Autoscaling group API

Creates an autoscaling group with bounded replica targets.

Send name, pinned template revision, region, required deployment inputs, minimum and maximum replicas, and metric policy. Bounds are non-negative integers with minimum no greater than maximum. Zero minimum requires an independent supported scale-from-zero metric.

To keep 2 to 8 worker replicas at a target CPU utilization of 65%, send:

{
  "name": "payments-workers",
  "template_revision_id": "trv_workers_4",
  "region_id": "reg_euwest",
  "min_replicas": 2,
  "max_replicas": 8,
  "inputs": { "interfaces": [{ "network_id": "net_production" }] },
  "metric_policy": { "metric": "droplet.cpu.utilization", "target": 0.65 },
  "paused": false
}

Returns 202 with group ID and reconciliation operation. Invalid metric units, missing template inputs, or unsupported policies return 422.

Returns desired, ready, pending-create, pending-remove counts, owned member IDs, metric evaluation, blockers, and ETag. Requires autoscaling:read. Unknown metrics prevent destructive scale-down rather than becoming zero load.

Updates autoscaling bounds, policy, paused state, or the template for future members.

Requires If-Match. Updates bounds, metric policy, paused state, or the template for new members. Supplied nested objects replace their full value. Existing members are not replaced merely because the template changed.

Requests a temporary override of the desired replica count.

Accepts desired replica count within bounds and a required override duration permitted by policy. Accepted work returns 202. On override expiry, automatic evaluation can choose another count.

Scale-in respects required drain and protection.

Deletes an autoscaling group under an explicit member-retention or removal plan.

Group deletion requires a reviewed member-retention or removal plan and cannot silently discard stateful members. See autoscaling, spend caps, and rolling replacement.