CostLensMIG Autoscaling (GCP)
MIG Autoscaling (GCP)
CostLens flags Managed Instance Groups running a fixed size with no autoscaler, so you can scale down during low demand.
Updated July 20261 min read
CostLens flags Managed Instance Groups (MIGs) with no autoscaler. A fixed-size group pays for every instance 24×7 even when demand drops; attaching an autoscaler lets it shrink during quiet periods and grow when needed.
How it works
Groups and autoscalers are enumerated
CostLens lists MIGs and autoscalers across regions via the Compute aggregated API, and matches each group to its autoscaler (if any).
Fixed-size groups are flagged
Groups with 2+ instances and no autoscaler are surfaced.
Savings are estimated
Priced from the group's instance-template machine type × instance count × a conservative autoscale-savings fraction (~20%).
What to do
- Confirm the workload has variable demand (autoscaling won't help a steady 100%-utilized group).
- Attach an autoscaler (Compute Engine → Instance groups → Edit → Autoscaling) keyed on CPU, load-balancing utilization, or a custom metric.
Severity levels
| Severity | Estimated monthly savings |
|---|---|
| critical | >$500 |
| high | $100–$500 |
| medium | $20–$100 |
| low | <$20 |
Required permissions
- Detect: Compute read (
roles/compute.viewer) — see GCP Accounts.
Limitations
- The savings figure is a conservative estimate — actual savings depend on the demand curve.
- A group that's genuinely at steady high utilization won't benefit from autoscaling.