GCP Source Plugin Configuration Reference


This example connects a single GCP project to a Postgres destination. The (top level) source spec section is described in the Source Spec Reference.

kind: source
  # Source spec section
  name: "gcp"
  path: "cloudquery/gcp"
  version: "v7.4.0"
  destinations: ["postgresql"]

    # GCP Spec section described below
    project_ids: ["my-project"]

GCP Spec

This is the (nested) spec used by GCP Source Plugin

  • project_ids ([]string) (default: empty. will use all available projects available to the current authenticated account)

    Specify specific projects to connect to. If either folder_ids or project_filter is specified, these projects will be fetched in addition to the projects from the folder/filter.

  • service_account_key_json (string) (default: empty).

    GCP service account key content. Using service accounts is not recommended, but if it is used it is better to use environment or file variable substitution.

  • folder_ids ([]string) (default: empty).

    cloudquery will sync from all the projects in the specified folders, recursively. folder_ids must be of the format folders/<folder_id> or organizations/<organization_id>. This feature requires the resourcemanager.folders.list permission. By default cloudquery will also sync from subfolders recursively (up to depth 100) - to reduce this, set folder_recursion_depth to a lower value (or 0 to disable recursion completely). Mutually exclusive with project_filter.

  • folder_recursion_depth (int) (default: 100).

    the maximum depth to recurse into subfolders. 0 means no recursion (only the top-level projects in folders will be used for sync).

  • project_filter (string) (default: empty).

    A filter to determine the projects that are synced. For instance, to only sync projects where the name starts with how-, set project_filter to name:how-*. Another example is: "name:how-* OR name:test-*". For syntax and example queries refer to API Reference here (opens in a new tab). Mutually exclusive with folder_ids.

  • backoff_delay (int) (default: 30). If specified APIs will be retried with exponential backoff if they are rate limited. This is the max delay (in seconds) between retries.

  • backoff_retries (int) (default: 0). If specified APIs will be retried with exponential backoff if they are rate limited. This is the max number of retries.

  • enabled_services_only (bool) (default: false). If enabled CloudQuery will skip any resources that belong to a service that has been disabled or not been enabled. If you use this option on a large organization (with more than 500 projects) you should also set the backoff_retries to a value greater than 0 otherwise the sync could fail because of rate limiting.

GCP + Kubernetes (GKE)

kind: source
  name: gcp
  path: "cloudquery/gcp"
  version: "v7.4.0"
  destinations: ["<destination>"]
kind: source
  name: k8s
  path: "cloudquery/k8s"
  version: "v3.1.0"
  destinations: ["<destination>"]

Kubernetes users may see the following message when running the K8s plugin on GKE Clusters:

WARNING: the gcp auth plugin is deprecated in v1.22+, unavailable in v1.26+; use gcloud instead.

As part of an initiative to remove platform specific code from Kubernetes, authentication will begin to be delegated to authentication plugins, starting in version 1.26.

What does this mean for CloudQuery users?

CloudQuery does not use any specific resources which hinder the upgrade.


The easiest way to upgrade, is to install gke-gcloud-auth-plugin from gcloud components on Mac or Windows:

gcloud components install gke-gcloud-auth-plugin

and apt on Deb based systems:

sudo apt-get install google-cloud-sdk-gke-gcloud-auth-plugin


Mac or Linux:

gke-gcloud-auth-plugin --version


gke-gcloud-auth-plugin.exe --version

Switch authentication methods

Set the flag:


Update components:

gcloud components update

Force credential update:

gcloud container clusters get-credentials {$CLUSTER_NAME}

Now you should be able to use kubectl as normal, and you should no longer see the warning in the CloudQuery output.

For more information, read Google's press release (opens in a new tab).