GKE

Module: GKE

Module flag: gke Enabled by default: yes

---

Overview

The GKE module provides cluster-level visibility and node pool management. Use it to inventory clusters, inspect node pool configuration, diagnose CPU/memory/error bottlenecks, and — with explicit confirmation — resize a node pool.

For Kubernetes workload-level visibility (Deployments, StatefulSets, Services, Ingresses), see GKE Workloads.

---

Enable This Module

GCP_PROJECT_ID=my-project ./aura-tracker-gcp --modules gke

---

Required GCP APIs

gcloud services enable container.googleapis.com monitoring.googleapis.com logging.googleapis.com

API	Used by
container.googleapis.com	All GKE tools
monitoring.googleapis.com	gcp_gke_get_cluster_bottlenecks (CPU/memory metrics)
logging.googleapis.com	gcp_gke_get_cluster_bottlenecks (error log signals)

---

Required IAM Permissions

Role	Tools that need it
roles/container.viewer	All read-only tools
roles/monitoring.viewer	gcp_gke_get_cluster_bottlenecks
roles/logging.viewer	gcp_gke_get_cluster_bottlenecks
roles/container.nodePoolAdmin	gcp_gke_scale_deployment (mutation)

---

Tools

gcp_gke_list_clusters

Lists GKE clusters in the project.

Parameters:

Name	Required	Description
project_id	yes	GCP project ID
location	no	Region or zone; - for all locations

Returns: Cluster name, location, Kubernetes version, node pool count, endpoint, status, labels.

---

gcp_gke_get_cluster_details

Gets full configuration for a single GKE cluster.

Parameters:

Name	Required	Description
project_id	yes	GCP project ID
location	yes	Region or zone
cluster_name	yes	Cluster name

Returns: Node pools (name, machine type, node count, autoscaling config), network, subnetwork, addons, workload identity, release channel, control plane endpoint, labels.

---

gcp_gke_get_cluster_bottlenecks

Identifies CPU, memory, and error-rate bottlenecks across all node pools in a cluster. Fans out 4 parallel metric/log queries within a shared 30-second timeout budget.

Parameters:

Name	Required	Description
project_id	yes	GCP project ID
location	yes	Region or zone
cluster_name	yes	Cluster name

Returns: Per-node-pool CPU utilization, memory utilization, pod restart rate, and recent error signals. Bottlenecks are ranked by severity.

---

gcp_gke_scale_deployment

Resizes a GKE node pool. Mutation — requires two-step confirmation.

Note: This tool resizes node pools via the GKE management API. Scaling Kubernetes Deployments (replica count) requires kubectl and is not currently implemented.

Parameters:

Name	Required	Description
project_id	yes	GCP project ID
location	yes	Region or zone
cluster_name	yes	Cluster name
node_pool_name	yes	Node pool name
node_count	yes	Desired node count
dry_run	no	true → preview only (Step 1)
confirm_plan_id	no	Plan ID from dry-run (Step 2)

Returns (dry_run=true): Description of the resize, plan_id, expires_in. Returns (confirmed): Previous count, new count, node pool name.

See Security and Safety#Two-Step Confirmation Flow.

---

Prompt Examples

Inventory

"List all GKE clusters in project my-project and their Kubernetes versions."

Bottleneck diagnosis

"Which node pools in the prod-cluster cluster in us-central1 are under CPU pressure?"

Autoscaling audit

"Which GKE clusters in project my-project do not have autoscaling enabled on any node pool?"

Node pool resize (safe two-step)

"I need to scale the general-pool node pool in prod-cluster (us-central1) from 3 to 5 nodes. Show me a preview first."

---

Related Modules

Module flag	Why you'd combine it
gke_workloads	List Kubernetes Deployments, StatefulSets, Services, Ingresses
gke_mesh	Inspect Istio/ASM service-to-service traffic topology
monitoring	Fetch raw CPU/memory metrics for cluster nodes
aura	Get composite health/efficiency score for GKE clusters