Rightsizing & waste

Two control sets with dollar estimates: - rightsize (41 controls): resources that are oversized or not provisioned to your standards. - waste (24 controls): resources that are billed but unused.

What's checked

Rightsizing and standards: - EC2: allowed instance families, previous generations, low CPU, Graviton candidates, reservation candidates, expiring RIs, delete-on-termination. - EBS: gp2→gp3, io1→io2, provisioned-IOPS→gp3, oversized volumes, high IOPS, low usage. - RDS: allowed classes, low CPU, reservation candidates, Graviton, versions in paid Extended Support, and Multi-AZ / Performance Insights / backup retention by environment. - ECS: task count by environment (spread across AZs), task vCPU and memory limits, autoscaling, cluster CPU. - Lambda, CloudWatch Logs, DynamoDB, ElastiCache, EKS, EMR, Redshift, CloudFront, Route 53.

Waste: - Unattached EBS volumes, and volumes attached only to stopped instances. - Old and orphaned snapshots and AMIs. - Load balancers without targets, unattached Elastic IPs, NAT gateways without traffic. - RDS instances with zero connections, and orphaned RDS snapshots. - Unused secrets, idle EMR clusters, unused Route 53 health checks, redundant CloudTrail trails, disabled KMS keys. - S3 buckets without lifecycle rules (or without noncurrent-version expiry), ECR repositories without lifecycle rules.

The full catalog, with parameters, is on the controls page.

Usage

spendlens rightsize                                   # alarms first, largest saving first
spendlens rightsize --summary --by service
spendlens rightsize -S rds -S ec2 --min-saving 50
spendlens rightsize --set sizing.ec2_cpu_low=10       # one-off threshold change
spendlens waste --summary --by service
spendlens waste -C ebs_volume_unattached -f json -o unattached.json
spendlens rightsize controls                          # list controls
Option Meaning
-S, --service Limit to services (ec2, rds, ebs, …)
-C, --control Limit to specific controls
--status alarm, ok, info, skip, error or all
--summary, --by Group counts and savings by service, control, resource_type, account_id or region
--min-saving Only findings saving at least this much per month
--fail-on-alarm Exit 3 if any alarm

How savings are estimated

Each control with an estimator prices its alarms from pricing: in finops.yaml. For example: - gp2→gp3: size × (gp2 price − gp3 price). - Unattached volume: the volume's full monthly cost. - Low-CPU instance: downsize_saving_pct of the instance price.

The defaults are approximate us-east-1 list prices. Override them for your region or discounts:

spendlens waste --set 'pricing.ebs_gb_month={gp2: 0.116, gp3: 0.0928}'

Estimates are indicative. Use them to rank work, then review before acting. For AWS's own usage-based figures, see Savings.