openova/platform/knative

Knative

Serverless platform for Kubernetes with scale-to-zero and event-driven capabilities. Application Blueprint (see docs/PLATFORM-TECH-STACK.md §4.6 — AI/ML). Used by bp-cortex (composite AI Hub Blueprint) as the serverless layer for KServe-managed model inference.

Status: Accepted | Updated: 2026-04-30

---

Blueprint chart

This folder ships an umbrella Helm chart at chart/ that wraps the upstream knative-operator chart (v1.21.1) under dependencies:. Catalyst-curated overlay templates render alongside:

• chart/templates/knativeserving.yaml — operator.knative.dev/v1beta1.KnativeServing CR pre-configured for istio-less mode (Cilium native Gateway-API ingress, no Knative-Istio sidecar). Domain template is sourced from knativeOverlay.knativeServing.sovereignFqdn — REQUIRED, no hardcoded fallback per docs/INVIOLABLE-PRINCIPLES.md #4.
• chart/templates/networkpolicy.yaml — locks the operator namespace down (DEFAULT FALSE).
• chart/templates/servicemonitor.yaml — operator metrics scrape (DEFAULT FALSE per docs/BLUEPRINT-AUTHORING.md §11.2; Capabilities-gated).
• chart/templates/hpa.yaml — operator Deployment HPA (DEFAULT FALSE; operator is leader-elected so HPA rarely makes sense).

Istio-less mode: the KnativeServing CR ships with ingress.istio.enabled: false, ingress.contour.enabled: false, ingress.kourier.enabled: false, and config.network.ingress-class: cilium.ingress.networking.knative.dev so Knative Routes resolve to Cilium HTTPRoute / Gateway-API objects.

---

Overview

Knative provides serverless capabilities on Kubernetes. Used independently for event-driven workloads and as the foundation for KServe model serving.

flowchart TB
    subgraph Knative["Knative"]
        Serving[Knative Serving]
        Eventing[Knative Eventing]
    end

    subgraph Use Cases
        KServe[KServe<br/>Model Serving]
        Functions[Serverless<br/>Functions]
        Events[Event-Driven<br/>Workloads]
    end

    Serving --> KServe
    Serving --> Functions
    Eventing --> Events
    Eventing --> KServe

---

Components

Component	Purpose
Knative Serving	Request-driven compute, scale-to-zero
Knative Eventing	Event-driven architecture, CloudEvents

---

Why Knative?

Feature	Benefit
Scale-to-zero	Cost savings for idle workloads
Auto-scaling	Handles traffic spikes automatically
Revisions	Traffic splitting, canary deployments
CloudEvents	Standard event format
KServe foundation	Required for ML model serving

---

Configuration

Helm Values

knative-serving:
  enabled: true
  config:
    network:
      ingress-class: "cilium"
      domain-template: "{{.Name}}.{{.Namespace}}.{{.Domain}}"
    autoscaler:
      enable-scale-to-zero: "true"
      scale-to-zero-grace-period: "30s"
      stable-window: "60s"

knative-eventing:
  enabled: true
  config:
    default-ch-webhook:
      default-ch-config: |
        clusterDefault:
          apiVersion: messaging.knative.dev/v1
          kind: InMemoryChannel        

---

Knative Service Example

apiVersion: serving.knative.dev/v1
kind: Service
metadata:
  name: my-service
  namespace: default
spec:
  template:
    metadata:
      annotations:
        autoscaling.knative.dev/min-scale: "0"
        autoscaling.knative.dev/max-scale: "10"
    spec:
      containers:
        - image: harbor.<location-code>.<sovereign-domain>/my-app:latest
          ports:
            - containerPort: 8080
          resources:
            requests:
              cpu: 100m
              memory: 128Mi

---

Traffic Splitting

apiVersion: serving.knative.dev/v1
kind: Service
metadata:
  name: my-service
spec:
  template:
    metadata:
      name: my-service-v2
    spec:
      containers:
        - image: harbor.<location-code>.<sovereign-domain>/my-app:v2
  traffic:
    - revisionName: my-service-v1
      percent: 90
    - revisionName: my-service-v2
      percent: 10

---

Event-Driven Architecture

apiVersion: eventing.knative.dev/v1
kind: Broker
metadata:
  name: default
  namespace: default
---
apiVersion: eventing.knative.dev/v1
kind: Trigger
metadata:
  name: my-trigger
spec:
  broker: default
  filter:
    attributes:
      type: my.event.type
  subscriber:
    ref:
      apiVersion: serving.knative.dev/v1
      kind: Service
      name: my-service

---

Integration with Cilium

Knative uses Cilium as the ingress class:

config:
  network:
    ingress-class: "cilium"

---

Monitoring

Metric	Query
Request count	revision_request_count
Request latency	revision_request_latencies
Pod count	autoscaler_actual_pods
Desired pods	autoscaler_desired_pods

---

Consequences

Positive:

• Scale-to-zero reduces costs
• Automatic scaling for traffic spikes
• Foundation for KServe model serving
• Event-driven architecture support
• Traffic splitting for canary deployments

Negative:

• Cold start latency for scale-to-zero
• Additional complexity
• Requires understanding of Knative concepts

---

Part of OpenOva