All Products
Search

This Product

    Container Service for Kubernetes:Create a workflow cluster

    Document Center

    Container Service for Kubernetes:Create a workflow cluster

    Last Updated:Mar 26, 2026

    Workflow clusters run Argo Workflows on a serverless architecture, using Elastic Container Instance (ECI) as the compute layer. The cluster automatically schedules large-scale workflows with preemptible instances to keep costs low.

    Prerequisites

    Before you begin, ensure that you have:

    Create a workflow cluster

    Use the console

    1. Log on to the ACK One console.

    2. On the Workflow Cluster page, click Create Workflow Cluster in the upper-right corner. Configure the parameters in the panel that appears, then click Create.

      ParameterDescription
      Cluster NameEnter a name for the cluster. The name must be 1–63 characters long, can contain letters, digits, underscores (_), and hyphens (-), and must start with a letter.
      RegionSelect the region to deploy the cluster.
      VPCSelect a VPC from the drop-down list.
      vSwitchSelect a vSwitch from the drop-down list.
      SLB Instance for API ServerNo configuration required. A Server Load Balancer (SLB) instance of type slb.s2.small is created automatically. Deleting this instance makes the API server inaccessible. For billing details, see Billing overview.
      Create and Associate EIPControls whether an elastic IP address (EIP) is associated with the SLB instance. If enabled, the system creates an EIP and exposes the API server to the Internet. The EIP cannot be disassociated afterward. If disabled, the API server cannot be accessed over the Internet. For billing details, see Billing overview.
      Enable Collection of Operation Logs and Auditing LogsEnables Simple Log Service. When enabled, the system creates a service named k8s-log-{ClusterID} and collects logs from kube-apiserver and kube-controller-manager for audit purposes. For billing details, see Billing overview.

    Use the Alibaba Cloud CLI

    1. Set your environment variables. Replace the placeholder values with your actual resource IDs.

      export REGION_ID=cn-zhangjiakou
export VPC_ID=<your-vpc-id>
export VSWITCH_IDS='["<vsw-id-1>","<vsw-id-2>"]'
export CLUSTER_NAME=workflow1

    2. Configure the CLI region and create a workflow cluster.

      ParameterRequiredDescription
      ProfileYesSet to XFlow.
      RegionIdYesThe region where the cluster is deployed.
      VpcIdYesThe ID of the VPC.
      VSwitchesYesA JSON array of vSwitch IDs. Specify vSwitches in different zones for high availability.
      NameNoThe cluster name.
      IsEnterpriseSecurityGroupYesSet to true to use an advanced security group.
      ApiServerPublicEipNoSet to true to expose the API server endpoint over the Internet using an EIP.
      aliyun configure set --region $REGION_ID
aliyun adcp CreateHubCluster \
  --Profile XFlow \
  --RegionId $REGION_ID \
  --VpcId $VPC_ID \
  --VSwitches "$VSWITCH_IDS" \
  --Name $CLUSTER_NAME \
  --ApiServerPublicEip true \
  --IsEnterpriseSecurityGroup true

      The command returns output similar to:

      {
    "ClusterId": "c1234abcd...",
    "RequestId": "A1B2C3D4...",
    "TaskId": "T9876..."
}

      Save the ClusterId value — you need it in the following steps.

      export CLUSTER_ID=<your-cluster-id>

    3. Wait for the cluster to reach the running state before proceeding.

      aliyun adcp DescribeHubClusterDetails --ClusterId $CLUSTER_ID | jq .Cluster.ClusterInfo

      Rerun the command until the cluster status shows running.

    4. Generate a kubeconfig file and configure your environment to use it.

      aliyun adcp DescribeHubClusterKubeconfig --ClusterId $CLUSTER_ID \
  | jq -r .Kubeconfig \
  | tee ack-argo-workflow-kubeconfig

export KUBECONFIG=ack-argo-workflow-kubeconfig

      After exporting KUBECONFIG, kubectl and the Argo CLI use this cluster by default.

    Delete a workflow cluster

    Important

    Before deleting a workflow cluster, delete all workflows, pods, and Elastic Container Instances running in the cluster.

    Use the console

    1. Log on to the ACK One console.

    2. Select the cluster to delete from the drop-down list in the upper-left corner.

    3. On the Workflow Cluster page, click Delete Workflow Cluster in the upper-right corner. Click OK to confirm.

    Use the Alibaba Cloud CLI

    1. Disable Argo Server and delete the associated SLB instance and Elastic Container Instances.

      aliyun adcp UpdateHubClusterFeature --ArgoServerEnabled false --ClusterId $CLUSTER_ID

    2. Delete the workflow cluster.

      aliyun adcp DeleteHubCluster --ClusterId $CLUSTER_ID

    What's next

    References