Terraform (or OpenTofu) stores state about managed infrastructure to map real-world resources to the configuration, keep track of metadata, and improve performance. Terraform stores this state in a local file by default, but you can also use a Terraform remote backend to store state remotely. By default, cdktn init will configure a HCP Terraform workspace and a corresponding remote backend to store state for the new project. If you run cdktn init --local to configure your new project to use a local backend to store state, you can still migrate the state to a remote backend later. You can configure your CDK Terrain (CDKTN) remote backend to be HCP Terraform, another Terraform supported backend, or a custom location.

When to Use Remote Backends

Consider using a remote backend when multiple individuals or teams need access to your infrastructure state data. Remote state makes it easier for teams to work together because all members have access to the latest state data in the remote store. It also allows you to share output values with other configurations, allowing groups to share infrastructure resources. For example, a core infrastructure team can handle building the core machines and then expose some information that other teams can use for their own infrastructure.

Define Remote Backends

You can define a JSON configuration for a remote backend with a TerraformBackend subclass or a JSON configuration file. The following example uses the TerraformBackend subclass CloudBackend.

import { Construct } from "constructs";
import {
  CloudBackend,
  TerraformStack,
  TerraformOutput,
  NamedCloudWorkspace,
  App,
} from "cdktn";

export class CloudBackendStack extends TerraformStack {
  constructor(scope: Construct, id: string) {
    super(scope, id);

    new CloudBackend(this, {
      hostname: "app.terraform.io",
      organization: "company",
      workspaces: new NamedCloudWorkspace("my-app-prod"),
    });

    new TerraformOutput(this, "dns-server", {
      value: "hello-world",
    });
  }
}

When you call cdktn synth, CDKTN stores the backend metadata like the organization name and workspace name in the cdk.tf.json file within the cdktf.out stack sub-directory containing the synthesized CDKTN code. For example, CDKTN creates the output for a stack called hello-terraform in cdktf.out/stacks/hello-terraform. The following example shows the stack output directory.

tree .
.
└── cdk.tf.json

The following example shows a relevant snippet of the generated cdk.tf.json file.

{
  "//": {
    "metadata": {
      "backend": "cloud",
      "cloud": "tfc",
      "stackName": "hello-terraform",
      "version": "0.20.11"
    }
  },
  "terraform": {
    "cloud": {
      "hostname": "app.terraform.io",
      "organization": "company",
      "workspaces": {
        "name": "hello-terraform"
      }
    }
  }
}

Initialize Remote Backends

All cdktn operations perform an automatic terraform init, but you can also initialize manually. To manually initialize a remote backend, go to the corresponding stack output directory in the cdktf.out folder and run terraform init.

$ cd cdkf.out/stacks/hello-terraform
$ terraform init

Migrate Local State Storage to Remote

After you define your remote backend, you can migrate existing local state files to the designated remote location. This requires moving Terraform state files to the CDKTN output directory. Consider an example project called hello-terraform that is using local storage to store the Terraform state. To migrate the local stage files to the remote backend:

Navigate into the main project directory.
Use CloudBackend to add a new remote backend.

class LocalBackendStack extends TerraformStack {
  constructor(scope: Construct, id: string) {
    super(scope, id);

    new TerraformOutput(this, "dns-server", {
      value: "local",
    });
  }
}

const app = new App();
const stack = new LocalBackendStack(app, "local-to-cloud-backend");
new CloudBackend(stack, {
  hostname: "app.terraform.io",
  organization: "company",
  workspaces: new NamedCloudWorkspace("my-app-prod"),
});

Run cdktn diff <stack name> --migrate-state to migrate the state into HCP Terraform or Terraform Enterprise.

Initializing Terraform Cloud...
Migrating from backend "local" to Terraform Cloud.
Do you wish to proceed?
              As part of migrating to Terraform Cloud, Terraform can optionally copy your
              current workspace state to the configured Terraform Cloud workspace.

              Answer "yes" to copy the latest state snapshot to the configured
              Terraform Cloud workspace.

              Answer "no" to ignore the existing state and just activate the configured
              Terraform Cloud workspace with its existing state, if any.

              Should Terraform migrate your existing state?

              Enter a value:
yes
Initializing provider plugins...
            - Reusing previous version of hashicorp/random from the dependency lock file
- Using previously-installed hashicorp/random v3.4.3
Terraform Cloud has been successfully initialized!

