Best practices for Config Connector

This page explains best practices you should consider when using Config Connector.

Manage API quota limits

If you've run into errors indicating that you've exceeded the API quota limit, it could be that you have created too many Config Connector resources of the same Kind under the same quota project. When you create a lot of resources, those resources can generate too many API requests to the same API endpoint because of the reconciliation strategy that Config Connector uses.

One way to resolve this issue is to request a quota increase. Besides a quota increase, if you've confirmed that the quota error is caused by GET requests against the Google Cloud resources managed by your Config Connector resources, you might consider one of the following options:

Increase the reconciliation interval

You can increase the time between Config Connector reconciling a resource to avoid hitting API quotas. The recommendation is to set the reconciliation interval to 1 hour.

To increase the reconciliation interval, follow the steps in Configuring the reconciliation interval.

Split your resources into multiple projects

This approach spreads out your Config Connector resources across different projects. This approach works well when adding new resources, but it can be risky to split existing resources because you need to delete the existing resources and recreate them under different projects. Deleting resources can cause data loss with some types of resources, such as SpannerInstance or BigtableTable resources. You should backup your data before deleting it.

To split existing Config Connector resources into different projects, complete the following steps:

  1. Decide which Config Connector resources you plan to move to different projects.
  2. Delete the Config Connector resources. Make sure the annotation is not set to abandon.
  3. Update the spec.projectRef field or annotation in the YAML configuration of the Config Connector resources you plan to move to the new projects.
  4. Grant the IAM service account used by Config Connector proper permissions on the new projects.
  5. Apply the updated YAML configuration to create the Config Connector resources.

Switch to namespaced mode

You can bind different IAM service accounts owned by different Google Cloud projects to different namespaces where Config Connector is installed in namespaced mode, and split your resources into different namespaces. To achieve this, complete the following steps:

  1. Configure Config Connector to run in namespaced mode. Create new IAM service accounts from different projects, and bind them to different namespaces following the instructions to configure Config Connector for each project.

  2. Grant the new IAM service accounts proper permissions to the project that contains the resources.

  3. Decide which Config Connector resources you plan to move to different namespaces.

  4. Update the YAML configuration of the Config Connector resources and set the annotation abandon.

  5. Apply the updated YAML configuration to update the Config Connector resources' deletion policy.

  6. Abandon the Config Connector resources.

  7. Update the metadata.namespace field in the YAML configuration of the Config Connector resources you plan to move to the different namespaces.

  8. Apply the updated YAML configuration to acquire the abandoned resources.

Manage node pools in GKE clusters

You might experience errors when you create a cluster by applying a ContainerCluster resource in Config Connector, and then attempt to update the nodeConfig or other node-related fields by applying an updated ContainerCluster configuration. These errors are due to immutible fields such as nodeConfig, nodeConfig.labels, nodeConfig.taint, which is a technical limitation of the underlying Google Cloud API.

If you need to update these fields, you can use the ContainerNodePool resource to manage node pools where these fields are not immutable. To manage node pools using the ContainerNodePool resource, you must specify an annotation "true". This annotation removes the default node pool that gets created during cluster creation. Then, to create separate node pools, specify nodeConfig fields in ContainerNodePool instead of in ContainerCluster. See the ContainerNodePool resource example for reference.

You should set the annotation absent for both the ContainerCluster and ContainerNodePool resources. This annotation avoids potential reconciliation errors during the interaction between the Config Connector controller and the underlying APIs.

The following examples show a ContainerCluster and a ContainerNodePool configuration with these annotations set:

kind: ContainerCluster
  name: containercluster-sample
  annotations: "true" absent
  description: A sample cluster.
  location: us-west1
  initialNodeCount: 1
kind: ContainerNodePool
    label-one: "value-one"
  name: containernodepool-sample
  annotations: absent
  location: us-west1
    minNodeCount: 1
    maxNodeCount: 3
    machineType: n1-standard-1
    preemptible: false
      - ""
      - ""
    name: containercluster-sample