About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.
If a DNSSEC-signed domain starts returning SERVFAIL on validating resolvers right after you moved to a new DNS provider, one cause is a Delegation Signer (DS) record at the registry that still points at the old provider's key. The new nameservers answer, but with different keys or no signatures, and every resolver that checks DNSSEC rejects the answer. Prove it with dig using +dnssec and then +cd. Fix it at the registrar by removing or replacing the DS record, then wait out its TTL.
What the symptom means
DNSSEC builds a chain of trust from the parent zone down. The parent holds a DS record for your zone. RFC 4034 says a DS record refers to a DNSKEY by "storing the key tag, algorithm number, and a digest of the DNSKEY RR." A validating resolver fetches the DS from the parent, fetches the DNSKEY from your nameservers, and checks that the two match and that the signatures verify.
RFC 4035 defines the outcomes. A response is "Secure" when the chain validates. It is "Insecure" when there is no chain, for example in an unsigned zone with no DS. It is "Bogus" when the resolver "believes that it ought to be able to establish a chain of trust but for which it is unable to do so." For a Bogus answer to a recursive query, RFC 4035 says the name server "MUST return RCODE 2", which is SERVFAIL, and "MUST return the full response if and only if the original query had the CD bit set."
That last rule is your test. A resolver with checking disabled (the CD bit) hands over the data instead of failing. If the same name works with +cd and fails without, the data is fine and DNSSEC validation is what breaks.
After a provider change, there are two ways, among others, to make the chain wrong:
- The DS at the registry points at the old provider's key. The new provider signs with its own key, or does not sign at all. RFC 6781 calls the state where the parent has a DS pointing to a nonexistent DNSKEY "security lameness", and says that if all DS records are in that state, "the child's zone" is "marked Bogus by verifying DNS clients."
- Caches. Some resolvers can fail while others recover. RFC 6840 notes that RFC 2308 "permits caching of server failures for up to five minutes."
Likely causes
| Cause | What you see | How to confirm |
|---|---|---|
| DS at registry is for the old provider's key; new provider signs with another key | SERVFAIL without +cd, answers with +cd; the DS key tag does not match any DNSKEY now served |
Compare dig DS at the parent with the DNSKEY from the new nameservers |
| New provider does not sign the zone, DS still published | Same SERVFAIL; dig DNSKEY at the new nameservers returns no key |
dig DNSKEY example.com @ns1.newprovider.example shows no answer |
| New DS entered wrongly (wrong key tag, algorithm or digest) | SERVFAIL after "fixing" the DS | Generate the DS from the new DNSKEY and compare |
| Nameservers disagree (old and new both still answering) | Intermittent SERVFAIL | Ask each delegated nameserver for DNSKEY and compare |
| Resolver caches | Some networks recover, others still fail | Query several resolvers; wait for the DS TTL |
| Expired signatures from an old copy of the zone | SERVFAIL with a signature expiry in the past | Check the RRSIG expiry fields in dig +dnssec output |
Prove the cause before changing anything
All of these only query DNS.
dig @192.0.2.53 example.com A +dnssec
dig @192.0.2.53 example.com A +cd
+dnssec sets the DNSSEC OK bit, "requests that DNSSEC records be sent". +cd "sets [or does not set] the CD (checking disabled) bit in the query. This requests the server to not perform DNSSEC validation of responses." Name a validating resolver with @ (here the documentation address 192.0.2.53; use your own), because without it dig asks whatever is in /etc/resolv.conf, which may not validate. Run both against the same validating resolver. The response status is in the header dig prints. SERVFAIL without +cd and a normal answer with +cd proves a validation failure. If both fail, the problem is not DNSSEC; read the page titled "DNS Changed but Users Still Reach the Old Application" and check the nameservers.
Next, see what the parent says. dig +trace follows referrals from the root and, per the manual, sets +dnssec too:
dig +trace example.com DS
Note the DS key tag, algorithm, digest type and TTL.
Then see what the nameservers publish. Find the delegated nameservers (dig NS example.com), and ask each one directly:
dig example.com DNSKEY @ns1.newprovider.example
dig example.com DNSKEY @ns2.newprovider.example
To turn the DNSKEY into the DS it should match, the BIND dnssec-dsfromkey manual gives this pattern, where -f - reads the zone data from standard input:
dig example.com DNSKEY @ns1.newprovider.example | dnssec-dsfromkey -f - example.com
By default it converts only keys with the KSK flag (257) and uses SHA-256 unless you choose another digest. Compare the key tag, algorithm and digest with the DS the parent holds. If they differ, the DS is stale. If the DNSKEY query returns nothing, the new provider is not signing the zone.
delv, BIND's validating lookup tool, simulates a validating resolver and reports why validation failed. Its manual notes that -i does not set the CD bit upstream, and that "when it is necessary to examine invalid data to debug a DNSSEC problem, use dig +cd."
