# Managed Instance Group

> A managed instance group is free as an object. Its cost is the Compute Engine VMs it maintains at the configured target size, 24 hours a day unless something changes that size. ZopNight discovers MIGs via Cloud Asset Inventory and manages capacity through GCP scaling schedules on the group, never stopping members individually.

Source: https://zop.dev/integrations/gcp/mig
Updated: 2026-08-19

---

A managed instance group runs a fleet of identical VMs from a template with autoscaling and autohealing. The group's size settings, not individual VMs, determine its compute spend.

## The group costs nothing; its target size costs everything

A MIG itself carries no charge: the bill is the sum of the Compute Engine instances it maintains, each metering vCPUs and memory per second plus a boot disk at provisioned size. What makes MIGs distinctive is that the spend is set by configuration, not by any single machine: the target size, the autoscaler's minimum, and the instance template's machine type together fix the fleet's burn rate. Change none of them and the group rebuilds any VM that disappears, restoring the cost along with the capacity.

## Why stopping a member VM is futile

Autohealing and autoscaling exist to defeat exactly the intervention a scheduler would make. Stop or delete a member and the group recreates it from the template within minutes. ZopDev therefore discovers MIGs via Cloud Asset Inventory, treats member VMs as group-managed, and works through the gcp:mig autoscaler provider using GCP scaling schedules on the group itself. Capacity is shaped by telling the group what size to be and when, never by touching members individually. This is also why MIG members are excluded from ordinary VM scheduling: two controllers issuing opposite instructions to the same machines helps no one.

## Group-shaped overspend

Look first at autoscaler minimums: a floor chosen during a traffic spike quietly becomes the permanent overnight size. Second, templates drift stale, leaving a group still stamping out the large machine type an old workload needed. Third, zombie groups: a service is retired, its load balancer deleted, and the MIG keeps three healthy instances serving nothing because nobody deleted the group.

## Reviewing groups in the console

Google Cloud console → Compute Engine → Instance groups shows each group with its current and target size and autoscaling mode. The Details tab exposes the autoscaler configuration and any scaling schedules, where a ZopNight-managed group's quiet-hours shape is visible.