runner.instance-evict

Drain a single runner instance (PENDING/IDLE/BUSY → DRAINING). Org-scoped. Reason is recorded on the transition event. Idempotent on already-terminal instances.

Drain a single instance.

Surgical removal — used for debugging stuck instances or removing one rogue VM from a healthy pool. The instance enters DRAINING immediately; if it's BUSY, the current job completes first. The next sweep terminates it. Replacement happens via the normal scale-check loop (sweep → terminate → drop below floor → top-up).

No-op (returns already_terminal) on instances already in DRAINING / FAILED / TERMINATED so re-running is safe.

Overview

Property	Value
Workflow type	Atomic
Library	App-runners
Version	1.0

Input Schema

Field	Type	Required	Default	Description
runner_instance_uuid	uuid	Yes	—	UUID of the RunnerInstance to evict, picked from the organization's (optionally group-scoped) list of instances.
reason	string	No	—	—

Output Schema

Field	Type	Required	Default	Description
runner_instance_uuid	uuid	No	—	UUID of the RunnerInstance that was evicted, echoed from input on output.
runner_group_uuid	uuid	No	—	UUID of the RunnerGroup the evicted instance belongs to, picked from the organization's list of runner groups.
previous_status	string	No	—	—
outcome	string	No	—	—
status	string	No	—	—
completed_at	string	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	—	—
reason	string	No	—	—

States

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

State Diagram

Transitions

From	Action	To	Description
pending	evict	completed	—
* (any state)	fail	failed	—

API Usage

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

{
  "workflow_type": "runner.instance-evict",
  "initial_data": {
    "runner_instance_uuid": "value"
  }
}