Cloud Custodian
July 5, 2026 ยท View on GitHub
Use Cloud Custodian or an equivalent scheduled job to clean resources that can outlive failed runner jobs, image builds, or tests.
What To Manage
| Resource | Why |
|---|---|
| old runner AMIs | Prevent unbounded image growth. |
| AMI snapshots | Avoid storage cost after AMI deletion. |
| stale EC2 instances | Catch failed termination paths. |
| stale launch templates | Remove leftovers from runner experiments. |
| old ECR tags | Keep artifact repositories manageable. |
| untagged resources | Preserve ownership and cost attribution. |
Blueprint
Copy from:
docs/operations/repo-blueprints/cloud-custodian
docs/operations/workflows/scheduled-maintenance.md
Run cleanup jobs in audit mode first. Move to destructive mode only after the policy output is reviewed.