Managed Certificate Validation and DNS Requirements

Managed Certificate Validation and DNS Requirements

Managed certificate validation removes most of the manual work from TLS, but it only keeps working while the right DNS records stay in place. If your certificates come from AWS Certificate Manager, Google Cloud Certificate Manager, Azure Front Door or Cloudflare, the provider handles issuance and renewal. Your DNS configuration still decides whether that renewal succeeds. This article covers which records each validation method needs, what breaks them, and how to keep auto-renewal working instead of finding out about a failure from an expired certificate.

How Managed Certificate Validation Works in DNS

Every public certificate requires domain control validation (DCV). The certificate authority has to confirm that whoever requests a certificate for shop.example.com actually controls that name. Managed services automate this, but almost all of them rely on a DNS record that you publish once and leave in place.

Each platform does this a little differently. AWS ACM gives you a CNAME like _a1b2c3.shop.example.com pointing to a target under acm-validations.aws. Google Cloud Certificate Manager uses a DNS authorization that maps _acme-challenge.shop.example.com via CNAME to a host under authorize.certificatemanager.goog. Azure Front Door asks for a TXT record at _dnsauth.shop.example.com. Cloudflare’s SaaS hostnames support either TXT or HTTP validation.

The common thread is persistence. ACM and Google re-check the same record at every renewal. If the record is deleted, the next renewal fails without much warning.

The Set-and-Forget Myth

A common belief is that a managed certificate renews forever once it is issued. That is not true. Renewal depends on three conditions: the validation record still resolves publicly, CAA records still allow the provider’s CA, and (for ACM) the certificate is still attached to a service such as a load balancer or CloudFront distribution. If ACM can’t validate a certificate during the renewal window, which starts 60 days before expiry, it sends notifications. Those emails often go to an account root address nobody reads.

Shorter certificate lifetimes make this worse. Under CA/Browser Forum ballot SC-081, maximum public TLS lifetimes dropped to 200 days on March 15, 2026. They fall to 100 days in March 2027 and 47 days in March 2029, and the period for reusing domain validation shrinks to 10 days. That means renewals, and validation checks against your DNS, happen much more often. A missing record used to cause one failure a year. By 2029 it can cause roughly eight.

A Typical Failure Scenario

Here is a pattern most cloud teams will recognize. A platform team moves its zone from Route 53 to another DNS provider. They export the records, but the migration script skips entries starting with an underscore because they “look like leftovers.” All ACM validation CNAMEs disappear. Nothing breaks that day or that month.

About ten months later, ACM’s renewal attempts start failing. Notifications go to a distribution list that was retired during a reorganization. The certificate on the customer API expires on a Saturday at 02:14 UTC. Mobile clients that pin to the chain start failing right away. Restoring the validation record takes five minutes, and issuance takes about 20 more. The cause took a day and a half to find because nobody connected the DNS migration to a certificate outage almost a year later. The checklist in migrating DNS providers without downtime or data loss exists to prevent this.

DNS Requirements Practitioners Verify First

When managed renewal fails, an experienced SRE checks these things before opening a support ticket:

1. Public resolution of the validation record. Run dig _acme-challenge.shop.example.com CNAME @8.8.8.8 and compare the answer against an authoritative server. Split-horizon setups often have the record only in the internal view, and the CA never sees that view.

2. CAA records at every level. A CAA record on example.com that only allows letsencrypt.org will block ACM, which needs amazon.com or amazontrust.com. Google needs pki.goog. CAs walk up the tree, so a restrictive record on the parent applies to every subdomain that doesn’t have its own CAA.

3. DNSSEC health. An expired RRSIG or a mismatched DS record after a key rollover makes validating resolvers return SERVFAIL. CAs validate DNSSEC, so issuance fails even though the record exists.

4. Multi-perspective consistency. Since 2025, CAs have to confirm validation from several network vantage points. Geo-steered DNS that answers differently by region, or nameservers blocked for some networks, can fail this check.

5. Delegation and lame nameservers. If one of four NS records points to a decommissioned server, validation attempts that hit it can time out.

Common Mistakes With Validation Records

The most frequent mistake is cleaning up underscore records during hygiene work. Records like _a1b2c3 or _dnsauth look meaningless in a zone file, and someone deletes them to tidy up. Tag them in your IaC (in a Terraform resource description, for example) so the next person knows they are load-bearing.

The second mistake is the reverse: leaving delegated validation CNAMEs after a service is retired. Teams often point _acme-challenge at a third-party DNS-01 provider or a SaaS vendor’s validation zone. If that vendor relationship ends and the target zone becomes claimable, whoever claims it can get valid certificates for your hostname. This is a dangling record with a TLS certificate attached. It belongs on the same cleanup list as the issues described in dangling CNAME records.

The third mistake is trusting the provider console instead of checking DNS. ACM can show “Issued” for months while the validation record is already gone, because the status only changes at the next renewal attempt. Check the DNS record itself on a schedule.

Edge Cases Where the Standard Advice Changes

Air-gapped and on-prem environments using an internal CA such as Microsoft AD CS or HashiCorp Vault PKI don’t use public DCV. Most of this article doesn’t apply to them, though CAA and DNSSEC hygiene still matter for any public names. Regulated environments that require OV or EV certificates can’t use most managed services, because ACM and Google only issue DV certificates.

Wildcard certificates need DNS-based validation everywhere, since HTTP-01 can’t prove control of *.example.com. Teams under 10 people with a dozen hostnames can usually get by with a quarterly manual check. Beyond about 50 certificates, keep a certificate inventory mapped to active subdomains so every validation record has a known owner and a known reason to exist.

FAQ

Can I delete the ACM validation CNAME after the certificate is issued?
You can, but the next managed renewal will then fail. Keep the record for as long as you want the certificate to renew automatically. One CNAME per domain name also covers other ACM certificates for that name in the same account.

Does a CAA record block my managed certificate provider?
Only if it exists and leaves out the provider’s CA. With no CAA record anywhere in the tree, any CA can issue. Once you add one, you have to list every CA you use, including amazon.com for ACM and pki.goog for Google.

Why does validation fail when the record resolves fine from my laptop?
Your laptop probably uses an internal resolver or a cached answer. CAs query authoritative servers from several public locations. Test against public resolvers and each authoritative nameserver directly, and check DNSSEC with a validating resolver.

Keeping Renewals Working

Treat validation records as part of production, alongside your A and CNAME records. Keep them in version control, comment them, and check from outside your network that they still resolve, that CAA allows the right CAs, and that DNSSEC validates. As certificate lifetimes approach 47 days, a validation failure will show up within weeks rather than the following year, so put these checks in place before March 2027.