[backport v2.12.5] [BUG] Rancher GKE Cluster Creation Fails with Shared VPC Subnetworks #16041
Description
This is a backport issue for #15432, automatically created via GitHub Actions workflow initiated by @gaktive
Original issue body:
JIRA: SURE-10594
Issue Description:
When creating a GKE cluster in a Shared VPC setup, Rancher attempts to reference the subnetwork using only its short name (e.g., gcp-non-prod-us-east4). GCP rejects this with a 404 Not Found, because in Shared VPCs subnetworks belong to the host project and must be referenced using the fully qualified path:
Env:
Rancher: 2.12.0
local cluster: Helm,
Rke2 version: 1.31
Steps to Reproduce:
-
Set up a GKE cluster creation via Rancher using a Shared VPC subnetwork.
-
Apply the cluster config with only the short subnetwork name.
-
Observe the failure with a 404 error from GCP.
Expected Behavior:
Rancher should detect when the subnetwork is in a Shared VPC and automatically resolve or expand the short name to the fully qualified subnetwork resource path.
Workaround:
Manually update the cluster config YAML to include the fully qualified subnetwork path before applying.
Additional Information:
time="2025-08-04T19:45:48Z" level=error msg="Error recording gkecc [ (id: )] failure message: Operation cannot be fulfilled on gkeclusterconfigs.gke.cattle.io \"c-n4l99\": the object has been modified; please apply your changes to the latest version and try again, original error: googleapi: Error 404: Not found: project \"logics-2-0-nonprod\" does not have a subnetwork named \"gcp-non-prod-us-east4\" in region \"us-east4\".\nDetails:\n[\n {\n \"@type\": \"type.googleapis.com/google.rpc.RequestInfo\",\n \"requestId\": \"0x48e0f3fcc65a172c\"\n }\n]\n, notFound"