Supported Backends

In addition to HCP Terraform, Terraform and CDKTN support the following backends.

local
```
new LocalBackend(stack, {...});
```
azurerm
```
new AzurermBackend(stack, {...});
```
consul
```
new ConsulBackend(stack, {...});
```
cos
```
new CosBackend(stack, {...});
```
gcs
```
new GcsBackend(stack, {...});
```
http
```
new HttpBackend(stack, {...});
```
oss
```
new OssBackend(stack, {...});
```
pg
```
new PgBackend(stack, {...});
```
s3
```
new S3Backend(stack, {...});
```

CDK Terrain v0.14 deprecated the artifactory, etcd, etcdv3, manta, and swift backends, and removed them in v0.20. Terraform removed these backends in v1.3. For migration paths from these removed backends, refer to Upgrading to Terraform v1.3.

State Locking with the S3 Backend

When several people or automated pipelines run against the same state, Terraform (or OpenTofu) uses state locking to prevent concurrent runs from corrupting the state file. The S3 backend supports two locking mechanisms:

S3-native locking (recommended) via the useLockfile property. Terraform writes a lock file alongside your state object in the same bucket — at <key>.tflock — and removes it when the operation completes. No additional infrastructure is required. S3-native locking is available in Terraform 1.10 and later and OpenTofu 1.10 and later.
DynamoDB locking via the dynamodbTable property, which points the backend at a DynamoDB table whose partition key is LockID. This is the legacy mechanism and is now deprecated in favor of useLockfile.

If you set neither property, state locking is disabled.

Enable S3-Native Locking

Set useLockfile to true on the S3Backend configuration.

import { Construct } from "constructs";
import { S3Backend, TerraformStack } from "cdktn";

export class S3BackendStack extends TerraformStack {
  constructor(scope: Construct, id: string) {
    super(scope, id);

    new S3Backend(this, {
      bucket: "my-terraform-state",
      key: "my-app/terraform.tfstate",
      region: "us-east-1",
      encrypt: true,
      useLockfile: true,
    });
  }
}

The identity that runs cdktn operations needs s3:GetObject, s3:PutObject, and s3:DeleteObject permissions on the lock file path so Terraform can create and release the lock.

S3-native locking requires Terraform 1.10 or OpenTofu 1.10 (or later). Declare the runtime versions your project supports in the targetVersions field of cdktf.json so that synthesis validates native locking against the versions you target — the default targets (terraform >=1.5.7, opentofu >=1.6.0) predate it. This is the same declarative, binary-free check CDK Terrain applies to function usage at synth time.

{
  "targetVersions": {
    "terraform": ">=1.10.0",
    "opentofu": ">=1.10.0"
  }
}

Migrate from DynamoDB to S3-Native Locking

If you already lock state with a DynamoDB table, you can move to S3-native locking without a window where state is unprotected by enabling both mechanisms for a single run, then removing the DynamoDB table.

Keep dynamodbTable and add useLockfile: true to your S3Backend configuration. With both set, Terraform acquires both locks.

new S3Backend(this, {
  bucket: "my-terraform-state",
  key: "my-app/terraform.tfstate",
  region: "us-east-1",
  dynamodbTable: "my-lock-table",
  useLockfile: true,
});

Run cdktn deploy (or cdktn diff) once so the change takes effect and the backend is reconfigured.
Remove dynamodbTable, leaving only useLockfile: true. After your next run no longer references the table, you can delete the DynamoDB table.

dynamodbTable remains available for backwards compatibility, but it is deprecated. Terraform deprecated the dynamodb_table argument in the S3 backend starting with v1.11 and may remove DynamoDB-based locking in a future release. Prefer useLockfile for new projects.

Escape Hatches

Escape hatches can add to or override existing resources, and you can use them for backends or backend constructs that CDKTN does not natively support. Escape hatch methods have an Override suffix (e.g., addOverride). The following example uses an escape hatch to add an unsupported remote backend on a Stack object.

stack.addOverride("terraform.backend", {
  atlas: {
    name: "example_corp/networking-prod",
    address: "https://app.terraform.io",
  },
});

​When to Use Remote Backends

​Define Remote Backends

​Initialize Remote Backends

​Migrate Local State Storage to Remote

​Supported Backends

​State Locking with the S3 Backend

​Enable S3-Native Locking

​Migrate from DynamoDB to S3-Native Locking

​Escape Hatches

When to Use Remote Backends

Define Remote Backends

Initialize Remote Backends

Migrate Local State Storage to Remote

Supported Backends

State Locking with the S3 Backend

Enable S3-Native Locking

Migrate from DynamoDB to S3-Native Locking

Escape Hatches