HorizontalPodAutoscaler

This declares one HorizontalPodAutoscaler to be managed by Orkestra.

Example:

onReconcile:
  hpa:
    - name: "{{ .metadata.name }}-hpa"
      deploymentRef: "{{ .metadata.name }}"
      minReplicas: "{{ .spec.minReplicas }}"
      maxReplicas: "{{ .spec.maxReplicas }}"
      targetCPUUtilizationPercentage: "80"
      forEach:
        field: spec.services
        as: item

Lifecycle

Declare this resource under onCreate for an idempotent, one-time create: Orkestra creates it on the first reconcile and leaves it untouched afterward. Set reconcile: true on the same entry to also apply it as drift correction on every subsequent reconcile. This is a shorthand for declaring the identical entry under onReconcile as well — there’s no need to do both.

Declare a resource under onDelete to run explicit cleanup before the CR’s finalizer is removed. Most resources need no onDelete entry — they are garbage-collected automatically through owner references when the CR itself is deleted.

Fields

name

Type: string

Name — HPA resource name. Default: “{{ .metadata.name }}-hpa”

namespace

Type: string

Namespace — target namespace. Default: CR namespace.

scaleTargetRef

Type: object

ScaleTargetRef — the target workload this HPA scales. Supports Deployment, ReplicaSet, StatefulSet, or any scalable resource. Supports template expressions: “{{ .metadata.name }}”

scaleTargetRef:
  apiVersion: apps/v1
  kind: Deployment
  name: "{{ .metadata.name }}-deployment"

minReplicas

Type: string

MinReplicas — minimum replica count as a string. Supports template expressions.

maxReplicas

Type: string

MaxReplicas — maximum replica count as a string. Supports template expressions.

targetCPUUtilizationPercentage

Type: string

TargetCPUUtilizationPercentage — CPU utilization target (0-100). Supports templates. When behavior.profile is set and this field is empty, the profile provides the default.

behavior

Type: object

Behavior — scale-up and scale-down behavior configuration. Set profile for a complete preset, or configure scaleUp/scaleDown explicitly. Profile and explicit fields are mutually exclusive.

behavior:
  profile: web

behavior:
  scaleUp:
    stabilizationWindowSeconds: 0
  scaleDown:
    stabilizationWindowSeconds: 300

labels

Type: map

Labels applied to HPA metadata. Values support template expressions.

reconcile

Type: boolean

Reconcile: true — also apply this declaration as drift correction on every reconcile. Equivalent to declaring the same entry under both onCreate and onReconcile. When false (default), only runs on onCreate (idempotent create).

when

Type: list

Conditions declares the set of runtime predicates that must all evaluate to true for this resource template to be applied during reconciliation.

Each condition inspects a field on the live Custom Resource using dot-notation (e.g. “spec.enabled”, “metadata.labels.tier”) and compares it against a value using the chosen operator. All conditions in the list are AND‑ed together.

If any condition fails, the resource is skipped for that reconcile cycle. This is not an error — it simply means “do not create/update this resource right now”. This enables expressive, data‑driven orchestration such as:

when:
  - field: spec.exposePublicly
    equals: "true"
  - field: spec.environment
    prefix: "prod"

Conditions allow templates to be selectively activated based on the CR’s state, enabling dynamic topologies, feature flags, environment‑specific behavior, and conditional provisioning without writing Go code.

anyOf

Type: list

AnyOf holds OR conditions — at least one must pass for this resource to be created. Works alongside the existing Conditions (when:) field which uses AND semantics.

anyOf:
  - field: spec.tier
    equals: pro
  - field: spec.tier
    equals: enterprise

forEach

Type: object

ForEach declares dynamic expansion over a list field. When set, one source declaration becomes N declarations — one per list element. .item and .<as> are available in template expressions within this declaration.

forEach:
  field: spec.services
  as: item

sleep

Type: string

Sleep injects an artificial delay into the reconcile of this resource. Useful for autoscale testing, latency simulation, and chaos engineering. Accepts extended duration units (s, m, h, d, w, mo, y).

Quick reference