Idle staging server needs attention.
Marshal connected one infrastructure change to its cost, exposure, utilization, and customer impact-then prepared one bounded decision. This page shows the evidence and guardrails a real brief should expose.
Decision
Review before Friday
No automatic change scheduled
What Marshal found
staging-worker-03 looks temporary, but it is still running and exposed to the internet.
The server appeared three days ago after a staging deploy. It averages 2% CPU, has no observed production dependency, and is projected to add $74 this month. No customer-facing outage was detected in the same window.
Added 3 days ago
Created by deploy-role in us-east-1
+$74 projected
Estimate from the current AWS rate and runtime
Public access on
Public IP and SSH open to the internet
None detected
No production dependency or monitor outage found
The proposed action
Stop staging-worker-03.
Stop the verified-idle EC2 instance. Its EBS disks survive and Marshal can start it again. This removes the runtime cost, but the open security group remains a separate finding until you restrict it.
- Risk
- Medium · service interruption
- Rollback
- Supported · start it again
- Cost effect
- About $74/month avoided
What happens after approval
Recheck
Confirm the resource and preconditions still match the reviewed plan.
Execute
Run the deterministic, allowlisted provider action-not generated shell code.
Verify
Read the provider state back and keep the issue open until it is confirmed.
See this loop with your evidence.
Start read-only. Marshal will show what each source has reported, what is missing, and what deserves the first decision. No credit card required.