Using Google Cloud Deploy execution environments

A Google Cloud Deploy execution environment is the environment in which Google Cloud Deploy executes its render, deploy, and verify operations. The execution environment consists of the following components:

  • The Cloud Build worker pool (default or private) in which Google Cloud Deploy executes render, deploy, and verify operations

  • The service account (default or alternate) that calls Google Cloud Deploy to perform these actions

  • The storage location (default or alternate) for rendered manifests in Cloud Storage

  • The Cloud Build timeout for operations (default or custom)

This article describes the default execution environment, service accounts, and storage for Google Cloud Deploy, as well as why and how you can change these defaults.

Defaults

The following are the defaults that Google Cloud Deploy uses to run, to execute rendering and deployment, and to store assets such as rendered manifests:

  • Default worker pool

    By default, Google Cloud Deploy runs in the default Cloud Build worker pool. However, you can configure Google Cloud Deploy to use a Cloud Build private worker pool.

    For more details about worker pools, see the Cloud Build Overview of default pools and private pools.

  • Default execution service account

    By default, Google Cloud Deploy uses the default Compute Engine service account.

  • Default Google Cloud Deploy storage location

    This value is the Cloud Storage bucket where Google Cloud Deploy stores your rendered manifests. By default, Google Cloud Deploy creates a Cloud Storage bucket, in the same region as the Google Cloud Deploy resources, taking the following form:

    <location>.deploy-artifacts.<project ID>.appspot.com

  • Default Cloud Build timeout

    By default, Cloud Build has a timeout of 1 hour on operations it performs for Google Cloud Deploy. You can change that timeout in the execution environment specification in target configuration.

The sections that follow describe the circumstances under which you would change any of these values, and links to instructions for doing so.

About Cloud Build worker pools

The Google Cloud Deploy execution environment can use one of the following:

  • The Cloud Build default pool

    The default worker pool is a secure, hosted environment with access to the public internet. Render, deploy, and verify operations are executed in that pool, isolated from other workloads.

  • A private pool

    Private worker pools are private, dedicated pools that can be customized more than the default worker pool. That customization can include the ability to access resources in a private network. Like the default worker pool, private worker pools are hosted and fully managed by Cloud Build. These pools can scale up or scale down to zero, with no infrastructure to set up, upgrade, or scale.

    The Cloud Build Private pools overview describes default worker pools and private worker pools more thoroughly, including a table comparing their features.

Changing the Google Cloud Deploy execution environment

You might change the Google Cloud Deploy execution environment under the following circumstances:

  • You want to deploy to a private Google Kubernetes Engine cluster

  • You want render, deploy, or verify operations, or a combination of the three, to be performed in an environment that's isolated from other organizations.

  • You want these operations to be performed in an environment that isn't connected to the public internet.

  • You want separate environments for render and deploy.

  • You want to use a dedicated service account with permissions that are more specific to your use than the permissions available in the default service account.

  • You want to store rendered manifests in a location different from the default Cloud Storage bucket.

Configuration of all three parts of the execution environment (worker pool, service account, and storage) is done per target, in each target's YAML configuration.

Changing from the default pool to a private pool

You configure worker pools per target, so that the pool is used for RENDER, DEPLOY, or VERIFY (or a combination of the three) for that target only.

To use the default worker pool for both render and deploy operations, you don't need to do anything.

The following is a sample target configuration that specifies a private worker pool for DEPLOY, and the default worker pool for RENDER and VERIFY:

executionConfigs:
- usages:
  - DEPLOY
  workerPool: "projects/p123/locations/us-central1/workerPools/wp123"
- usages:
  - RENDER
  - VERIFY

For more information about how to configure private pools for targets, see Delivery pipeline configuration documentation.

Changing from the default to custom execution service account

As with the worker pool, you can specify an alternate service account to use for rendering or deploying (or both) per target. To do so, add the following line to the target configuration, after the workerPool element:

serviceAccount: "[name]@[project_name].iam.gserviceaccount.com"

The specified service account must include the clouddeploy.jobRunner role, as described in the Google Cloud Deploy service accounts document.

See Target definitions for more details on this configuration.

Changing the storage location

To change the storage bucket from the Google Cloud Deploy default, add the following line to the target definition in the workerPool stanza:

artifactStorage: "gs://[bucket_name]/[dir]"

This configuration changes where the rendered manifests are stored, but does not affect where the rendering source is stored.

Using Google Cloud Deploy in a VPC Service Controls perimeter

Google Cloud Deploy supports VPC Service Controls.

You can follow the VPC Service Controls quickstart to set up a service perimeter.

Limitations

  • You must use a Cloud Build private worker pool for the target's execution environment—not the default worker pool.

  • The project that contains the worker pool and the project that contains your Google Cloud Deploy resources must remain in the same VPC Service Controls security perimeter.

  • Any GKE cluster you deploy to in the VPC Service Controls perimeter must be a private cluster.

    To set up a private pool for a private cluster, see this tutorial.

What's next