Skip to content

Route 53

First PublishedByAtif Alam

Amazon Route 53 is a DNS web service: hosted zones, records, health checks, and traffic policies. It sits in front of many architectures — Elastic Load Balancing, CloudFront, API Gateway. For TCP/HTTP concepts, see TCP/IP Primer and HTTP for Operators.

TypeResolves for
Public hosted zoneInternet clients (authoritative for your domain on the public internet).
Private hosted zoneResources inside one or more VPCs (split-horizon: same name, different answers inside vs outside).

Associate a private zone with VPCs in the same or other accounts (with RAM / sharing where needed).

RecordUse
A / AAAAName → IPv4 / IPv6.
CNAMEName → another DNS name (cannot be used at zone apex for DNS standards — use alias on AWS).
Alias (Route 53)Map to AWS resources (ALB, NLB, CloudFront, S3 website, API Gateway). No charge for queries to AWS aliases. Works at apex (example.com).
NS / SOADelegation and zone metadata.
PolicyWhen to use
SimpleSingle target; no health-based routing.
WeightedSplit traffic by percentage (canary, migration).
LatencySend users to the lowest-latency region among healthy endpoints.
FailoverPrimary/secondary with Route 53 health checks.
GeolocationRoute by user’s geographic location (compliance, localization).
GeoproximityTraffic flow with bias toward regions (Traffic Flow).
MultivalueReturn multiple healthy records (not a substitute for full LB logic).

Combine with health checks so failover and weighted routing skip unhealthy targets.

Route 53 can probe HTTP/HTTPS/TCP endpoints (or calculated checks combining others). Failover records depend on check results. Align check path and expected status with a cheap endpoint (not a heavy DB query).

  • Inbound Resolver endpoints — On-premises DNS forwards queries into Route 53 for private zones.
  • Outbound Resolver — VPC workloads resolve on-prem names via conditional forwarding rules.
  • Resolver DNS Firewall — Optional filtering on DNS queries.

When documenting hybrid DNS, spell out who is authoritative for which suffix, and forwarding vs recursive paths.

Terminal window
dig example.com A
dig @8.8.8.8 example.com A # specific resolver
dig +trace example.com # delegation chain from roots
dig -x 203.0.113.10 # PTR (reverse)
SymptomLikely causes
NXDOMAINWrong zone, typo, delegation not updated at registrar, private record not associated with this VPC.
Wrong IP / stale answerTTL caching at resolvers or clients; recent change not propagated.
Works from laptop, not from EC2Split-horizon private zone, different resolver (/etc/resolv.conf / VPC DNS), hybrid forwarding.
Alias to ALB wrongDualstack vs IPv4-only; wrong region; health all targets down (see ELB page).
HTTPS works by IP but not nameDNS fixed but certificate SAN does not match hostname.

Lower TTL before migrations for faster cutover; higher TTL reduces query load and vendor API rate limits. Trade agility vs cache stability based on how often you need to change records.

Latency-based or failover records in front of regional ALBs are a standard active/active or active/passive story. Mention health checks and data replication assumptions (DNS does not fix app consistency).