2026-04-23 01:33:55 +00:00
|
|
|
apiVersion: http.keda.sh/v1alpha1
|
2026-04-23 01:22:28 +00:00
|
|
|
kind: HTTPScaledObject
|
2026-04-22 17:07:05 +00:00
|
|
|
metadata:
|
|
|
|
|
name: openclaw-brain-a100-scaling
|
|
|
|
|
namespace: customer1
|
|
|
|
|
spec:
|
2026-04-23 01:42:03 +00:00
|
|
|
# The host header the interceptor should match
|
2026-04-23 01:37:09 +00:00
|
|
|
|
2026-04-23 01:42:03 +00:00
|
|
|
# Target to scale + routing info (old style)
|
2026-04-22 17:07:05 +00:00
|
|
|
scaleTargetRef:
|
2026-04-23 01:42:03 +00:00
|
|
|
deployment: openclaw-brain-vllm # your Deployment name
|
|
|
|
|
service: openclaw-brain-service # your Service name
|
|
|
|
|
port: 8000 # ← change if your container port is different (e.g. 8080, 11434 for vLLM, etc.)
|
2026-04-23 01:43:57 +00:00
|
|
|
host: openclaw-brain-service.customer1.svc.cluster.local
|
2026-04-23 01:42:03 +00:00
|
|
|
# Replica settings
|
2026-04-23 01:37:09 +00:00
|
|
|
replicas:
|
|
|
|
|
min: 0
|
|
|
|
|
max: 1
|
|
|
|
|
|
2026-04-23 01:42:03 +00:00
|
|
|
# Scaling metric - best for "wake on first request" use case
|
|
|
|
|
targetPendingRequests: 1
|
2026-04-23 01:37:09 +00:00
|
|
|
|
2026-04-23 01:42:03 +00:00
|
|
|
# Cooldown before scaling down (15 minutes idle)
|
|
|
|
|
scaledownPeriod: 900
|
