Skip to content

Aura Score

Jump to bottom
Anna edited this page Jun 25, 2026 · 1 revision

Module: Aura Score

Module flag: aura Enabled by default: yes

Overview

The Aura Score module computes a composite health and efficiency score (0–100) for Cloud Run services, Cloud SQL instances, BigQuery datasets, and GKE clusters. It surfaces the most actionable issues first, sorted worst-to-best.

Score bands:

Score Band Meaning
80–100 🟢 Green Healthy and efficient
50–79 🟡 Yellow Degraded or over-provisioned
0–49 🔴 Red Action required

Enable This Module

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

Enable Cloud Recommender integration for efficiency signals (optional — requires an extra IAM role):

RECOMMENDER_ENABLED=true GCP_PROJECT_ID=my-project ./aura-tracker-gcp --modules aura

Required GCP APIs

gcloud services enable monitoring.googleapis.com container.googleapis.com \
  run.googleapis.com sqladmin.googleapis.com bigquery.googleapis.com

For Recommender:

gcloud services enable recommender.googleapis.com

Required IAM Permissions

Role Tools that need it
roles/monitoring.viewer Metrics for all resource types
roles/container.viewer GKE control plane health
roles/run.viewer Cloud Run service metadata
roles/cloudsql.viewer Cloud SQL instance metadata
roles/bigquery.metadataViewer BigQuery dataset/job metadata
roles/recommender.viewer Cloud Recommender (only if RECOMMENDER_ENABLED=true)

Score Computation

Health Score (60% of total)

Resource Signals and weights
Cloud Run Error rate 50% + CPU utilization 30% + P99 latency 20%
Cloud SQL CPU utilization 40% + Memory utilization 30% + Disk utilization 30%
BigQuery Job failure rate 60% + Slot utilization 20% + Storage bytes 20%
GKE Node CPU 35% + Node memory 35% + Pod restart rate 20% + Control plane health 10%

Efficiency Score (40% of total)

  • Derived from utilization metrics
  • Override: If Cloud Recommender API is enabled, GCP's own cost recommendations replace the metric-based estimate
  • Idle resources score 20; over-provisioned resources score 45; well-sized resources score higher

Caching

Scores are cached in-process for 5 minutes per resource. Repeated calls within the cache window are free (no Cloud Monitoring reads).

Tools

gcp_get_aura_score

Computes the Aura Score for a single resource.

Parameters:

Name Required Description
project_id yes GCP project ID
resource_kind yes cloud_run, cloud_sql, bigquery, or gke
region yes Resource region
resource_name yes Resource name

Returns: Score (0–100), band (green/yellow/red), health score, efficiency score, list of actionable reasons (e.g. "CPU at 2% — consider min-instances=0"), Recommender recommendations (if enabled).

gcp_project_aura_summary

Scores all Cloud Run services, Cloud SQL instances, BigQuery datasets, and GKE clusters in the project. Returns results sorted worst-first.

Parameters:

Name Required Description
project_id yes GCP project ID
resource_kind no Filter to a single resource type

Returns: All scored resources with score, band, and top reason per resource. Total count per band.

Prompt Examples

Project health dashboard

"Give me an Aura Score summary for project my-project — sorted by worst score first."

Red resource investigation

"Which resources in project my-project have a red Aura Score? What's the top issue for each one?"

Cost optimization

"Which Cloud Run services in project my-project have an efficiency score below 50 — are any over-provisioned or idle?"

GKE health

"What is the Aura Score for the prod-cluster GKE cluster in us-central1, and what's causing any degradation?"

Related Modules

Module flag Why you'd combine it
cloudrun Drill into the service details behind a low score
monitoring Fetch the raw metrics that drive the Aura Score
gke Get cluster and node pool details for GKE scores
coverage Combine health score with observability completeness

Clone this wiki locally