What a neglected cloud account looks like: $24,000 a month

Here's what a neglected cloud account actually looks like when you open it.

Naming and Tagging

No consistent naming convention. Resources called things like test-new-2, db-final-v3, loadbalancer-dont-delete. The names tell you nothing about purpose, environment, or owner.

Tags: either completely absent or present on 30 percent of resources with no consistent schema. Some resources tagged with env:prod. Others with environment:production. Others with nothing.

Regions and Orphaned Resources

Multiple regions active with no clear reason. The main workload runs in us-east-1. There are resources in eu-west-1, ap-southeast-2, and us-west-2. Some of them might be for users in those regions. Most of them are from experiments that ran in those regions because someone had a reason at the time and the resources never got cleaned up.

Orphaned resources everywhere:

• EBS volumes not attached to any instance.
• Elastic IPs not associated with any resource.
• Load balancers with no healthy targets.
• Security groups referenced by nothing.

Logging and Databases

Log groups without retention policies, ingesting data at rates that don't match the traffic the services should be generating.

RDS instances with automated backups retained for 35 days when the company's recovery requirement is 7 days. Multiple database snapshots taken manually for reasons nobody remembers. The oldest ones: from two years ago.

Pricing and Spend

Reserved instances that expired and weren't renewed, leaving workloads back on on-demand pricing without anyone noticing for four months.

The total monthly spend: $24,000. The team's honest estimate of what it should cost: $10,000 to $12,000.

The $12,000 to $14,000 gap is not from bad engineering decisions. It's from the accumulated weight of good engineers moving fast and never having the mandate, the time, or the tools to look back.

This is what the starting point looks like. The good news is that starting points are just starting points.