You need to enable JavaScript to run this app.
Volcengine Kubernetes Engine

Volcengine Kubernetes Engine

Copy page
Download PDF
VCI Agent Sandbox
Installation guide
Copy page
Download PDF
Installation guide
This topic primarily describes the installation method for VCI Agent Sandbox.​
Prerequisites​
  • The vci-virtual-kubelet add-on has been installed in the cluster. For detailed instructions, see vci-virtual-kubelet.​
Procedure​
  1. Log in to the VKE console.​
  1. In the left navigation bar, click Application center > Application template. On the page that appears, click the AI tab.​
  1. Find agent-sandbox, hover the pointer over the application template card, and then click Deploy.​
  1. On the Deploy application page, specify the agent-sandbox parameter.​
  • ​
    Parameter​
    Description​
    Orchestration template​
    ​
    Application templates​
    Select a Chart version of agent-sandbox. We recommend the latest version.​
    Parameter configuration​
    The YAML file content of the VCI Agent Sandbox controller. Click the edit icon next to values.yaml to view the YAML file content.​
    tip​
    For the descriptions of the YAML file, see the values.yaml file content and parameter descriptions section of this topic.​
    Name​
    Specify a custom name for the application based on the on-screen naming rules.​
    Deployment environment​
    ​
    Cluster type​
    The type of cluster for which you want to deploy agent-sandbox. Select Managed version.​
    Cluster​
    Select the VCI-based cluster that you created in Prerequisites.​
    Namespace​
    The namespace where the agent-sandbox application is deployed in the target cluster.​
    Protocol​
    Read and understand the agreement before selecting it.​
    ​
  • values.yaml file content and parameter descriptions​
warning​
  • Update the userConfig configuration in the YAML file based on your business requirements. Resource settings can be verified and adjusted based on the actual concurrency scenario. For other parameters, refer to the YAML parameter descriptions to specify custom configurations, or keep the default values.​
  • The following section provides a YAML example and parameter descriptions for the latest Chart version of agent-sandbox. Feature parameters supported by the latest version are not built into some earlier versions. We recommend using the latest Chart version to obtain the latest features and performance optimizations.​
  • ​
    Complete YAML example
    YAML parameter description
    ​
    ​
    userConfig:​
    # The Access Key of Volcano Engine. It needs to be explicitly provided during the deployment of the application template.​
    ak: ""​
    # The Secret Key of Volcano Engine. It needs to be explicitly provided during the deployment of the application template.​
    sk: ""​
    region: "cn-beijing"​
    # The VCI API URL. For details, visit https://www.volcengine.com/docs/6460/1899855.​
    vciHost: "open.volcengineapi.com"​
    # The deployment mode. Valid values: Node (deployed on ECS nodes) and VirtualNode (deployed based on VCIs).​
    deployNodeType: "Node"​
    ​
    namespace:​
    # create controls whether the namespace is created by the current Chart.​
    create: true​
    name: agent-sandbox-system​
    ​
    controller:​
    enableExtensions: true​
    # If you do not specify whether to create a Service for the Sandbox CR, the defaultSandboxService determines whether to create a Service by default.​
    defaultSandboxService: true​
    replicaCount: 1​
    resources:​
    requests:​
    cpu: 500m​
    memory: 512Mi​
    ​
    sandboxClaimConcurrentWorkers: 32​
    sandboxWarmPoolConcurrentWorkers: 32​
    sandboxTemplateConcurrentWorkers: 32​
    sandboxConcurrentWorkers: 32​
    ​
    kubeApiQps: 100​
    kubeApiBurst: 200​
    ​
    image:​
    registry: vke-cn-beijing.cr.volces.com/vke​
    image: agent-sandbox-controller​
    tag: "v0.4.6-vke.1"​
    pullPolicy: Always​
    vciPodAnnotations:​
    vke.volcengine.com/burst-to-vci: enforce​
    ##### Parameters that are not specified will be omitted.​
    # clusterDomain: "cluster.local"​
    leaderElect: true​
    ​
    # leaderElectionNamespace: ""​
    # enableTracing: false​
    # enablePprof: false​
    # enablePprofDebug: false​
    # pprofBlockProfileRate: 1000000​
    # pprofMutexProfileFraction: 10​
    ​
    ##### The following Workers are active only if controller.enableExtensions=true.​
    ##### extraArgs is used to pass additional flags not listed above, such as Zap logging flags.​
    extraArgs: []​
    ​
    nodeSelector: {}​
    ​
    tolerations: []​
    ​
    affinity: {}​
    ​
    e2bConfig:​
    enableE2BServer: false​
    ​
    serviceAccountName: agent-sandbox-controller-e2b​
    # The domain name suffix of the E2B service. It will be passed to the control plane and data plane. This will affect domain name concatenation and access behavior.​
    domain: e2b.com​
    ​
    controlPlane:​
