Create a node pool
Creates a worker node pool and starts provisioning. The pool takes the cluster’s Kubernetes version.
Returns 400 for, among others: parent cluster not found; a duplicate
pool name in the cluster; missing nico.instanceTypeID
or nico.release; not exactly one of
subnetIDs and vpcPrefixIDs; a pod CIDR too small for the requested
nodes.
Authentication
A JWT from an issuer the Operator has configured, carried as Authorization: Bearer <token>.
Request
Prefix for generated node names: <cluster name>-<nodeNamePrefix>-<suffix>.
The controller assigns suffix as a 4-digit decimal number (0000–9999).
Generated names must fit within 63 characters (Kubernetes Node name limit).
Maximum prefix length depends on cluster name length.
The desired number of nodes in the pool. For control-plane pools, aggregate capacity above 1 requires controlPlaneVIP on the cluster.
Annotations set on CAPI Machines only when they are created, including Machines created during scale-out or replacement. Fixed at node-pool creation. Not reconciled after provisioning. Keys under nke.nvidia.com, nke.io, nvke.nvidia.com, kubernetes.io, k8s.io, cluster.x-k8s.io and their subdomains require an explicit deployment allowlist entry for this metadata type.
Labels set on CAPI Machines only when they are created, including Machines created during scale-out or replacement. Fixed at node-pool creation. Not reconciled after provisioning. Keys under nke.nvidia.com, nke.io, nvke.nvidia.com, kubernetes.io, k8s.io, cluster.x-k8s.io and their subdomains require an explicit deployment allowlist entry for this metadata type.
Labels applied to the pool’s nodes at creation, as a map of keys to values. Keys under nvke.nvidia.com, nke.nvidia.com (legacy), nke.io (legacy), kubernetes.io, k8s.io and their subdomains are reserved, unless the deployment allows specific Kubernetes keys.
Taints applied to the pool’s nodes at creation. Not allowed on control-plane pools. Keys under nvke.nvidia.com, nke.nvidia.com (legacy), nke.io (legacy), kubernetes.io, k8s.io and their subdomains are reserved. Each key and effect pair may appear only once.
Free-text description.
Disables auto-mode (automatic node recycling).
Response
Node pool created; provisioning has started
Node names are <cluster>-<prefix>-<NNNN> (four-digit suffix), at most 63 characters in total.
Lifecycle conditions: provisioning, provisioned, deleting.
Disables auto-mode (automatic node recycling).
Annotations set on CAPI Machines only when they are created, including Machines created during scale-out or replacement. Fixed at node-pool creation. Not reconciled after provisioning. Keys under nke.nvidia.com, nke.io, nvke.nvidia.com, kubernetes.io, k8s.io, cluster.x-k8s.io and their subdomains require an explicit deployment allowlist entry for this metadata type.
Labels set on CAPI Machines only when they are created, including Machines created during scale-out or replacement. Fixed at node-pool creation. Not reconciled after provisioning. Keys under nke.nvidia.com, nke.io, nvke.nvidia.com, kubernetes.io, k8s.io, cluster.x-k8s.io and their subdomains require an explicit deployment allowlist entry for this metadata type.
Free-text description.
Nodes allowed above desiredNodeCount during a rollout. 0 means sequential: each node is torn down before its replacement is provisioned.