node.md

subcategory	Managed Kubernetes
page_title	Nodes
description	Tools for listing, inspecting, recreating and deleting individual Kubernetes worker nodes in IONOS CLOUD.

Nodes

These tools act on one worker node at a time. They differ in ordering, and that difference is the whole reason to prefer one over the other:

	Sequence	Capacity	Pool size afterwards	Cost
recreate_k8s_node	replacement joins the cluster, then the old node drains	never dips	unchanged	one extra billable node while both run
delete_k8s_node	node goes first, backfill may follow	runs one node short	not guaranteed to recover — see below	no extra node

delete_k8s_node is not a reliable way to replace a node. Whether the pool backfills depends on the pool. Without an autoscaler, node_count is the desired size and the pool should return to it. With an active autoscaler, the autoscaler owns the count — node_count is no longer authoritative — and it can legitimately hold the pool at the smaller size. Observed live: a 2-node pool with min 1 went to 1 node and stayed there.

Use recreate_k8s_node to replace a node, and update_k8s_nodepool to change the size on purpose. The delete_k8s_node preview warns when an autoscaler is active.

Both recreate_k8s_node and delete_k8s_node require IONOS_MCP_TOOL_SCOPE to include destructive, and both are two-phase: the first call previews the node and returns a one-time confirmation_token, and only a second call carrying that token mutates anything.

Asynchronous operations

Both endpoints answer 202 Accepted, so the tool returns before the node has actually gone. Follow progress with list_k8s_nodepool_nodes and read each node's metadata.state. Node state is its own, smaller enum — not the cluster/node pool one:

State	Meaning
READY	node is in service
PROVISIONING, PROVISIONED	replacement coming up
REBUILDING	being recreated
TERMINATING	being removed
BUSY	a modification is pending

Leave at least 30 seconds between polls. A recreate takes several minutes, because the replacement must be provisioned and join the cluster before the old node is drained.

list_k8s_nodepool_nodes

Lists all individual nodes in a Kubernetes node pool, including their state and public/private IP addresses.

Parameters:

Name	Type	Required	Description
k8s_cluster_id	string	Yes	The ID of the Kubernetes cluster
nodepool_id	string	Yes	The ID of the Kubernetes node pool
depth	integer	No	Nesting depth of returned objects (0–5, default 1).

Example:

{
  "name": "list_k8s_nodepool_nodes",
  "arguments": {
    "k8s_cluster_id": "3f2e4b1c-8d9a-4f7e-b3c2-1a5d6e9f0b4c",
    "nodepool_id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890"
  }
}

API Reference: k8sNodepoolsNodesGet

---

get_k8s_node

Gets detailed information about a specific node in a Kubernetes node pool.

Parameters:

Name	Type	Required	Description
k8s_cluster_id	string	Yes	The ID of the Kubernetes cluster
nodepool_id	string	Yes	The ID of the Kubernetes node pool
node_id	string	Yes	The ID of the Kubernetes node
depth	integer	No	Nesting depth of returned objects (0–5).

Example:

{
  "name": "get_k8s_node",
  "arguments": {
    "k8s_cluster_id": "3f2e4b1c-8d9a-4f7e-b3c2-1a5d6e9f0b4c",
    "nodepool_id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
    "node_id": "f9e8d7c6-b5a4-3210-fedc-ba9876543210"
  }
}

API Reference: k8sNodepoolsNodesFindById

---

recreate_k8s_node

Recreates one worker node. Requires IONOS_MCP_TOOL_SCOPE to include destructive.

IONOS provisions a replacement node, waits for it to register with the cluster, and only then drains and destroys the old one — so capacity is preserved throughout. Pods on the old node are evicted and anything on its local storage is lost.

Use this for a node that is unhealthy or stuck. Prefer it over delete_k8s_node, which removes the node before the replacement exists.

Cost: while the replacement runs alongside the old node, the node pool carries an extra billable active node.

The tool is a POST but is classified destructive, because the existing node is thrown away. It is not idempotent: a second call recreates the replacement in turn.

Parameters:

Name	Type	Required	Description
k8s_cluster_id	string	Yes	The ID of the cluster the node belongs to
nodepool_id	string	Yes	The ID of the node pool the node belongs to
node_id	string	Yes	The ID of the node to recreate
confirmation_token	string	No	Omit on the first call; pass the token returned by the preview on the second

Example:

{
  "name": "recreate_k8s_node",
  "arguments": {
    "k8s_cluster_id": "3f2e4b1c-8d9a-4f7e-b3c2-1a5d6e9f0b4c",
    "nodepool_id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
    "node_id": "f9e8d7c6-b5a4-3210-fedc-ba9876543210"
  }
}

API Reference: k8sNodepoolsNodesReplacePost

---

delete_k8s_node

Deletes one worker node from a node pool. Requires IONOS_MCP_TOOL_SCOPE to include destructive. Irreversible.

The node is drained and destroyed, its pods are evicted, and data on its local storage is lost.

The pool is left a node short, and may stay that way. The node is removed first, and whether a replacement follows depends on whether an autoscaler is active — see the table at the top of this page. To change a pool's size on purpose, use update_k8s_nodepool; to replace a node, use recreate_k8s_node.

Blocked by the autoscaler minimum. The API refuses to remove a node when doing so would take the pool below auto_scaling.min_node_count, and reports it as [VDC-14-1826] Operation failed, as last node can not be deleted from nodepool — even when several nodes remain. A pool of 2 pinned at min 2 cannot have a node deleted at all. The tool reads the pool during the preview and says this plainly rather than letting you spend a confirmation token on a request the API will refuse. Lower min_node_count first, or use recreate_k8s_node, which is never blocked because it adds the replacement before removing the old node.

Parameters:

Name	Type	Required	Description
k8s_cluster_id	string	Yes	The ID of the cluster the node belongs to
nodepool_id	string	Yes	The ID of the node pool the node belongs to
node_id	string	Yes	The ID of the node to delete
confirmation_token	string	No	Omit on the first call; pass the token returned by the preview on the second

Example:

{
  "name": "delete_k8s_node",
  "arguments": {
    "k8s_cluster_id": "3f2e4b1c-8d9a-4f7e-b3c2-1a5d6e9f0b4c",
    "nodepool_id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
    "node_id": "f9e8d7c6-b5a4-3210-fedc-ba9876543210"
  }
}

API Reference: k8sNodepoolsNodesDelete