From 240620c3a4728d1faa04cd1e2d449884e39bcf19 Mon Sep 17 00:00:00 2001
From: Sirius Claw <sirius.claw@siriusdevops.local>
Date: Wed, 22 Apr 2026 17:07:05 +0000
Subject: [PATCH 1/2] feat: add KEDA scaling for A100 vLLM

- Scale to 0 when idle (saves money)
- 15 min cooldown before scale-down
- Triggers on 3+ pending HTTP requests
- Targets openclaw-brain-vllm deployment
---
 apps/base/customer1/openclaw/keda-vllm.yaml | 21 +++++++++++++++++++++
 1 file changed, 21 insertions(+)
 create mode 100644 apps/base/customer1/openclaw/keda-vllm.yaml

diff --git a/apps/base/customer1/openclaw/keda-vllm.yaml b/apps/base/customer1/openclaw/keda-vllm.yaml
new file mode 100644
index 0000000..09dadc2
--- /dev/null
+++ b/apps/base/customer1/openclaw/keda-vllm.yaml
@@ -0,0 +1,21 @@
+apiVersion: keda.sh/v1alpha1
+kind: ScaledObject
+metadata:
+  name: openclaw-brain-a100-scaling
+  namespace: customer1
+spec:
+  scaleTargetRef:
+    name: openclaw-brain-vllm
+    kind: Deployment
+  service:
+    name: openclaw-brain-service
+    port: 8000
+  minReplicaCount: 0
+  maxReplicaCount: 1
+  cooldownPeriod: 900  # 15 minutes idle before scale-down
+  triggers:
+    - type: http
+      metadata:
+        path: "/v1/models"
+        host: "openclaw-brain-service.customer1.svc.cluster.local"
+        requestCount: "3"  # Scale up when 3+ requests pending

From 9673b76fc74f04281235a56d098b1cfd1d5ec2c4 Mon Sep 17 00:00:00 2001
From: Sirius Claw <sirius.claw@siriusdevops.local>
Date: Wed, 22 Apr 2026 17:22:22 +0000
Subject: [PATCH 2/2] fix: trigger KEDA scale-up on 1 request instead of 3

- Wake script handles cold start, so scale immediately on first request
- Reduces latency for /wake command and manual requests
---
 apps/base/customer1/openclaw/keda-vllm.yaml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/apps/base/customer1/openclaw/keda-vllm.yaml b/apps/base/customer1/openclaw/keda-vllm.yaml
index 09dadc2..c26cf69 100644
--- a/apps/base/customer1/openclaw/keda-vllm.yaml
+++ b/apps/base/customer1/openclaw/keda-vllm.yaml
@@ -18,4 +18,4 @@ spec:
       metadata:
         path: "/v1/models"
         host: "openclaw-brain-service.customer1.svc.cluster.local"
-        requestCount: "3"  # Scale up when 3+ requests pending
+        requestCount: "1"  # Scale up on 1+ request (wake script handles cold start)