle/oci-cluster-autoscaler:1.29.0-10

      Tip
      
      

      If you want to deploy the Kubernetes Cluster Autoscaler on a Kubernetes cluster that is not in the same region as any of the Oracle repositories containing Cluster Autoscaler images, we recommend you push the image to a repository that is in the same region as the cluster, as follows:

      i. Pull the image from an Oracle repository using the docker pull command. See Pulling Images Using the Docker CLI.

      ii. Tag the image (using the docker tag command), and then push the image to a repository in Oracle Cloud Infrastructure Registry that is in the same region as the cluster in which you want to run the Kubernetes Cluster Autoscaler (using the docker push command). See Pushing Images Using the Docker CLI.

      iii. Specify the location of the image in the cluster-autoscaler.yaml file.

      Note
      
      

      If you want to deploy the Kubernetes Cluster Autoscaler on a Kubernetes cluster where you have enabled image verification, do not simply specify an image path from one of the Oracle repositories in the cluster-autoscaler.yaml file. Instead, do the following:

      i. Pull the image from an Oracle repository using the docker pull command. See Pulling Images Using the Docker CLI.

      ii. Tag the image (using the docker tag command), and then push the image to a repository in Oracle Cloud Infrastructure Registry that is in the same region as the cluster in which you want to run the Kubernetes Cluster Autoscaler (using the docker push command). See Pushing Images Using the Docker CLI.

      iii. Sign the image using a master key and key version in the Vault service, creating an image signature. See Signing Images for Security.

      iv. Specify the location of the signed image in the cluster-autoscaler.yaml file. Reference the image using the image digest rather than the image tag (see Enforcing the Use of Signed Images from Registry).

   3. Save the cluster-autoscaler.yaml file.

4. In the cluster-autoscaler.yaml file you created, specify each of the cluster's node pools that you want the Kubernetes Cluster Autoscaler to manage.

   You can specify multiple node pools in the cluster-autoscaler.yaml file. Note the recommendation is to always have at least one node pool that is not managed by the Kubernetes Cluster Autoscaler. Also note that it is your responsibility to manually scale any node pools you do not specify in the configuration file.

   1. In the cluster-autoscaler.yaml file, locate the following template line:

      - --nodes=1:5:{{ node pool ocid 1 }}

      The --nodes parameter has the following format:

      --nodes=<min-nodes>:<max-nodes>:<nodepool-ocid>

      where:

      • <min-nodes> is the minimum number of nodes allowed in the node pool. The Kubernetes Cluster Autoscaler will not reduce the number of nodes below this number.
      • <max-nodes> is the maximum number of nodes allowed in the node pool. The Kubernetes Cluster Autoscaler will not increase the number of nodes above this number. Make sure the maximum number of nodes you specify does not exceed the tenancy limits for the worker node shape defined for the node pool.
      • <nodepool-ocid> is one or more node pool OCIDs.

   2. Change the value of the --nodes parameter to specify:

      • The minimum number of nodes allowed in the node pool. For example, 1.
      • The maximum number of nodes allowed in the node pool. For example, 5.
      • The OCID of the node pool you want the Kubernetes Cluster Autoscaler to manage.

      For example:

      --nodes=1:5:ocid1.nodepool.oc1.iad.aaaaaaaaaeydq...

   3. If you only want the Kubernetes Cluster Autoscaler to manage one node pool in the cluster, locate the following line in the cluster-autoscaler.yaml file and remove it:

      - --nodes=1:5:{{ node pool ocid 2 }}

   4. If you want the Kubernetes Cluster Autoscaler to manage a second node pool in the cluster, locate the following line in the cluster-autoscaler.yaml file and set appropriate values for the --nodes parameter:

      - --nodes=1:5:{{ node pool ocid 2 }}

   5. If you want the Kubernetes Cluster Autoscaler to manage more node pools, insert additional --nodes parameters in the cluster-autoscaler.yaml file and set appropriate values for them.
   6. Save the cluster-autoscaler.yaml file.

5. In the cluster-autoscaler.yaml file you created, confirm that the default values of the CPU and memory limit parameters are sufficient for the number of node pools that you want the Kubernetes Cluster Autoscaler to manage. The default limits are relatively low, so consider increasing the limits if you want the Kubernetes Cluster Autoscaler to manage a large number of node pools. Note that it is your responsibility to set the limits to suitable values.

