Container Service for Kubernetes:Deploy a Stable Diffusion application in an ACK Serverless cluster

Log on to the ACK console. In the left navigation pane, click Clusters.

On the Clusters page, find your cluster and click its name. In the left navigation pane, choose Workloads > Deployments.

On the Deployments page, click Create from Image.

On the Basic Information step, set Name to stable-diffusion and Replicas to 1. Click Next.

On the Container step, configure the following parameters. Keep the default values for all other parameters. Click Next.

For the internal network image, replace <Region ID> with your cluster's region ID. Supported values: cn-beijing, cn-shanghai, cn-hangzhou, cn-shenzhen. The internal network endpoint pulls the image faster and avoids internet bandwidth charges.

On the Advanced page, click Create next to Services to expose the application. Configure the Service with the following settings:

Return to the Advanced page and click Create.

Go back to the Deployments page and click the application name. On the Pods tab, wait until the pod status changes to Running. Then click the Access Method tab and copy the IP address from the External Endpoint field.

The application typically takes several minutes to start. The Stable Diffusion model must load into memory before the readiness probe on port 7860 passes and the pod status changes to Running.

Connect to the ACK Serverless cluster using kubectl.

Create a file named stable-diffusion.yaml with the following content: The YAML uses a public image endpoint by default. To pull the image over an internal network, replace the image value with the endpoint for your cluster region:

The Deployment requests 8 vCPU, 16 GB memory, and 50 GiB ephemeral storage. These resources are needed to load and run the Stable Diffusion model for inference. Insufficient resources may cause out-of-memory errors or startup failures.

   apiVersion: apps/v1
   kind: Deployment
   metadata:
     labels:
       app: stable-diffusion
     name: stable-diffusion
     namespace: default
   spec:
     replicas: 1
     selector:
       matchLabels:
         app: stable-diffusion
     template:
       metadata:
         labels:
           app: stable-diffusion
       spec:
         containers:
         - args:
           - --listen
           - --skip-torch-cuda-test
           - --no-half
           command:
           - python3
           - launch.py
           image: yunqi-registry.cn-shanghai.cr.aliyuncs.com/lab/stable-diffusion:v1.2.0
           imagePullPolicy: IfNotPresent
           name: stable-diffusion
           ports:
           - containerPort: 7860
             protocol: TCP
           readinessProbe:
             tcpSocket:
               port: 7860
           resources:
             limits:
               cpu: "8"
               memory: 16Gi
               ephemeral-storage: 50Gi
   ---
   apiVersion: v1
   kind: Service
   metadata:
     annotations:
       service.beta.kubernetes.io/alibaba-cloud-loadbalancer-address-type: internet
       service.beta.kubernetes.io/alibaba-cloud-loadbalancer-instance-charge-type: PayByCLCU
     name: stable-diffusion-svc
     namespace: default
   spec:
     externalTrafficPolicy: Local
     ports:
     - port: 7860
       protocol: TCP
       targetPort: 7860
     selector:
       app: stable-diffusion
     type: LoadBalancer

Deploy the application: Expected output:

   kubectl apply -f stable-diffusion.yaml

   deployment.apps/stable-diffusion created
   service/stable-diffusion-svc created

Verify that the Deployment is ready: Expected output:

The pod may take several minutes to reach the 1/1 READY state. The application needs time to load the model into memory.

   kubectl get deployment stable-diffusion

   NAME               READY   UP-TO-DATE   AVAILABLE   AGE
   stable-diffusion   1/1     1            1           103m

Retrieve the external IP address of the Service: Expected output: Copy the EXTERNAL-IP value. Use this IP address to access the application in the next step.

   kubectl get service stable-diffusion-svc

   NAME                   TYPE           CLUSTER-IP     EXTERNAL-IP      PORT(S)          AGE
   stable-diffusion-svc   LoadBalancer   172.x.x.x      47.x.x.x        7860:xxxxx/TCP   103m