Deploying MaxAI STO Model

Create Amazon EFS Access Points for AION STO

To create EFS Access Point for AION STO, follow the steps below:

1. Log in to the AWS Management Console.
2. Navigate to EFS via the search bar.
3. Select Access Points from the left navigation menu.
4. Click Create Access Point.
5. Provide the following details:

   Field	Value
   File system	Use the default or an existing one
   Name (optional)	cdp-aion-sto
   Root directory path	/app/AION/model/sto
   User ID	0
   Group ID	0
   Secondary group IDs	Leave blank
   Owner User ID	0
   Owner Group ID	0
   Access point permissions	0775

6. Click Create Access Point.

Deploy STO Model via Devtron

To deploy the STO model, follow the steps below:

1. Step 1: Access Devtron
   • Open Devtron dashboard, and navigate to the Chart Store and select Generic Helm Chart.
2. Step 2: Configure and Deploy
   • Click Configure & Deploy.
   • Under the Configuration section, paste and modify the following YAML template according to your environment.
   • STO Helm Configuration Template is provided below.

     data:
       - apiVersion: v1
     	kind: PersistentVolume
     	metadata:
       	name: aion-sto-pv
     	spec:
       	capacity:
         	storage: 50Gi
       	accessModes:
         	- ReadWriteMany
       	persistentVolumeReclaimPolicy: Retain
       	storageClassName: efs-sc
       	csi:
         	driver: efs.csi.aws.com
         	volumeHandle: <AWS Access Point Id>::<AWS File System ID>
       - apiVersion: v1
     	kind: PersistentVolumeClaim
     	metadata:
       	name: aion-sto-pvc
     	spec:
       	accessModes:
         	- ReadWriteMany
       	resources:
         	requests:
           	storage: 50Gi
       	storageClassName: efs-sc
      
       - apiVersion: apps/v1
     	kind: Deployment
     	metadata:
       	name: cdp-sto-deployment
     	data:
       	config.ini: |
         	[AUTH]
         	client_id=test
         	client_secret=test
     	spec:
       	replicas: 1
       	selector:
         	matchLabels:
           	app: cdp-sto
       	template:
         	metadata:
           	labels:
             	app: cdp-sto
         	spec:
           	containers:
             	- name: cdp-sto-container
               	Image: { provide image repository: image tag}
               	imagePullPolicy: Always
               	ports:
                 	- containerPort: 8000
               	volumeMounts:
                 	- mountPath: /app/AION/config
                   	name: aion-recom-volume
                 	- mountPath: /app/AION/model
                   	name: aion-pvc-volume
         	
           	volumes:
             	- name: aion-recom-volume
               	configMap:
                 	name: aion-config
             	- name: aion-pvc-volume
               	persistentVolumeClaim:
                 	claimName: aion-sto-pvc
      
       - apiVersion: v1
     	kind: Service
     	metadata:
       	name: cdp-sto-service
     	spec:
       	ports:
         	- name: http
           	port: 8000
           	protocol: TCP
           	targetPort: 8000
       	selector:
         	app: cdp-sto
      
       - apiVersion: networking.k8s.io/v1
     	kind: Ingress
     	metadata:
       	Annotations: {Add annotation related to environment}
       	labels:
         	app.kubernetes.io/instance: cdp-sto
       	name: cdp-sto
     	spec:
       	ingressClassName: alb
       	rules:
         	- host: <HOST_DETAILS>
           	http:
             	paths:
               	- backend:
                   	service:
                     	name: cdp-sto-service
                     	port:
                       	number: 8000
                 	path: /*
                 	pathType: ImplementationSpecific
       - apiVersion: keda.sh/v1alpha1
     	kind: ScaledObject
     	metadata:
       	name: cdp-sto-scaledobject
       	namespace: default
     	spec:
       	scaleTargetRef:
         	name: cdp-sto-deployment
       	pollingInterval: 30
       	cooldownPeriod: 300
       	enabled: true
       	minReplicaCount: 1
       	maxReplicaCount: 5
       	idleReplicaCount: 0
       	restoreToOriginalReplicaCount: true
       	advanced:
         	horizontalPodAutoscalerConfig:
           	behavior:
             	scaleDown:
               	stabilizationWindowSeconds: 300
               	policies:
                 	- type: Percent
                   	value: 100
                   	periodSeconds: 60
       	triggers:
         	- type: prometheus
           	metadata:
             	serverAddress: <PROMETHEUS_SERVER_URL>
             	metricName: cpuusage
             	query: sum(rate(container_cpu_usage_seconds_total{pod=~"cdp-sto.*",container_name!="POD"}[5m]))*100
             	threshold: "80"
         	- type: prometheus
           	metadata:
             	serverAddress: <PROMETHEUS_SERVER_URL>
             	metricName: memoryusage
             	query: avg(container_memory_usage_bytes{pod=~"cdp-sto.*"}) / sum(kube_pod_container_resource_limits{pod=~"cdp-sto.*",resource="memory"}) * 100
             	threshold: "80"

     Note: Replace placeholders such as <IMAGE_REPOSITORY>, <IMAGE_TAG>, <AWS_ACCESS_POINT_ID>, <HOST_DETAILS>, <PROMETHEUS_SERVER_URL> and <EFS_FILE_SYSTEM_ID> with your actual values.

Post-Deployment Verification