   1. In the cluster-autoscaler.yaml file, locate the following lines:

                resources:
                  limits:
                    cpu: 100m
                    memory: 300Mi

   2. Set the CPU and memory limits to values that are appropriate for the number of node pools that you want the the Kubernetes Cluster Autoscaler to manage. For example:

                resources:
                  limits:
                    cpu: 200m
                    memory: 600Mi

   3. Save the cluster-autoscaler.yaml file.

6. In the cluster-autoscaler.yaml file you created, specify other parameters for the Kubernetes Cluster Autoscaler. For information about the parameters you can set, see Supported Kubernetes Cluster Autoscaler Parameters.

7. Save and close the cluster-autoscaler.yaml file.

1. If you haven't already done so, follow the steps to set up the cluster's kubeconfig configuration file and (if necessary) set the KUBECONFIG environment variable to point to the file. Note that you must set up your own kubeconfig file. You cannot access a cluster using a kubeconfig file that a different user set up. See Setting Up Cluster Access.
2. Deploy the Kubernetes Cluster Autoscaler on the cluster by entering:

   kubectl apply -f cluster-autoscaler.yaml

3. View the Kubernetes Cluster Autoscaler logs to confirm that it was successfully deployed and is currently monitoring the workload of node pools in the cluster, by entering:

   kubectl -n kube-system logs -f deployment.apps/cluster-autoscaler

4. Identify which one of the three Kubernetes Cluster Autoscaler pods defined in the cluster-autoscaler.yaml file is currently performing actions, by entering:

   kubectl -n kube-system get lease

5. Obtain a high-level view of the Kubernetes Cluster Autoscaler's state from the configmap in the kube-system namespace, by entering:

   kubectl -n kube-system get cm cluster-autoscaler-status -oyaml

You can watch the Kubernetes Cluster Autoscaler you have deployed as it automatically scales worker nodes in a node pool. To make the scaling operation more obvious, consider the following suggestions (note these are for observation purposes only, and might be contrary to recommendations shown in Recommendations when using the Kubernetes Cluster Autoscaler in Production Environments):

• Observe a cluster that has a single node pool (the node pool being managed by the Kubernetes Cluster Autoscaler).
• If the cluster you want to observe has more than one node pool, restrict pods to running on nodes on the single node pool being managed by the Kubernetes Cluster Autoscaler. See Assigning Pods to Nodes in the Kubernetes documentation.
• Start with one node in the node pool being managed by the Kubernetes Cluster Autoscaler.
• In the Kubernetes Cluster Autoscaler configuration file, you specify the maximum number of nodes allowed in the node pool. Make sure the maximum number of nodes you specify does not exceed the tenancy limit for the worker node shape defined for the node pool.

To view the Kubernetes Cluster Autoscaler automatically scaling worker nodes:

1. Confirm the current total number of worker nodes in the cluster by entering:

   kubectl get nodes

2. Define a sample Nginx application by creating a file called nginx.yaml in a text editor, with the following content:

   apiVersion: apps/v1
   kind: Deployment
   metadata:
     name: nginx-deployment
   spec:
     selector:
       matchLabels:
         app: nginx
     replicas: 2
     template:
       metadata:
         labels:
           app: nginx
       spec:
         containers:
         - name: nginx
           image: nginx:latest
           ports:
           - containerPort: 80
           resources:
             requests:
               memory: "500Mi"

   Notice that a resource request limit has been set.

3. Deploy the sample application by entering:

   kubectl create -f nginx.yaml

4. Increase the number of pods in the deployment to 100 (from 2) by entering:

   kubectl scale deployment nginx-deployment --replicas=100

   The Kubernetes Cluster Autoscaler now adds worker nodes to the node pool to meet the increased workload.

5. Observe the status of the deployment by entering:

   kubectl get deployment nginx-deployment --watch

6. After a few minutes, view the increased total number of worker nodes in the cluster by entering:

   kubectl get nodes

   Note that the number of worker nodes that you see will depend on the worker node shape and the maximum number of nodes specified in the Kubernetes Cluster Autoscaler configuration file.

1. Delete the sample Nginx application by entering:

   kubectl delete deployment nginx-deployment

2. After ten minutes, confirm that the worker nodes have reduced to the original number, by entering:

   kubectl get nodes

Note that after deleting the sample Nginx application and waiting, you might see fewer worker nodes but still more than the original number. This is probably because kube-system pods have been scheduled to run on those nodes. kube-system pods can prevent the Kubernetes Cluster Autoscaler from removing nodes because the Autoscaler's skip-nodes-with-system-pods parameter is set to true by default.