kubernetes.node.drain

Cordon a Node and evict its Pods (kubectl drain equivalent)

Cordons a Kubernetes Node (marks it unschedulable) and then evicts all eligible Pods via the Eviction API, equivalent to kubectl drain.

Behaviour

• Cordon — patches spec.unschedulable: true on the node so no new Pods are scheduled while evictions are in flight.
• Eviction filter — pods are classified on every iteration:
  • Completed / Succeeded pods are always skipped.
  • DaemonSet-owned pods are skipped when ignore_daemonsets=true (default). When false, their presence is treated as a blocking error.
  • Unowned pods (no ownerReferences) are evicted only when force=true. With force=false (default), they cause an immediate FAILED transition.
• Polling — after all initial evictions are issued the workflow enters the WAITING state and polls every poll_interval seconds. The drain is considered complete when no evictable pods remain on the node. If timeout_seconds elapses before that, the workflow transitions to COMPLETED with remaining_pods populated (non-empty list = soft timeout).

Inputs:

• connection_uuid: Cloud connection UUID (required)
• name: Node name (required)
• grace_period_seconds: Eviction grace period in seconds (default 30)
• ignore_daemonsets: Skip DaemonSet-owned pods (default true)
• force: Evict pods with no owning controller (default false)
• timeout_seconds: Hard polling timeout, capped at 1800 (default 300)
• poll_interval: Seconds between polls, capped at 60 (default 5)

Outputs:

• pods_evicted: Number of pods for which an eviction was issued.
• pods_skipped: Number of pods intentionally skipped.
• remaining_pods: JSON list of {namespace, name} still present at timeout (empty list when drain completed cleanly).

Plugin required: context.get_plugin("connection").kubernetes_runner_clients(connection_uuid, organization_uuid) must expose:

• core_v1 (CoreV1Api) with read_node, patch_node, list_pod_for_all_namespaces, create_namespaced_pod_eviction.

Overview

Property	Value
Workflow type	Linear
Library	App-kubernetes
Version	1.0

Input Schema

Field	Type	Required	Default	Description
connection_uuid	uuid	Yes	—	—
name	string	Yes	—	—
grace_period_seconds	integer	No	—	—
ignore_daemonsets	boolean	No	—	—
force	boolean	No	—	—
timeout_seconds	integer	No	—	—
poll_interval	integer	No	—	—
organization_uuid	uuid	No	—	—

Output Schema

Field	Type	Required	Default	Description
connection_uuid	uuid	No	—	—
organization_uuid	uuid	No	—	—
name	string	No	—	—
grace_period_seconds	integer	No	—	—
ignore_daemonsets	boolean	No	—	—
force	boolean	No	—	—
timeout_seconds	integer	No	—	—
poll_interval	integer	No	—	—
pods_evicted	integer	No	—	—
pods_skipped	integer	No	—	—
remaining_pods	json	No	—	—
drain_started_at	float	No	—	—
failure_reason	string	No	—	—
failure_type	string	No	—	—
failed_action	string	No	—	—
failed_at_state	string	No	—	—
failed_step	string	No	—	—
failed_layer	string	No	—	—
error	string	No	—	—
error_type	string	No	—	—

States

State	Initial	Terminal	Success	Auto-advance	Description
pending	Yes	No	—	execute	—
waiting	No	No	—	—	—
completed	No	Yes	Yes	—	—
failed	No	Yes	No	—	—

State Diagram

Transitions

From	Action	To	Description
pending	execute	waiting	—
waiting	poll	waiting	—
waiting	complete	completed	—
* (any state)	fail	failed	—

API Usage

POST /api/workflows/start
Content-Type: application/json

{
  "workflow_type": "kubernetes.node.drain",
  "initial_data": {
    "connection_uuid": "value",
    "name": "value"
  }
}