Non-prod redundancy
Resilience costs money: a second availability zone, a geo-redundant backup, a read replica, a higher SLA tier. In production that cost buys real protection. In dev, test, and staging — and for some non-critical prod workloads — you are often paying for redundancy the workload’s recovery objectives never call for.
What we look for
Section titled “What we look for”- HA where it isn’t needed — a multi-AZ database, zone-redundant service, or uptime-SLA tier on a resource tagged or named non-prod.
- Geo-redundant storage/backups (GRS) on workloads that don’t require cross-region recovery, where local redundancy (LRS) is sufficient.
- Extra replicas / multi-region writes carried in environments that don’t serve them.
These are judgement calls about recovery objectives, so each candidate is gated on a clear non-prod or low-criticality signal and framed for owner/compliance sign-off — we never quietly reduce the resilience of something that might need it.
How to act
Section titled “How to act”Some redundancy flags are reversible online; others (a backup storage-redundancy flag, once data exists) require a fresh resource and re-protect — the candidate calls out which. leancosts drafts the change, the compliance pre-check, and the rollback path. Follow Act on a cost finding.