tegwick
76cb9c7ad9
Declare maxReplicas activator 20->3, webhook 5->2, 3scale-kourier-gateway 10->3 and 10m CPU requests for all six substrate containers (220m->60m), per the resource-control node budget of 2026-10-21 (CUST-WP-0071, RESOURCE-WP-0008). Read-only BEFORE evidence under docs/evidence/2026-10-21-hpa-caps-before/. Not applied; railiance-cluster overlays must follow (handoff in T02). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Assistant: claude-code Assistant-Model: opus Assistant-Process: 8340@bnt-lap001 Assistant-Session: 2e192b3d-b3e8-4403-94d4-d55a2c8b476c
37 lines
1.2 KiB
YAML
37 lines
1.2 KiB
YAML
# Declared HPA ceilings for the Knative Serving + Kourier v1.22.0 substrate on a
|
|
# single node (railiance01). Strategic-merge patches over the upstream HPAs in
|
|
# serving-core.yaml and kourier.yaml; only maxReplicas changes. minReplicas (1)
|
|
# and the CPU target (100% of request) stay upstream.
|
|
# The HPAs scale on percent of the CPU request, so after the 10m request trim
|
|
# (cpu-requests.patch.yaml) the same load reads ~5x higher. These caps bound the
|
|
# worst-case scale-out on one node to 3+2+3 pods instead of 20+5+10.
|
|
# RAIL-KNATIVE-WP-0003; evidence: resource-control
|
|
# docs/evidence/2026-10-21-node-allocation-review.md ("Knative HPA scale-out
|
|
# headroom" 1520m -> 230m).
|
|
---
|
|
# upstream v1.22.0: minReplicas 1, maxReplicas 20
|
|
apiVersion: autoscaling/v2
|
|
kind: HorizontalPodAutoscaler
|
|
metadata:
|
|
name: activator
|
|
namespace: knative-serving
|
|
spec:
|
|
maxReplicas: 3
|
|
---
|
|
# upstream v1.22.0: minReplicas 1, maxReplicas 5
|
|
apiVersion: autoscaling/v2
|
|
kind: HorizontalPodAutoscaler
|
|
metadata:
|
|
name: webhook
|
|
namespace: knative-serving
|
|
spec:
|
|
maxReplicas: 2
|
|
---
|
|
# upstream v1.22.0: minReplicas 1, maxReplicas 10
|
|
apiVersion: autoscaling/v2
|
|
kind: HorizontalPodAutoscaler
|
|
metadata:
|
|
name: 3scale-kourier-gateway
|
|
namespace: kourier-system
|
|
spec:
|
|
maxReplicas: 3
|