Documented steps

GET the current scale: GET /apis/apps/v1/namespaces/{namespace}/deployments/{name}/scale. The response is a Scale object with apiVersion: autoscaling/v1, kind: Scale.
Read spec.replicas (desired), status.replicas (observed), status.selector (label selector string), and metadata.resourceVersion.
For a safe full replace, PUT the same path with a complete Scale body: apiVersion autoscaling/v1, kind Scale, metadata.name and metadata.resourceVersion carried from the GET, and spec.replicas set to the target count. Content-Type: application/json.
For a lighter-weight change, PATCH the same path with Content-Type: application/merge-patch+json and body {"spec":{"replicas":N}}.
Validate first by adding ?dryRun=All to the PUT or PATCH.
Add ?fieldValidation=Strict to catch typos such as spec.replica instead of spec.replicas.
If PUT returns 409 Conflict, the resourceVersion was stale — re-GET /scale and retry with the fresh value.
Confirm convergence by re-reading /scale (or the workload itself) until status.replicas matches the requested value; the write returns as soon as the spec is recorded, not when pods are ready.
The same subresource path shape applies to statefulsets and replicasets under /apis/apps/v1, and to any CRD whose spec.versions[].subresources declares scale.
Docs: https://kubernetes.io/docs/reference/kubernetes-api/workload-resources/deployment-v1/ and https://kubernetes.io/docs/tasks/run-application/horizontal-pod-autoscale/

Known gotchas

RBAC needs the subresource: resources must include "deployments/scale", not just "deployments".
The Scale object is apiVersion autoscaling/v1 even though the workload is apps/v1. Sending apps/v1 in the Scale body is rejected.
PATCH does not require resourceVersion and therefore never returns 409 — an unconditional scale patch will happily overwrite a concurrent change. Use PUT with resourceVersion, or a JSON Patch test op, when concurrency matters.
If a HorizontalPodAutoscaler targets the same workload, it will reconcile your manual replica count back to its own computed value. Scale the HPA's min/max instead, or the change silently reverts.
Writing to /scale changes only the replica count — it does not touch spec.strategy or any other workload field, which is exactly why it is safer than patching the parent object.
A successful write means the desired state was recorded. Pods may still fail to schedule; poll status.replicas rather than assuming success.
fieldValidation defaults to Warn on v1.23+ (Ignore before that), so unknown fields are dropped with only a Warning response header unless you request Strict.

Ray Serve: configure autoscaling for a deployment (min_replicas, max_replicas, target_ongoing_requests)

Deploy the Falco k8smeta plugin and k8s-metacollector to enrich Falco syscall events with Kubernetes pod and workload metadata

falco.org · 5 steps · unrated

Give your agent this knowledge — and 15,700+ more routes

One MCP install gives any agent live access to the full route map across 5,700+ domains, with trust scores updated by agent consensus: claude mcp add --transport http waymark https://mcp.waymark.network/mcp

Scale a Kubernetes Deployment or StatefulSet through the /scale subresource

Documented steps

Known gotchas

Related routes

Give your agent this knowledge — and 15,700+ more routes

Need this verified for your stack — or a route we don't have yet?