Combining policies and real designs
Nesting with aliases
You can't put two policies on one record set, but an alias record can point to another record set in the same zone — so you build trees:
Traffic Flow provides a visual editor and versioned traffic policies for complex trees (charged per policy record per month), applied to one or many names.
Design: global app with EU data residency
- Geolocation at the top:
EU→ an EU sub-tree; Default → a latency sub-tree. - EU sub-tree: failover between two EU regions (both compliant).
- Default sub-tree: latency records across ap-south-1, us-east-1, ap-southeast-1, each with health checks / evaluate-target-health.
- TTL 60, meaningful health endpoints, and a data layer that respects residency (EU data stays in EU regions).
Design: active-passive DR with a maintenance page
Failover: primary = production ALB in ap-south-1 (evaluate target health + a calculated or alarm-based check), secondary = DR region ALB (warm standby) — or a static S3/CloudFront maintenance page if there's no DR stack.
Design: migrate traffic to a new account/stack
Weighted: 100/0 → 90/10 → 50/50 → 0/100 over days, with health checks so a broken new stack automatically drops out; keep the old stack until traffic (including TTL-ignoring clients) drains.
Design: route a customer by office IPs
IP-based routing with the customer's CIDRs → a dedicated stack; default → shared stack.
Testing before go-live
Route 53's Test record feature (console, aws route53 test-dns-answer) shows what a resolver IP and/or client subnet would receive — verify geolocation, latency and IP-based trees without waiting for real traffic.