app/services/ai/ops/ml-deployment-optimizer.yaml

apiVersion: v1
kind: Namespace
metadata:
  name: aiops-system
  labels:
    name: aiops-system
    purpose: "intelligent-operations"
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: zenith-ml-deployment-optimizer
  namespace: aiops-system
  labels:
    app: ml-deployment-optimizer
    component: aiops
spec:
  replicas: 2
  selector:
    matchLabels:
      app: ml-deployment-optimizer
  template:
    metadata:
      labels:
        app: ml-deployment-optimizer
        component: aiops
    spec:
      containers:
      - name: ml-optimizer
        image: zenith/ml-deployment-optimizer:latest
        ports:
        - containerPort: 8080
          name: http
        - containerPort: 8081
          name: grpc
        env:
        - name: MODEL_PATH
          value: "/models/deployment-optimization"
        - name: PROMETHEUS_URL
          value: "http://prometheus.monitoring.svc.cluster.local:9090"
        - name: KUBERNETES_SERVICE_HOST
          valueFrom:
            fieldRef:
              fieldPath: status.hostIP
        - name: KAFKA_BOOTSTRAP_SERVERS
          value: "kafka.kafka.svc.cluster.local:9092"
        - name: REDIS_URL
          value: "redis://redis-cluster.zenith-production.svc.cluster.local:6379"
        - name: LOG_LEVEL
          value: "INFO"
        - name: OPTIMIZATION_INTERVAL
          value: "300s"
        - name: CONFIDENCE_THRESHOLD
          value: "0.85"
        resources:
          requests:
            cpu: 1000m
            memory: 2Gi
          limits:
            cpu: 4000m
            memory: 8Gi
        volumeMounts:
        - name: model-storage
          mountPath: /models
          readOnly: true
        - name: config-volume
          mountPath: /config
          readOnly: true
        livenessProbe:
          httpGet:
            path: /health
            port: 8080
          initialDelaySeconds: 60
          periodSeconds: 30
          timeoutSeconds: 10
        readinessProbe:
          httpGet:
            path: /ready
            port: 8080
          initialDelaySeconds: 30
          periodSeconds: 10
          timeoutSeconds: 5
      - name: ml-trainer
        image: zenith/ml-trainer:latest
        env:
        - name: TRAINING_INTERVAL
          value: "24h"
        - name: DATA_RETENTION_DAYS
          value: "30"
        - name: MODEL_BACKUP_PATH
          value: "/backup/models"
        resources:
          requests:
            cpu: 2000m
            memory: 4Gi
          limits:
            cpu: 6000m
            memory: 16Gi
        volumeMounts:
        - name: model-storage
          mountPath: /models
        - name: backup-storage
          mountPath: /backup
      volumes:
      - name: model-storage
        persistentVolumeClaim:
          claimName: ml-models-pvc
      - name: backup-storage
        persistentVolumeClaim:
          claimName: ml-backup-pvc
      - name: config-volume
        configMap:
          name: ml-optimizer-config
---
apiVersion: v1
kind: Service
metadata:
  name: ml-deployment-optimizer
  namespace: aiops-system
  labels:
    app: ml-deployment-optimizer
spec:
  selector:
    app: ml-deployment-optimizer
  ports:
  - name: http
    port: 8080
    targetPort: 8080
    protocol: TCP
  - name: grpc
    port: 8081
    targetPort: 8081
    protocol: TCP
  type: ClusterIP
---
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: ml-models-pvc
  namespace: aiops-system
spec:
  accessModes:
    - ReadWriteOnce
  storageClassName: gp3
  resources:
    requests:
      storage: 100Gi
---
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: ml-backup-pvc
  namespace: aiops-system
spec:
  accessModes:
    - ReadWriteOnce
  storageClassName: gp3
  resources:
    requests:
      storage: 500Gi
---
apiVersion: v1
kind: ConfigMap
metadata:
  name: ml-optimizer-config
  namespace: aiops-system
data:
  optimization-config.yaml: |
    # ML Deployment Optimization Configuration
    models:
      deployment_success:
        type: "gradient_boosting"
        features:
          - cpu_utilization
          - memory_usage
          - request_rate
          - error_rate
          - deployment_size
          - complexity_score
          - test_coverage
        target: "deployment_success_rate"
        retrain_interval: "24h"
        
