When you have finished setting up cluster node machines,
[your network](https://docs.cloud.google.com/kubernetes-engine/distributed-cloud/bare-metal/docs/how-to/plan-ip-addresses.md.txt), and
the other [prerequisites](https://docs.cloud.google.com/kubernetes-engine/distributed-cloud/bare-metal/docs/installing/install-prereq.md.txt), you're
almost ready to install Google Distributed Cloud. The next step is to decide what
kinds of clusters to create and choose which tool to use.

## Choose a cluster type

You can create different kinds of clusters in Google Distributed Cloud, including
**admin** clusters (to control the resources of your clusters) and **user**
clusters (to run workloads). We recommend admin and user cluster deployments
if you have multiple clusters in the same data center that you want to manage
from a centralized place, and for larger deployments that need isolation between
different teams or between development and production workloads.

You can also run Google Distributed Cloud as a single **standalone** cluster,
which serves as a user cluster and as an admin cluster. A standalone cluster
supports the edge profile, which has significantly reduced system resource
requirements and is recommended for edge devices with high resource constraints.
In addition, Google Distributed Cloud lets you create **hybrid** clusters that
combine administration tasks and workloads, as well as controlling other user
clusters.

Each of these configurations has their own advantages and benefits. For more
information on deciding which configuration to develop, see
[Choose a deployment model](https://docs.cloud.google.com/kubernetes-engine/distributed-cloud/bare-metal/docs/installing/install-prep.md.txt).

## Choose a tool to create clusters

You have your choice of tools for creating clusters and managing
cluster lifecycle:

- The command-line tool `bmctl`, which you run on your admin workstation in your on-premises data center.
- The Google Cloud console, Google Cloud CLI, or [Terraform](https://www.terraform.io/). These standard tools use the GKE On-Prem API, which runs on Google Cloud infrastructure, and collectively they are referred to as the *GKE On-Prem API clients*.

For information on deciding which tool best suits your needs, see
[Choose a tool to create clusters](https://docs.cloud.google.com/kubernetes-engine/distributed-cloud/bare-metal/docs/installing/cluster-lifecycle-management-tools.md.txt).

## More information

For more information on creating and configuring clusters, see the
following:

- Admin clusters:

  - [Create admin clusters using GKE On-Prem API clients](https://docs.cloud.google.com/kubernetes-engine/distributed-cloud/bare-metal/docs/installing/creating-clusters/create-admin-cluster-api.md.txt)
  - [Create admin clusters using `bmctl`](https://docs.cloud.google.com/kubernetes-engine/distributed-cloud/bare-metal/docs/installing/creating-clusters/admin-cluster-creation.md.txt)
- User clusters

  - [Create user clusters using GKE On-Prem API clients](https://docs.cloud.google.com/kubernetes-engine/distributed-cloud/bare-metal/docs/installing/creating-clusters/create-user-cluster-api.md.txt)
  - [Create user clusters using `bmctl`](https://docs.cloud.google.com/kubernetes-engine/distributed-cloud/bare-metal/docs/installing/creating-clusters/user-cluster-creation.md.txt)
- [Create standalone clusters](https://docs.cloud.google.com/kubernetes-engine/distributed-cloud/bare-metal/docs/installing/creating-clusters/standalone-cluster-creation.md.txt)

- [Create hybrid clusters](https://docs.cloud.google.com/kubernetes-engine/distributed-cloud/bare-metal/docs/installing/creating-clusters/hybrid-cluster-creation.md.txt)

## About the creation process

The cluster creation process includes preflight checks and machine
initialization. If cluster creation fails after the machine initialization phase
(even if preflight checks passed without errors), you must
[delete the cluster](https://docs.cloud.google.com/kubernetes-engine/distributed-cloud/bare-metal/docs/how-to/reset-nodes.md.txt). This returns the node to
a clean state. After deleting the cluster, you can re-attempt to create the
cluster after making any needed configuration changes.

The cluster creation process runs health checks when the cluster has been
created. This last step verifies that the cluster is in good operating
condition. If the cluster doesn't pass all health checks, the create operation
fails. When all health checks pass, the create operation finishes successfully.

## Plan for scalability

Before you create a new cluster, you need to understand the dimensions that
affect scalability. Some dimensions, such as the number of IP addresses that you
can reserve for Pods in your cluster (`clusterNetwork.pods.cidrBlocks`), are
immutable, so you need to plan for them when you create a cluster. For more
information about scalability dimensions and how to scale up a cluster, see
[Scale up Google Distributed Cloud
clusters](https://docs.cloud.google.com/kubernetes-engine/distributed-cloud/bare-metal/docs/how-to/scale). For information about some of the
restrictions and recommendations for maximum settings for your cluster, see
[Quotas and limits](https://docs.cloud.google.com/kubernetes-engine/distributed-cloud/bare-metal/docs/limits.md.txt).