Running dev and test Compute Engine VMs with no off-hours schedule

Office-hours VMs billed around the clock

A VM stopped overnight costs nothing for compute during those hours. Google’s instance lifecycle page lists CPU as billable only in RUNNING and PENDING_STOP. A development VM used from nine to six, Monday to Friday, is busy for roughly a quarter of the week and billed for all of it.

Compute Engine has a native way to fix this: instance schedules, a resource policy that starts and stops attached VMs on cron expressions in a time zone you choose.

Spotting unscheduled dev and test VMs

gcloud compute instances list --filter="status=RUNNING" \
  --format="table(name,zone,labels.list())"
gcloud compute resource-policies list

Look for names or labels that say dev, test, QA or staging, and check whether they appear in any instance schedule.

Evidence the VM is non-production

The VM must not be stopped and must look non-production: a name containing dev, test, qa, staging, sandbox or demo, or a dev or test environment label. A VM already carrying a schedule label is left alone. ZopNight then needs a measured off-hours pattern from the VM’s usage heatmap and a suggested stop and start schedule built from it.

Guard rails before any schedule is proposed

Production signals win: a production-looking name or environment label ends the check. VMs whose name carries a dr or standby token are excluded as disaster recovery capacity, since they must be ready at any hour. Self-managed databases, caches and message brokers are excluded too, because a quiet window does not prove durable state is safe to stop. A VM idle for 95% or more of the week is not a scheduling case at all; it goes to Idle GCP VM Instance.

Hours the schedule actually removes

saving = monthly VM cost x share of the week the suggested schedule keeps the VM stopped

ZopNight prices only what the proposed schedule does, not the VM’s total idle time.

Putting the VM on an instance schedule

1. Review the proposed start, stop and time zone with the team that uses the VM.

2. Create a schedule, following Google’s scheduling guide:

   gcloud compute resource-policies create instance-schedule dev-hours --region=REGION \
     --vm-start-schedule='45 7 * * 1-5' --vm-stop-schedule='0 19 * * 1-5' --timezone=Europe/London

3. Attach it: gcloud compute instances add-resource-policies VM_NAME --zone=ZONE --resource-policies=dev-hours.

4. Grant the Compute Engine Service Agent permission to start and stop instances, or the schedule fails.