      performance_prediction:
        type: "lstm"
        features:
          - historical_performance
          - traffic_patterns
          - resource_usage
          - error_trends
          - deployment_frequency
        target: "response_time_p95"
        sequence_length: 100
        retrain_interval: "168h"
      
      resource_optimization:
        type: "neural_network"
        features:
          - pod_cpu_utilization
          - pod_memory_usage
          - request_latency
          - throughput
          - error_rate
          - cost_metrics
        targets:
          - optimal_cpu_limit
          - optimal_memory_limit
          - optimal_replica_count
        retrain_interval: "72h"
    
    optimization_strategies:
      rollout_strategy:
        enabled: true
        weights:
          success_rate: 0.4
          performance: 0.3
          cost: 0.2
          risk: 0.1
        max_rollout_percentage: 20
        min_promotion_delay: "300s"
        
      resource_tuning:
        enabled: true
        cpu_buffer_percentage: 10
        memory_buffer_percentage: 15
        auto_scale_range:
          min_replicas: 2
          max_replicas: 50
        optimization_window: "3600s"
        
      failure_prevention:
        enabled: true
        prediction_horizon: "1800s"
        confidence_threshold: 0.85
        auto_rollback_threshold: 0.95
        alternative_strategies:
          - canary
          - blue_green
          - rolling
    
    monitoring:
      metrics_collection:
        interval: "30s"
        retention: "30d"
        prometheus_query_timeout: "10s"
        
      model_performance:
        accuracy_threshold: 0.90
        drift_detection_interval: "3600s"
        alert_on_degradation: true
        
      business_metrics:
        - deployment_success_rate
        - mean_recovery_time
        - cost_savings
        - performance_improvement
        - user_satisfaction
---
apiVersion: batch/v1
kind: CronJob
metadata:
  name: ml-model-retraining
  namespace: aiops-system
spec:
  schedule: "0 2 * * *"  # Daily at 2 AM
  jobTemplate:
    spec:
      template:
        spec:
          containers:
          - name: model-trainer
            image: zenith/ml-trainer:latest
            command:
            - /bin/sh
            - -c
            - |
              echo "Starting ML model retraining..."
              
              # Retrain deployment success model
              python3 -m zenith.ml.train \
                --model deployment_success \
                --data-source prometheus \
                --lookback-days 30 \
                --output-path /models/deployment_success_latest.pkl
              
              # Retrain performance prediction model
              python3 -m zenith.ml.train \
                --model performance_prediction \
                --data-source prometheus \
                --lookback-days 90 \
                --sequence_length 100 \
                --output-path /models/performance_prediction_latest.pkl
              
              # Retrain resource optimization model
              python3 -m zenith.ml.train \
                --model resource_optimization \
                --data-source prometheus \
                --lookback-days 60 \
                --output-path /models/resource_optimization_latest.pkl
              
              echo "Model retraining completed"
              
              # Validate models
              python3 -m zenith.ml.validate \
                --models deployment_success,performance_prediction,resource_optimization \
                --threshold 0.90
              
              echo "Model validation completed"
            env:
            - name: PROMETHEUS_URL
              value: "http://prometheus.monitoring.svc.cluster.local:9090"
            - name: KUBERNETES_CONFIG
              value: "/config/kubeconfig"
            volumeMounts:
            - name: model-storage
              mountPath: /models
            - name: config-volume
              mountPath: /config
            resources:
              requests:
                cpu: 4000m
                memory: 8Gi
              limits:
                cpu: 8000m
                memory: 16Gi
          volumes:
          - name: model-storage
            persistentVolumeClaim:
              claimName: ml-models-pvc
          - name: config-volume
            configMap:
              name: ml-optimizer-config
          restartPolicy: OnFailure
---
apiVersion: monitoring.coreos.com/v1
kind: ServiceMonitor
metadata:
  name: ml-deployment-optimizer-metrics
  namespace: aiops-system
spec:
  selector:
    matchLabels:
      app: ml-deployment-optimizer
  endpoints:
  - port: http
    path: /metrics
    interval: 30s
    scrapeTimeout: 10s