    name: agent-sandbox-control-plane​
    serviceName: agent-sandbox-control-plane​
    replicas: 1​
    resources:​
    requests:​
    cpu: 500m​
    memory: 512Mi​
    ​
    # Whether to enable control-plane authentication. After this feature is enabled, the provided API key will be validated during subsequent access to E2B APIs.​
    enableAuth: true​
    # The administrator key. If it is not specified, a random API key will be generated and saved in the e2b-key-store Secret.​
    adminKey: ""​
    ​
    # The idle connection timeout. It is a Go duration string, such as 600s.​
    idleTimeout: 600s​
    # The response write timeout. It is a Go duration string. If the value is small, long-running requests may be interrupted.​
    writeTimeout: 600s​
    # The sandbox lifecycle timeout. It is a Go duration string. If the value is small, APIs used for creating or modifying a sandbox may time out before completion.​
    sandboxTimeout: 600s​
    ​
    # The QPS limit for Kubernetes API access.​
    restClientQps: 50​
    # The maximum burst requests to the Kubernetes API.​
    restClientBurst: 100​
    ​
    service:​
    port: 8081​
    ​
    image:​
    name: control-plane-gateway​
    tag: v0.4.6-vke.1​
    pullPolicy: Always​
    ​
    dataPlane:​
    name: agent-sandbox-data-plane​
    serviceName: agent-sandbox-data-plane​
    ​
    replicas: 1​
    resources:​
    requests:​
    cpu: 500m​
    memory: 512Mi​
    ​
    service:​
    port: 8080​
    image:​
    name: data-plane-gateway​
    tag: v0.4.6-vke.1​
    pullPolicy: Always​
    ​
    agentRuntime:​
    image:​
    name: agent-runtime​
    tag: v0.0.4​
    ​
    crdUpgrade:​
    enabled: true​
    image:​
    tag: v0.1.0-vke.1​
    pullPolicy: Always​
    ​
    ​
  1. Click Confirm to start deployment.​
  • After the application is deployed, it is displayed on the Helm page. Select the target project, cluster, and namespace, find the application deployed based on the agent-sandbox template, and view the application status. If Deployed is displayed in the Status column, it indicates the application has been deployed.​
Validating results​
After the application is deployed, you can verify the installation result by using the following method:​
  1. On the details page of the agent-sandbox application, click the Overview tab and click the name of the associated cluster.​
  1. On the cluster management page, click Workload > Pod in the left navigation bar. Check whether a pod named agent-sandbox-controller in the Running state exists under agent-sandbox-system or the namespace you specify.​
  • ​​​
What to do next​
Expose the E2B service via the Ingress​
If the E2B service is enabled in the deployed VCI Agent Sandbox controller (e2bConfig.enableE2BServer: true), you need to use an ALB Ingress to route the root domain name and wildcard domain name of the E2B service to the control plane and data plane services. Then, point the DNS records of the root domain name and wildcard domain name to the IP address exposed by the ALB Ingress to expose the E2B service externally.​
  1. Create an ALB Ingress. The parameters mentioned below need to be configured as described. Other parameters can be configured by referring to Creating an ALB Ingress through the console.​
  • ​​​
  • ​
    Parameter​
    Description​
    Forwarding configuration​
    ​
    Custom forwarding rules​
    Configure the following two custom forwarding rules:​
    • Custom forwarding rule 1​
    • Forwarding conditions: Select Domain name and enter the root domain name configured in e2bConfig.domain (the default root domain name is e2b.com).​
    • Forwarding actions: Select Forward to and configure the following forwarding actions.​
    • Service name: Set the value to agent-sandbox-control-plane.​
    • Port: Only one port is available, which is the control-plane service port configured in the YAML file of the VCI Agent Sandbox controller.​
    • Weight: Keep the default value.​
    • Custom forwarding rule 2​
    • Forwarding conditions: Select Domain name and enter the wildcard domain name configured in e2bConfig.domain (the default wildcard domain name is *.e2b.com).​
    • Forwarding actions: Select Forward to and configure the following forwarding actions.​
    • Service name: Set the value to agent-sandbox-data-plane.​
    • Port: Only one port is available, which is the data-plane service port configured in the YAML file of the VCI Agent Sandbox controller.​
    • Weight: Keep the default value.​
    ​
  1. Configure DNS resolution. Point the DNS resolution (A records) of the root domain name and wildcard domain name to the IP address exposed by the ALB Ingress. This routes external traffic to ALB.​
  • You can use the capabilities of the VolcanoEngine DNS service to configure DNS resolution. You need to add an A record for both the root domain name and the wildcard domain name (a total of two records). For details, see Adding A records. ​
Last updated: 2026.07.24 16:18:07