Deploy and manage workloads

Edge Appliance uses Google Distributed Cloud Virtual for Bare Metal to provide you with a single-node Kubernetes cluster running on the appliance, with access to the appliance filesystem.

Your Anthos cluster is listed in the Google Cloud console. From there, you can deploy and manage workloads using Kubernetes.

Work with Kubernetes

Edge Appliance comes with Kubernetes pre-installed, including the kubectl command line interface.

The Kubernetes documentation is the best resource for information about deploying and managing workloads.

The cluster on your Edge Appliance uses a storage class name of standard. You must specify this value in your configuration file when setting up any workload with a persistent volume claim on the appliance.

Installed CLIs

The Edge Appliance comes with the following command-line interfaces installed:

  • bmctl for GKE on Bare Metal management.
  • kubectl to communicate with your Kubernetes cluster's control plane, using the Kubernetes API.
  • ta cluster to accomplish Edge Appliance-specific cluster management tasks.

You should not install or use the gcloud CLI on your appliance, as doing so requires logging in with your Google Account credentials and could pose a security risk.

Cluster commands

The following ta cluster commands are supported:

  • ta cluster print_token prints the cluster token.
  • ta cluster config sets up cluster configuration.
  • ta cluster create creates a cluster.
  • ta cluster reset resets the cluster state and removes GKE Hub Membership.
  • ta cluster reset --node_only resets the cluster state.
  • ta cluster upgrade upgrades cluster to new Anthos version.

The following cluster operation commands are supported with the bmctl tool:

  • bmctl get config
  • bmctl update

Configure DNS

To configure DNS for your cluster, refer to Configure DNS for a cluster in the GKE on Bare Metal documentation.

Offline vs. online capabilities

When Edge Appliance is not connected to the network, cluster management using the Google Cloud console is not available. In this case, you must use the CLI locally on the appliance.

The Google Cloud console interface does not save state, so any commands sent from the Google Cloud console while the appliance is offline will fail. You should retry your commands once the appliance is online.

Logs and metrics collected while the appliance is offline are not sent to Cloud Monitoring when the appliance comes back online.