feat(DM01-6276): add keepAlive option to GKE cluster service

infrabox/local-dev/docker-compose.yml

@@ -16,7 +16,7 @@ services:
             - "5432:5432"
 
     minio:
-        image: minio/minio
+        image: quay.io/minio/minio
         command: server /data
         environment:
             - MINIO_ACCESS_KEY=AKIAIOSFODNN7EXAMPLE

infrabox/test/api/docker-compose.yml

@@ -7,7 +7,7 @@ services:
             dockerfile: ./src/postgres/Dockerfile
 
     minio:
-        image: minio/minio
+        image: quay.io/minio/minio
         command: server /data
         environment:
             - MINIO_ACCESS_KEY=AKIAIOSFODNN7EXAMPLE

infrabox/utils/storage/docker-compose.yml

@@ -2,7 +2,7 @@ version: "3.2"
 
 services:
     minio:
-        image: minio/minio
+        image: quay.io/minio/minio
         command: server /data
         environment:
             - MINIO_ACCESS_KEY=AKIAIOSFODNN7EXAMPLE

src/services/gcp/README.md

@@ -84,6 +84,69 @@ or using
 export KUBECONFIG="/var/run/infrabox.net/services/$SERVICE_NAME/kubeconfig"
 ```
 
+## Spec Reference
+
+| Field | Type | Default | Description |
+|---|---|---|---|
+| `zone` | string | — | GCP zone (required) |
+| `machineType` | string | — | GKE node machine type |
+| `numNodes` | int | — | Number of nodes |
+| `diskSize` | int | — | Boot disk size in GB |
+| `clusterVersion` | string | — | Kubernetes version (e.g. `1.37`) |
+| `preemptible` | bool | false | Use preemptible nodes |
+| `enableAutoscaling` | bool | false | Enable cluster autoscaling |
+| `minNodes` | int | — | Minimum nodes (requires `enableAutoscaling`) |
+| `maxNodes` | int | — | Maximum nodes (requires `enableAutoscaling`) |
+| `enableNetworkPolicy` | bool | false | Enable Kubernetes network policy |
+| `disableLegacyAuthorization` | bool | false | Disable legacy ABAC authorization |
+| `enablePodSecurityPolicy` | bool | false | Enable Pod Security Policy |
+| `stackType` | string | — | Set to `ipv4-ipv6` for dual-stack networking |
+| `enableManagedPrometheus` | bool | false | Enable Google Managed Prometheus |
+| `serviceCidr` | string | `/18` | Services IPv4 CIDR |
+| `clusterCidr` | string | `/18` | Pods IPv4 CIDR |
+| `keepAlive` | bool | false | Keep cluster running after job ends (see below) |
+
+## Long-Running Clusters (keepAlive)
+
+By default, GKE clusters are deleted as soon as the InfraBox job finishes. Set `keepAlive: true` to keep the cluster running after the job ends — useful for debugging or extended testing scenarios.
+
+```json
+"services": [{
+    "apiVersion": "gcp.service.infrabox.net/v1alpha1",
+    "kind": "GKECluster",
+    "metadata": {
+        "name": "my-cluster"
+    },
+    "spec": {
+        "zone": "us-east1-b",
+        "machineType": "n1-standard-1",
+        "numNodes": 1,
+        "diskSize": 100,
+        "clusterVersion": "1.37",
+        "keepAlive": true
+    }
+}]
+```
+
+When `keepAlive: true`:
+- The cluster is tagged with the GCP label `infrabox-keep-alive=true` at creation time
+- InfraBox removes its CRD record when the job ends but does **not** delete the GKE cluster
+- The cluster is excluded from InfraBox's GC loop
+
+**Finding keepAlive clusters:**
+```bash
+gcloud container clusters list \
+    --filter="labels.infrabox-keep-alive=true" \
+    --format="table(name,zone,status,createTime)"
+```
+
+**Deleting a keepAlive cluster:**
+```bash
+gcloud container clusters delete <name> --zone <zone>
+```
+
+> **Note:** keepAlive clusters are not managed by InfraBox after the job ends. You are responsible for deleting them to avoid ongoing GCP costs.
+
 ## Install
 To install the service in your Kubernetes cluster you have to first create a GCP Service Account with `Kubernetes Engine Admin` and `Service Account User` roles.
 Download the service account json file and save it as `service_account.json`. Then create a secret for it:

src/services/gcp/infrabox-service-gcp/templates/crd.yaml

@@ -54,6 +54,8 @@ spec:
                   type: boolean
                 stackType:
                   type: string
+                keepAlive:
+                  type: boolean
             status:
               x-kubernetes-preserve-unknown-fields: true
   names:

src/services/gcp/pkg/apis/gcp/v1alpha1/types.go

@@ -32,6 +32,7 @@ type GKEClusterSpec struct {
 	ServiceCidr                string `json:"serviceCidr,omitempty"`
 	ClusterCidr                string `json:"clusterCidr,omitempty"`
 	EnableManagedPrometheus    bool   `json:"enableManagedPrometheus,omitempty"`
+	KeepAlive                  bool   `json:"keepAlive,omitempty"`
 }
 
 type GKEClusterStatus struct {

src/services/gcp/pkg/stub/handler.go

@@ -240,6 +240,9 @@ func createCluster(cr *v1alpha1.GKECluster, log *logrus.Entry) (*v1alpha1.GKEClu
 	if !cr.Spec.EnableManagedPrometheus {
 		args = append(args, "--no-enable-managed-prometheus")
 	}
+	if cr.Spec.KeepAlive {
+		args = append(args, "--labels", "infrabox-keep-alive=true")
+	}
 	master_authorized_networks := os.Getenv("ALLOW_IPS")
 	if master_authorized_networks == "" {
 		master_authorized_networks = "0.0.0.0/0"
@@ -546,6 +549,15 @@ func checkTimeout(cr *v1alpha1.GKECluster, log *logrus.Entry) error {
 }
 
 func deleteGKECluster(cr *v1alpha1.GKECluster, log *logrus.Entry) error {
+	if cr.Spec.KeepAlive {
+		if err := cleanUpCrd(cr, log); err != nil {
+			log.Errorf("Failed to remove GKECluster CRD: %v", err)
+			return err
+		}
+		log.Infof("GKE cluster %s kept alive (keepAlive=true)", cr.Status.ClusterName)
+		return nil
+	}
+
 	// Get the GKE Cluster
 	gkecluster, err := getRemoteCluster(cr.Status.ClusterName, log)
 	if err != nil && !errors.IsNotFound(err) {
@@ -895,7 +907,7 @@ func cleanUpClusters(maxAge string, log *logrus.Entry) {
 
 func getOutdatedClusters(maxAge string, log *logrus.Entry) ([]RemoteCluster, error) {
 	cmd := exec.Command("gcloud", "container", "clusters", "list",
-		"--filter", fmt.Sprintf("createTime<-P%s AND name:ib-*", maxAge),
+		"--filter", fmt.Sprintf("createTime<-P%s AND name:ib-* AND NOT labels.infrabox-keep-alive=true", maxAge),
 		"--format", "json")
 
 	out, err := cmd.Output()