Scaling Node Pools

Find out how to scale up and scale down the node pools you've created using Kubernetes Engine (OKE).

To optimize resource usage, you can scale a node pool up and down to change the number of worker nodes in the node pool, and the availability domains and subnets in which to place them.

For general information about modifying node pools and worker nodes, see Modifying Node Pool and Worker Node Properties. In particular, note the following:

• Any changes you make to worker node properties will only apply to new worker nodes. You cannot change the properties of existing worker nodes.
• If you change a node pool's placement configuration (the availability domains, fault domains, and subnets in which worker nodes are placed, but not the node pool's capacity type), existing worker nodes are terminated and new worker nodes are created in the new locations.
• Changing the capacity type associated with a node pool placement configuration applies to new worker node instances. Existing worker node instances are not moved automatically. To update the capacity type for existing nodes, terminate and replace the existing worker node instances. For more information, see Terminating and Replacing Worker Nodes.
• If a capacity reservation is specified for a node pool, note that the node shape, availability domain, and fault domain in the node pool's placement configuration must always match the capacity reservation's instance type, availability domain, and fault domain respectively. See Using Capacity Reservations to Provision Managed Nodes.
• If a compute host group is specified for a node pool placement configuration, the compute host group must be in the same availability domain as the placement configuration. The node pool shape must also match the compute host group shape. See Using Compute Host Groups to Provision Managed Nodes.
• If a compute cluster is specified for a node pool, the node pool shape must support RDMA and compute clusters, and the placement configuration must include only the availability domain that contains the compute cluster. Fault domains cannot be specified for node pools that use compute clusters. See Using Compute Clusters to Provision Managed Nodes.
• If a compute cluster is specified for a node pool, new worker node instances are launched in the same compute cluster. You cannot change the compute cluster when scaling a node pool. To use a different compute cluster, create a new node pool.
• Do not use the kubectl delete node command to scale down or terminate worker nodes in a cluster that was created by Kubernetes Engine. Instead, reduce the number of worker nodes by changing the corresponding node pool properties using the Console or the API. The kubectl delete node command does not change a node pool's properties, which determine the desired state (including the number of worker nodes). Also, although the kubectl delete node command removes the worker node from the cluster's etcd key-value store, the command does not delete the underlying instance or virtual node.
• When scaling down a node pool, note that the node pool's Cordon and drain properties determine when and how worker nodes are terminated. See Cordoning and Draining Managed Nodes Before Shut Down or Termination.