Prepare to set up Cloud Service Mesh with proxyless gRPC services

Setting up Cloud Service Mesh includes several phases. This document describes the first phase: instructions for preparing to configure Cloud Service Mesh with proxyless gRPC applications. This document applies when you are using the load balancing APIs. However, we strongly recommend that you use the service routing APIs. The other phases are covered by the platform-specific guides listed in Continue the setup process later in this document.

Before you read this guide, familiarize yourself with the following documents, which provide an overview of using Cloud Service Mesh with proxyless gRPC applications:


Prepare your environment by completing the following tasks:

  1. Enable billing.
  2. Grant the required permissions.
  3. Enable the Traffic Director API for your project.
  4. Ensure that the service account has sufficient permissions to access the Traffic Director API.

The following sections provide instructions for each task.

Enable billing

Make sure that billing is enabled for your Google Cloud project. For more information, see Enable, disable, or change billing for a project.

Grant the required IAM permissions

You must have sufficient Identity and Access Management (IAM) permissions to create VM instances and modify a network to configure Cloud Service Mesh. If you have the role of project Owner or Editor (roles/owner or roles/editor) in the project where you are enabling Cloud Service Mesh, you automatically have the correct permissions.

Otherwise, you must have all of the IAM roles shown in the following table. If you have these roles, you also have their associated permissions, as described in the Compute Engine IAM documentation.

Task Required role
Set IAM policy for a service account. Service Account Admin
Enable Cloud Service Mesh. Service Usage Admin
Create networks, subnets, and load balancer components. Compute Network Admin
Add and remove firewall rules. Compute Security Admin
Create instances. Compute Instance Admin
Create and modify a GKE cluster, if using Pods. Cluster Admin
Allows access to service accounts. Service Account User

The Compute Engine VMs must have the scope. For more information, see Troubleshooting deployments that use proxyless gRPC.

With xDS v3, grant the service account used by Cloud Service Mesh gRPC clients the role roles/trafficdirector.client.

Enable the Traffic Director API


  1. In the Google Cloud console, go to the API Library page for your project.

    Go to the API Library

  2. In the Search for APIs & Services field, enter Traffic Director.

  3. In the search results list, click Traffic Director API. If you don't see the Traffic Director API listed, that means that you don't have the necessary permissions to enable the Traffic Director API.

  4. On the Traffic Director API page, click Enable.


Run the following command:

gcloud services enable

Enable the service account to access the Traffic Director API

When you set up your data plane and connect it to Cloud Service Mesh, your xDS clients connect to the xDS server. These xDS clients present a service account identity to the xDS server to ensure that communications between the data plane and the control plane are properly authorized.

For a Compute Engine VM, the xDS client uses the service account assigned to the VM.

You need the following permissions. The xDS protocol version is specified in the bootstrap file. Only xDS v3 is supported.

If you are using xDS v2, you must migrate to xDS v3. For information about how to migrate, see Migrate from xDS v2 to xDS v3.

When you use xDS v3, the service account used by your gRPC applications must have the trafficdirector.networks.reportMetrics and trafficdirector.networks.getConfigs permissions. You can use the IAM Cloud Service Mesh Client role (roles/trafficdirector.client), which wraps both permissions.


  1. In the Google Cloud console, go to the IAM & Admin page.

    Go to IAM & Admin

  2. Select your project.

  3. Identify the service account to which you want to add a role:

    • If the service account isn't already on the Members list, it doesn't have any roles assigned to it. Click Add and enter the email address of the service account.
    • If the service account is already on the Members list, it has existing roles. Select the service account, and then click the Roles tab.
  4. Expand the role. For the service account that you want to edit, click Edit.

  5. Select the Other > Cloud Service Mesh Client role.

  6. To apply the role to the service account, click Save.


Run the following command:

gcloud projects add-iam-policy-binding PROJECT \
    --member serviceAccount:SERVICE_ACCOUNT_EMAIL \

Replace the following:

  • PROJECT: enter gcloud config get-value project
  • SERVICE_ACCOUNT_EMAIL: the email associated with the service account

Next, follow this overall procedure for setting up proxyless gRPC applications in a service mesh:

  1. Update your gRPC clients to the latest version of gRPC, with the most recent patch.
  2. Update your clients' gRPC name resolver scheme when you create a channel and specify a Cloud Service Mesh bootstrap file.
  3. Configure Cloud Service Mesh and Cloud Load Balancing resources.

