Idle & orphaned resources
Some resources keep billing while doing no work: a virtual machine nobody logs into, a load balancer with no healthy targets, a disk left behind when its VM was deleted, a public IP that points at nothing. This is usually the fastest, lowest-risk saving available — you are removing waste, not trading off performance.
What we look for
Section titled “What we look for”leancosts flags a resource as idle or orphaned only when the evidence is unambiguous over a sustained window, never on a single quiet day:
- Idle — the activity metric that defines “in use” for that resource type (CPU, requests, connections, reads/writes, sessions) stays at or near zero across the measurement window.
- Orphaned — the resource has no parent or no consumers: an unattached managed disk, an unassociated public IP, a NAT gateway on an empty subnet, a load balancer whose backends are all gone.
Low-confidence candidates are withheld by default — a resource that merely looks quiet is not surfaced until the signal is strong enough to defend.
How to act
Section titled “How to act”Most idle and orphaned resources can be stopped (reversible) before they are deleted (often not). leancosts drafts a guided change with a pre-check to confirm the resource really is unused, a snapshot step for anything holding data, and a grace window before any destructive step. Follow Act on a cost finding.