fix(neuron-mcp): blue/green deployment, memory limits, preStop drain hook

- Rename Deployment to neuron-mcp-blue with slot=blue label
- Add neuron-mcp-green Deployment scaffold (replicas=0, ready to receive new image)
- Service selector now routes by slot label; annotation tracks active slot
- Memory: 768Mi → 1536Mi limit, 256Mi → 512Mi request (fixes OOM pod restarts / 502s)
- Add JAVA_TOOL_OPTIONS: container-aware GC + ExitOnOutOfMemoryError
- Add preStop sleep 8s so Traefik drains in-flight connections before pod terminates
- terminationGracePeriodSeconds: 30
- Add scripts/blue-green-deploy.sh: automated slot-swap with health check and rollback

servers/legion/k8s/neuron-technologies/prod/mcp-deployment-green.yaml

@@ -0,0 +1,76 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: neuron-mcp-green
+  namespace: neuron-prod
+  labels:
+    app: neuron-mcp
+    slot: green
+    env: prod
+spec:
+  # Scaled to 0 when inactive. Deploy process scales to 1, verifies health, then switches service.
+  replicas: 0
+  strategy:
+    type: Recreate
+  selector:
+    matchLabels:
+      app: neuron-mcp
+      slot: green
+  template:
+    metadata:
+      labels:
+        app: neuron-mcp
+        slot: green
+        env: prod
+    spec:
+      terminationGracePeriodSeconds: 30
+      containers:
+        - name: neuron-mcp
+          # IMAGE UPDATED HERE before scaling to 1
+          image: registry.neuralplatform.ai/neuron-technologies/neuron-mcp:v0.8.3
+          imagePullPolicy: Always
+          ports:
+            - name: http
+              containerPort: 8080
+          env:
+            - name: OTEL_EXPORTER_OTLP_ENDPOINT
+              value: "http://alloy.monitoring.svc.cluster.local:4318"
+            - name: JAVA_TOOL_OPTIONS
+              value: "-XX:+UseContainerSupport -XX:MaxRAMPercentage=75.0 -XX:+ExitOnOutOfMemoryError"
+          envFrom:
+            - configMapRef:
+                name: neuron-prod-config
+            - secretRef:
+                name: neuron-prod-secrets
+          volumeMounts:
+            - name: data
+              mountPath: /data
+          resources:
+            requests:
+              cpu: 250m
+              memory: 512Mi
+            limits:
+              cpu: 1500m
+              memory: 1536Mi
+          lifecycle:
+            preStop:
+              exec:
+                command: ["sh", "-c", "sleep 8"]
+          livenessProbe:
+            httpGet:
+              path: /actuator/health/liveness
+              port: http
+            initialDelaySeconds: 30
+            periodSeconds: 30
+            failureThreshold: 3
+          readinessProbe:
+            httpGet:
+              path: /actuator/health/readiness
+              port: http
+            initialDelaySeconds: 10
+            periodSeconds: 10
+            failureThreshold: 3
+      volumes:
+        - name: data
+          persistentVolumeClaim:
+            claimName: neuron-prod-data

servers/legion/k8s/neuron-technologies/prod/mcp-deployment.yaml

@@ -1,10 +1,11 @@
 apiVersion: apps/v1
 kind: Deployment
 metadata:
-  name: neuron-mcp
+  name: neuron-mcp-blue
   namespace: neuron-prod
   labels:
     app: neuron-mcp
+    slot: blue
     env: prod
 spec:
   replicas: 1
@@ -13,14 +14,15 @@ spec:
   selector:
     matchLabels:
       app: neuron-mcp
+      slot: blue
   template:
     metadata:
       labels:
         app: neuron-mcp
+        slot: blue
         env: prod
-      annotations:
-        kubectl.kubernetes.io/restartedAt: "2026-04-24T10:00:00Z"
     spec:
+      terminationGracePeriodSeconds: 30
       containers:
         - name: neuron-mcp
           image: registry.neuralplatform.ai/neuron-technologies/neuron-mcp:v0.8.3
@@ -31,6 +33,8 @@ spec:
           env:
             - name: OTEL_EXPORTER_OTLP_ENDPOINT
               value: "http://alloy.monitoring.svc.cluster.local:4318"
+            - name: JAVA_TOOL_OPTIONS
+              value: "-XX:+UseContainerSupport -XX:MaxRAMPercentage=75.0 -XX:+ExitOnOutOfMemoryError"
           envFrom:
             - configMapRef:
                 name: neuron-prod-config
@@ -42,10 +46,15 @@ spec:
           resources:
             requests:
               cpu: 250m
-              memory: 256Mi
+              memory: 512Mi
             limits:
-              cpu: 1000m
-              memory: 768Mi
+              cpu: 1500m
+              memory: 1536Mi
+          lifecycle:
+            preStop:
+              exec:
+                # Give Traefik time to drain in-flight connections before pod terminates
+                command: ["sh", "-c", "sleep 8"]
           livenessProbe:
             httpGet:
               path: /actuator/health/liveness
@@ -57,7 +66,7 @@ spec:
             httpGet:
               path: /actuator/health/readiness
               port: http
-            initialDelaySeconds: 15
+            initialDelaySeconds: 10
             periodSeconds: 10
             failureThreshold: 3
       volumes:

servers/legion/k8s/neuron-technologies/prod/services.yaml

@@ -21,9 +21,14 @@ metadata:
   namespace: neuron-prod
   labels:
     app: neuron-mcp
+  annotations:
+    # Blue/green: change slot to "green" then back to "blue" to switch traffic.
+    # Process: deploy green → verify health → patch this selector → scale blue to 0.
+    neuron.ai/active-slot: "blue"
 spec:
   selector:
     app: neuron-mcp
+    slot: blue
   ports:
     - name: http
       port: 8080