This document provides information for completing the first two steps. The configuration process that you use for step 3 depends on whether your deployment uses Compute Engine VMs or GKE network endpoint groups (NEGs).

Supported gRPC versions and languages

gRPC is an open source project and its release support is described on the gRPC FAQ. We recommend that you use the most recent version of gRPC to ensure that known security vulnerabilities are mitigated. This also ensures that your applications have access to the latest features supported by Cloud Service Mesh. The service mesh features supported in various implementations and versions of gRPC are listed on GitHub. For a list of gRPC languages and features supported with Cloud Service Mesh and proxyless gRPC services, see Cloud Service Mesh features.

Cloud Service Mesh maintains compatibility with current and supported versions of gRPC and endeavors to be compatible with gRPC versions less than one year old, subject to the Google Cloud Platform Terms of Service.

Update your gRPC clients

Update the gRPC library in your application to the version that supports the features that you require. For details, see the previous section.

Add the xDS name-resolver as a dependency to your gRPC applications. Requirements per language for Java and Go are shown in the following sections. Other languages don't have any additional requirements.

Java requirements

In Java, if you're using Gradle, add the grpc-xds dependency to your build.gradle file. Replace LATEST_GRPC_VERSION with the latest version of gRPC.

dependencies {
  runtimeOnly 'io.grpc:grpc-xds:LATEST_GRPC_VERSION'

If you're using Maven, add the following to the <dependencies> section of pom.xml. Replace LATEST_GRPC_VERSION with the latest version of gRPC.


Go requirements

If you're using Go, import the xds Go package.

Set gRPC name resolver to use xds

Set or change your gRPC applications to use the xds name resolution scheme in the target URI, rather than DNS or any other resolver scheme. You do this by using the prefix xds:/// in the target name when you create a gRPC channel. Load balancing for gRPC clients is on a per-channel basis.

Include the service name used in the target URI in the Cloud Service Mesh configuration. For example, in Java, you create the channel by using this structure, in which the service name is helloworld:


Create and configure a bootstrap file

The xds resolver scheme tells the gRPC application to connect to Cloud Service Mesh to obtain configuration information for the target service. Therefore, do the following:

  • Create a bootstrap file, as shown in the following sample. This file tells gRPC to connect to an xDS server (Cloud Service Mesh) to get the configuration for specific services.
  • Define an environment variable named GRPC_XDS_BOOTSTRAP, with the bootstrap filename as the value of the environment variable.

The setup instructions have examples that show how to generate the bootstrap file. For your convenience, you can use the latest version of Cloud Service Mesh gRPC bootstrap generator.

A bootstrap file containing the information needed to connect to Cloud Service Mesh must be included alongside the application. A sample bootstrap file looks like this:

  "xds_servers": [
      "server_uri": "",
      "channel_creds": [
          "type": "google_default"
      "server_features": ["xds_v3"]
  "node": {
    "id": "projects/123456789012/networks/default/nodes/b7f9c818-fb46-43ca-8662-d3bdbcf7ec18",
    "metadata": {
    "locality": {
      "zone": "us-central1-a"

The following table explains the fields in the bootstrap file.

Field Value and description
xds_servers A list of xDS servers. gRPC uses only the first one in the list.
server_uri Specify at least one. gRPC tries to connect to only the first xDS server in the list of xds_servers. The default value is
channel_creds Credentials to use with the xDS server.
type Use the value google_default. For more information about how credentials are obtained, see Getting started with authentication.
server_features A list of features supported by the server, such as xDS v3 support. The default value is empty.
node Information about the client connecting to the xDS server.

The id must be in the following format as shown in the preceding example:


Provide a unique string as the value of ID. This helps identify the gRPC client that is connecting to Cloud Service Mesh.

metadata Information specific to the xDS server.
TRAFFICDIRECTOR_NETWORK_NAME If the field is empty or not specified, then the value is set to default.
locality The Google Cloud zone in which the gRPC client is running.

Continue the setup process

After you complete the prerequisites described in this document, continue with one of these documents: