:User guide for RDS Custom container series nodes

Key concepts

Step 1: Create an ACK Pro cluster

Create an ACK Managed Cluster Pro Edition. Then, on the Basic Information page of the cluster in the console, obtain the cluster ID, VPC information, and vSwitch information. For details, see Create an ACK managed cluster.

Skip this step if you already have an ACK managed cluster.

Step 2: Create an RDS Custom container node

1. Log on to the ApsaraDB RDS Console. In the navigation pane on the left, click RDS Custom > Custom Instance List.

2. Click Create Instance and configure the parameters on the buy page. For other parameter descriptions, see Create an RDS Custom instance.

   Parameter	Description
   Database type	Select RDS Custom.
   Region	Select the region where your ACK managed cluster is deployed.
   Product series	Select container series.
   ACK cluster	Select the ACK managed cluster that you created in Step 1.
   Connection	You must provide the KubeConfig the first time you create an RDS Custom container series node. Click Get Connection Information to open the ACK Cluster Information page. Select the Connection tab and click Get Permanent KubeConfig. From the intranet access section, copy the entire client-certificate-data. Paste the client-certificate-data information into the first input box in the Connection section on the RDS Custom buy page. From the intranet access section, copy the entire client-key-data. Paste the client-key-data information into the second input box in the Connection section on the RDS Custom buy page.
   VPC	Select the VPC of the ACK managed cluster.
   vSwitch	Select a vSwitch in the same availability zone as the container series node.
   Instance type	Select either AI node or Lingjun node.

3. Click Confirm Order and complete the payment.

4. Return to the console. In the navigation pane on the left, click RDS Custom > Custom Instance List. In the top navigation bar, select the region where you created the instance. Find the newly created instance based on its creation time.

   Note

   Instance creation takes 1 to 10 minutes. Refresh the page to check the status.

Step 3: Use container nodes in an ACK cluster

1. Log on to the Container Service for Kubernetes (ACK) console. In the navigation pane on the left, choose Clusters.

2. On the Clusters page, click the name of your target cluster. In the navigation pane on the left, choose Nodes > > Nodes to view node information.

3. In the navigation pane on the left, choose Basic Information.

4. On the Basic Information page, click Manage Cluster in Workbench in the upper-right corner.

5. In the terminal, run a kubectl command to verify the cluster connectivity.

   For example, the following command queries node information.

   kubectl get node vn-frzqv6g28****

   Expected output:

   NAME              STATUS        ROLES        AGE          VERSION
   vn-frzqv6g28****  Active        <none>       25m        V1.24.2-sharer.1.1-082dafbcabcd88

Step 4: Schedule pods to a container node

1. Log on to the Container Service for Kubernetes (ACK) console. In the navigation pane on the left, choose Clusters.

2. On the Clusters page, click the name of your target cluster. In the navigation pane on the left, choose Workloads > > Pods.

3. On the Pods page, click Create from YAML. Configure the template and then click Create.

   • Set the image address in the spec.containers.image parameter. For a list of image addresses for the RDS Custom container series in different regions.

     Note

     For users who work with large language models, RDS Custom provides container images for both SgLang and vLLM environments. You can also use images from your private image repository to launch containers.

   • To schedule a Pod to an RDS Custom container node, you must configure a specific nodeSelector and tolerations. For more information, see Configure nodeSelector and tolerations.

     Note

     To prevent system component Pods from consuming resources on RDS Custom container nodes, each node has the following default taints and labels:

     • Label: alibabacloud.com/virtual-node: true

     • Taint:

       effect: NoSchedule
       key: virtual-kubelet.io/provider
       value: aliclouddb

     • Set the nodeSelector to alibabacloud.com/virtual-node: "true". Alternatively, find the vnodeId of the target node and set rm.alibaba-inc.com/vnodeId: "vn-e45b77245bb09b5e".

     • Set the tolerations to:

       tolerations:
         - effect: NoSchedule
           key: virtual-kubelet.io/provider
           value: aliclouddb

   The following YAML is an example of a Pod configuration for an AI node:

   Note

   This example configures a Pod named test-pod, using resource limits and node affinity to deploy a containerized service.

   # Pod configuration
   apiVersion: v1
   kind: Pod
   metadata:
     name: test-pod   # Use a unique name to avoid conflicts.
   spec:
     containers:
     - command:
       - sh
       - -c
       - echo hello world; sleep infinity;
       image: aliclouddb-pub-registry-vpc.cn-beijing.cr.aliyuncs.com/aliclouddb-public/des-ai-nv:25.05-sglang0.4.6.post4-pytorch2.6-cu124-20250513-serverless
       imagePullPolicy: IfNotPresent
       name: sglang
       ports:
       - containerPort: 8000  # The port that the container listens on.
         name: restful
         protocol: TCP
       resources:
         requests:
           cpu: "20"       
           memory: "100Gi"   
         limits:
           cpu: "20"       
           memory: "100Gi"  
     restartPolicy: Always
     nodeSelector:
       alibabacloud.com/virtual-node: "true"   # Schedules the Pod only to virtual nodes with this label.
     tolerations:
     - effect: NoSchedule
       key: virtual-kubelet.io/provider
       value: aliclouddb

Related topics

What is RDS Custom?