# Route 53 Records

> zop.dev manages individual Route 53 record sets (name, type, value, and TTL) through the route53-record resource type, with zone CRUD alongside as route53-zone. Records are free; queries bill per million. TTL is the setting with memory: a change propagates only as caches expire, up to 86400 seconds on day-long TTLs.

Source: https://zop.dev/integrations/provision/aws/route-53-records
Updated: 2026-08-19

---

DNS records point names at your load balancers, databases, and endpoints. zop.dev manages individual Route 53 records with full lifecycle support, independent of zone provisioning.

## Record sets managed alongside zone CRUD

Route 53 record sets within a hosted zone (resource type route53-record), alongside zone CRUD (route53-zone).

## Record name, type, value, and TTL

```text
Record name, type, value, and TTL.
```

## Terraform handles records only inside modules

Available through the cloud-resource-provisioner API; the Terraform layer manages records only as part of larger modules.

## Records without a module around them

The Terraform layer touches DNS records only as parts of larger modules: a cluster's ingress
records, a zone's delegation set. The `route53-record` resource type covers everything outside
those shapes: pointing a marketing domain at a load balancer, a TXT record for domain
verification, an MX set for mail, a CNAME for a vendor integration. Full lifecycle support
means these are managed records, not console one-offs. They can be listed, updated, and
removed by the same automation that created them, which is how DNS avoids becoming a pile of
records nobody remembers authoring.

## TTL debt is paid at migration time

TTL is permission to cache. A record with an 86400-second TTL tells every resolver on the
internet it may serve the old value for a full day after you change it, so the moment of
editing a record is not the moment the world sees the edit. The standard playbook for a planned
move: lower the TTL well ahead of time, wait out the old TTL so caches drain, make the change,
then raise it again. Skipping the first step is how a "5-minute cutover" turns into a day of
split traffic between the old and new targets.

## Cheap to hold, instant to break

Records themselves cost nothing; Route 53 bills per million queries against the zone. The real
cost of record edits is risk, not money. A fat-fingered value propagates to resolvers
immediately and caches for the TTL you set, and DNS sits upstream of every other debugging
step. The zone-plus-record CRUD pairing is what makes careful automation possible: converge
records from configuration, and the blast radius of a typo shrinks to a bad commit that gets
reverted instead of a mystery outage.