Orphan EBS volumes: ₹80K/mo silent drain

Source: Dev.to

Problem Overview

Every AWS audit I run, I find the same thing:

• 15‑40 % of EBS volumes are orphaned – detached, still running, and still being billed.

The team’s reaction is always identical: “Oh, we’ll run a cleanup script this weekend.”
They do. Two months later, the orphans are back.

The cleanup script is a downstream fix; it doesn’t stop the upstream leak.

Root Causes

• Auto‑scaling groups terminate instances but can’t delete attached volumes.
• Terraform runs that recreate resources but leave old volumes dangling.
• Developer scripts that spin up one‑off EBS for testing, never deleted.
• CloudFormation stacks partially destroyed, leaving volumes behind.

Real‑World Impact

In an audit last month, a company was paying ₹80 K/month for 47 orphaned gp2 volumes.
Three of those were 2 TB snapshots tagged from an engineer who left in 2024.

Policy Recommendations

1. Tag every volume with an owner at creation.
2. IaC: enable ebs_auto_delete on instance termination.
3. Service Control Policy: block creation of untagged volumes entirely.
4. Weekly report: flag volumes unused for 7+ days; auto‑stop if the owner doesn’t respond within 14 days.

The scripts we write to clean up are proof that our policies are broken.

Takeaway

If this reminds you of a dashboard you’ve been putting off, share it. There’s likely a VPE or CTO in your network burning ₹5 L/year on this exact pattern.

Tags: AWS, FinOps, DevOps, CloudCost, InfrastructureAsCode, EBS, Kubernetes, IndiaSaaS, Founders