How Google Cloud Recommender (Active Assist) identifies waste, suggests optimizations, and helps you act on cost, security, and performance improvements.
What Is Recommender?
Recommender (part of Active Assist) is a Google Cloud service that analyzes your resource usage and configuration, then generates actionable recommendations. It runs automatically in the background and surfaces insights in the Console, CLI, and API.
You do not need to install anything or configure anything. Recommendations appear based on your existing usage patterns.
Types of Recommendations
Recommender covers six categories:
| Category | What It Finds | Examples |
|---|---|---|
| Cost | Wasted spend, idle resources, optimization opportunities | Idle VMs, unused disks/IPs, oversized machine types, CUD purchase suggestions |
| Security | Overly permissive policies, security risks | IAM role recommendations, firewall rule cleanup, service account key rotation |
| Performance | Resources that could perform better with different configuration | Machine type changes, Cloud SQL wellness |
| Reliability | Availability and resilience risks | Single-zone dependencies, missing health checks |
| Sustainability | Carbon footprint reduction | Workload placement in lower-carbon regions |
| Management | Policy and organization optimization | Unused projects, org policy improvements |
Cost Recommendations in Detail
The most commonly used cost recommendations:
| Recommendation | What It Detects | What It Suggests |
|---|---|---|
| Idle VMs | VMs with low CPU, network, and disk usage over 1-14 days | Stop or delete the VM |
| Idle resources | Unattached Persistent Disks, unused static IPs, unused custom images | Delete the resource |
| VM rightsizing | VMs using a fraction of their CPU/memory capacity | Resize to a smaller machine type |
| CUD suggestions | Stable usage patterns that would benefit from a commitment | Purchase a 1-year or 3-year CUD |
| Unattended projects | Projects with no activity or user logins | Archive or delete the project |
In practice: Idle VM and idle resource recommendations are the quickest wins. Check Recommendation Hub monthly and clean up what you are not using.
How to Access Recommendations
| Method | How |
|---|---|
| Recommendation Hub | Google Cloud Console > Active Assist > Recommendations. Centralized view across all categories and projects. |
| In-context | Recommendations appear directly in service UIs (e.g., the Compute Engine VM list shows rightsizing suggestions inline). |
| gcloud CLI | gcloud recommender recommendations list --recommender=google.compute.machineType.Recommender --project=MY_PROJECT |
| Recommender API | REST API at recommender.googleapis.com for programmatic access. List, get, mark claimed, or dismiss recommendations. |
Each recommendation includes:
- Description of the issue
- Primary impact (estimated cost savings, risk reduction, etc.)
- Specific actions to take
- Supporting insights that led to the recommendation
Caution: Do Not Apply Blindly
Recommendations are suggestions, not commands. Review them in context before acting.
Rightsizing and CUDs: VM rightsizing recommendations do not account for existing flexible CUDs. If you downsize a VM to a smaller machine type, it may no longer match your commitment, and the smaller VM + the commitment cost could exceed the original on-demand cost of the larger VM.
Idle VMs with low CPU: A VM with low CPU usage is not necessarily idle. It might be performing important low-CPU work: monitoring, log aggregation, configuration management, or serving as a standby. Check what the VM is actually doing before deleting it.
IAM recommendations: Removing IAM roles can break service accounts and automation. Test changes in a non-production environment first, or use the recommended role for a transition period before removing the old one.
General approach:
- Review each recommendation and understand the reasoning
- Verify it makes sense for your specific workload
- Test changes in a non-production environment when possible
- Apply incrementally, not all at once
- Monitor for unintended consequences after applying
TL;DR
- Recommender (Active Assist) automatically generates optimization suggestions based on your usage patterns.
- Cost recommendations find idle VMs, unused resources, rightsizing opportunities, and CUD purchase suggestions.
- Access via Recommendation Hub in Console, gcloud CLI, or the Recommender API.
- Review before applying: Rightsizing does not account for CUDs, idle VMs might do important low-CPU work, and IAM changes can break automation.
- Check recommendations monthly for quick wins, especially idle resource cleanup.
- Recommendations cover cost, security, performance, reliability, sustainability, and management.
Resources
Active Assist Overview Official documentation for the Recommender and Active Assist portfolio.
Idle VM Recommendations How idle VM detection works and what triggers recommendations.
View and Apply Idle Resource Recommendations Guide for reviewing idle disks, IPs, and images.
CUD Recommendations Blog Post How Recommender suggests CUD purchases based on your usage patterns.
Committed-Use Discounts When Recommender suggests a CUD, understand what you are committing to.
Cost Optimization Overview of all cost levers on Google Cloud.