Attempting to perform a GCP XPN internal cluster installation, the install fails when the master nodes are added to a second [internal] instance group (k8s-ig-xxxx).

always 

1. The following install config was used:

additionalTrustBundlePolicy: Proxyonly
apiVersion: v1
baseDomain: installer.gcp.devcluster.openshift.com
credentialsMode: Passthrough
featureSet: TechPreviewNoUpgrade
compute:
- architecture: amd64
  hyperthreading: Enabled
  name: worker
  platform: {}
  replicas: 3
controlPlane:
  architecture: amd64
  hyperthreading: Enabled
  name: master
  platform: {}
  replicas: 3
metadata:
  creationTimestamp: null
  name: bbarbach-xpn
networking:
  clusterNetwork:
  - cidr: 10.124.0.0/14
    hostPrefix: 23
  machineNetwork:
  - cidr: 10.128.0.0/16
  networkType: OVNKubernetes
  serviceNetwork:
  - 172.30.0.0/16
platform:
  gcp:
    projectID: openshift-installer-shared-vpc
    region: us-central1
    network: bbarbach-internal-vpc
    computeSubnet: bbarbach-internal-vpc
    controlPlaneSubnet: bbarbach-internal-vpc
    networkProjectID: openshift-dev-installer
publish: Internal

2. This is a shared VPC install so the service and host projects need to be used in the install-config above.

3. Set the release image to 4.13-nightly

4. openshift-install create cluster --log-level=DEBUG

ERROR                                              
ERROR Error: Error waiting for Updating RegionBackendService: Validation failed for instance 'projects/openshift-installer-shared-vpc/zones/us-central1-a/instances/bbarbach-xpn-4t8zl-master-0': instance may belong to at most one load-balanced instance group. 
ERROR                                              
ERROR                                              
ERROR   with google_compute_region_backend_service.api_internal, 
ERROR   on main.tf line 13, in resource "google_compute_region_backend_service" "api_internal": 
ERROR   13: resource "google_compute_region_backend_service" "api_internal" { 
ERROR                                              
FATAL failed disabling bootstrap load balancing: failed to apply Terraform: exit status 1 
FATAL                                              
FATAL Error: Error waiting for Updating RegionBackendService: Validation failed for instance 'projects/openshift-installer-shared-vpc/zones/us-central1-a/instances/bbarbach-xpn-4t8zl-master-0': instance may belong to at most one load-balanced instance group. 
FATAL                                              
FATAL                                              
FATAL   with google_compute_region_backend_service.api_internal, 
FATAL   on main.tf line 13, in resource "google_compute_region_backend_service" "api_internal": 
FATAL   13: resource "google_compute_region_backend_service" "api_internal" { 
FATAL                                              
FATAL  

Successful install

The normal GCP internal cluster installation succeeds. Checking the instance groups, the internal cluster creates the k8s-ig-xxxx instance groups where the workers are added to each respective group. The masters are NOT added to the instance groups. The failure during the xpn install occurs because these masters are added to the instance groups.