Enrich — Orkestra

enrich tells Orkestra which secondary Kubernetes objects to fetch at reconcile time and embed into the child resource’s template context. Without it, notes like podCount, hasWarnings, and deploymentReplicaSetCount return zero or empty — they have no data to work with.

Why it exists. Orkestra’s informer watches your CR. When something changes — a pod crashes, a new ReplicaSet is created, an event is emitted — the CR is requeued. But the reconciler only reads the CR itself from the cache. The child objects (pods, events, ReplicaSets) exist in Kubernetes, not in the operator’s in-memory cache. enrich is the explicit instruction to go fetch them and attach them to the child map that notes and status templates consume.

enrich:
  - pods     # always fetch pod list for .children.deployment
  - events   # always fetch warning events

Or conditionally:

enrich:
  - pods
  - events:
      when:
        - field: "{{ replicasReady .children.deployment }}"
          equals: "false"

Where enrichment sits in the pipeline

informer cache → DeepCopy → normalize → mutation → validation
    → template reconcile
        → read children
        → enrich ← you are here
        → status fields (now have _pods, _warnings, etc.)

Enrichment runs after children are read and before status fields are written. The _pods, _warnings, _replicaSets keys are available in every status field template that reconcile cycle.

The cost

Each enrichment target is one or more Kubernetes API calls per reconcile. This is the most important thing to understand about enrichment.

Target	API calls	When it matters
`pods`	1 list-pods per child	Every reconcile — keep this cheap
`events`	1 list-events per child	Events are verbose — gate this
`replicasets`	1 list-replicasets per child	Only useful during rollouts
`pvc`, `storageclass`	1 get per object	Usually cheap
`endpoints`	1 list-endpointslices per service	Scales with service count

A Deployment with enrich: [pods] reconciling every 15 seconds = 4 pod-list calls per minute per CR. With 50 CRs that is 200 calls/minute purely for pod enrichment. Gate what you can.

See Cost and When to Use for the full cost model and gating patterns.

Try it

ork init --pack use-cases
cd enrich/01-pod-health         # always-on pod enrichment — count, readiness, crash detection
cd enrich/02-warning-events     # conditional event enrichment — only fetched when degraded
cd enrich/03-rollout-observer   # conditional replicaset enrichment — only during rollouts
ork run

Where to go next

Cost and When to Use — API call budget, enrichAll warning, the gating pattern
Targets — all enrichment targets, what they embed, which notes they unlock
Conditional Enrichment — gate patterns, anyOf, combining conditions