Resize control-plane nodes

Changes the instance type of every control plane. On a running cluster the control planes are resized one at a time behind an etcd quorum check and a drain, and the response carries the operation_id to follow (clusters with fewer than three control planes are refused); on a stopped cluster the new type is applied at the next start and no operation is returned.

PUT /api/v1/clusters/digitalocean/{cluster_id}/control-plane/instance-type

Resize control-plane nodes

Python

import requests

url = "https://platform.ankra.app/api/v1/clusters/digitalocean/{cluster_id}/control-plane/instance-type"

payload = { "instance_type": "<string>" }
headers = {
    "Authorization": "Bearer <token>",
    "Content-Type": "application/json"
}

response = requests.put(url, json=payload, headers=headers)

print(response.text)

200

{
  "mode": "offline",
  "new_instance_type": "<string>",
  "previous_instance_type": "<string>",
  "updated": 123,
  "operation_id": "3c90c3cc-0d44-4b50-8888-8dd25736052a"
}

Authorizations

Authorization · string · header · required

Bearer authentication header of the form Bearer <token>, where <token> is your auth token.

Headers

authorization · string | null

x-ankra-organisation-id · string | null

Path Parameters

cluster_id · string<uuid> · required

Body

application/json

instance_type · string · required

The new instance type for every control plane node. On a stopped cluster it is applied at the next start; on a running cluster with at least three controllers it is rolled out live, one controller at a time, each drained and waited back to healthy before the next. On a running Hetzner cluster with a single controller it resizes that server in place, which takes the Kubernetes API down while it reboots and needs accept_api_downtime.

Minimum string length: 1

accept_api_downtime · boolean · default: false

Accepts that resizing a RUNNING cluster's only control plane takes the Kubernetes API down for a few minutes while that server reboots on the new type. Required for that resize (Hetzner), which is refused with 409 without it; ignored on every other lane. Per request, never stored as a setting.

force_drain · boolean · default: false

Rolling lane only. Bypass PodDisruptionBudgets when draining each controller. Off by default: a PDB is a workload owner's stated availability requirement.

Response

200

application/json

Successful Response

mode · enum<string> · required

Which lane the request took: 'offline' rewrote a stopped cluster's plan, 'rolling' is the live one-controller-at-a-time path on a running cluster, 'live_with_api_downtime' resizes a running cluster's only controller in place with the Kubernetes API down while it reboots. It describes the lane, not whether work was dispatched - read operation_id for that.

Available options: offline, rolling, live_with_api_downtime

new_instance_type · string · required

previous_instance_type · string · required

updated · integer · required

operation_id · string<uuid> | null

The operation to poll, when the request dispatched one. Null whenever nothing was dispatched: on the offline lane, which applies the new size at the next cluster start (its own operation), and on either lane when the control planes already run the requested type (updated is 0). Decide whether to poll by testing operation_id for null, not by reading mode.