Example (illustrative): example.com was signed at the old provider with a DS at the registry, TTL 86,400 seconds (24 hours; 86,400 ÷ 3,600 = 24). On Monday the administrator changes the nameservers to ns1.newprovider.example and ns2.newprovider.example and does not touch the DS. The new provider has signing off. dig example.com A +dnssec against a validating resolver returns SERVFAIL; with +cd it returns the address. dig +trace example.com DS shows the parent still has a DS, and dig example.com DNSKEY @ns1.newprovider.example returns nothing. The DS has nothing to match, so the zone is Bogus.
Fix
Decide which end state you want: two are valid.
Option A: the zone is unsigned at the new provider (for now)
Remove the DS record at the registrar. RFC 6781 describes the same move as the only workable path when the old operator does not cooperate: the zone goes Insecure for the duration of the change, and the registry is asked "to remove the DS RR pointing to the losing operator's DNSKEY and to change the NS RRset to point to the gaining operator. Once this has propagated through the DNS, the registry should be asked to insert the DS record pointing to the (newly signed) zone at operator B."
Disruptive step: removing the DS turns DNSSEC protection off for the domain. Answers are no longer validated, which is exposure against spoofed answers for as long as it lasts. To back out, add the DS again (keep a copy of the old values). Resolvers that cached the old DS keep failing until it expires. Wait at least the DS TTL you noted, counted from when the parent stops publishing the old DS; in the example that is 24 hours.
Option B: the new provider signs the zone
Get the DS values, or the DNSKEY, from the new provider's control panel, or generate the DS with dnssec-dsfromkey as above. Replace the old DS at the registrar. If the registrar requires the DNSKEY instead of the DS, give it the KSK. Check the algorithm number and digest type entered match the generated values.
Disruptive step: replacing the DS changes the trust anchor for the zone. A wrong value makes the zone Bogus again. Back out by restoring the previous DS, which only helps if the old nameservers still serve that key. Wait for the old DS to leave caches before judging the result.
The safe order for future moves
RFC 6781 section 4.3.5.2 describes the case where the old provider will not hand over keys: the zone goes Insecure, the DS is removed and the NS RRset changed, and the DS for the new zone is inserted after "this has propagated through the DNS". The order that follows is that section's, with the build-and-test step and the waits for caches added; the RFC does not specify those.
- Note the DS TTL, the NS TTL and the TTL of the zone's DNSKEY and main records. Keep a copy of the current DS values.
- Risk: build the complete zone at the new provider, with every record from the old one and signing on if it will be signed, but do not change the nameservers yet. Impact: nothing changes for the public, because no one is delegated there yet. Back-out: delete the zone at the new provider. Test it directly: ask each new nameserver for every record the zone serves (
dig example.com A @ns1.newprovider.example, and the same for MX, TXT and the rest), compare with the old nameservers, and confirmdig DNSKEYreturns keys. Generate the new DS from that key withdnssec-dsfromkeyand keep it for step 5. - Disruptive step: remove the DS at the registrar. Wait until
dig +trace example.com DSshows the parent no longer publishes it (registries differ in how long that takes), then at least the DS TTL, and as a precaution the longest of the other TTLs you noted; the RFCs give no wait times. Impact: DNSSEC protection is off for the domain until step 5. Back-out: add the DS back with the values you noted, which only helps while the old nameservers still serve the matching key. - Disruptive step: change the nameservers. Keep the old provider serving the zone, unchanged, until the NS TTL has passed, counted from when
dig +trace example.com NSshows the new set, and check that every delegated nameserver gives the same answers. Impact: resolvers move to the new nameservers as the old NS records expire, and a wrong nameserver name or an incomplete zone there breaks the domain. Back-out: change the nameservers back to the previous ones, which the old provider must still serve. - Disruptive step: publish the new DS at the registrar. Impact: validating resolvers start to check the zone against it, and a wrong key tag, algorithm or digest makes the zone Bogus again. Back-out: remove the DS, which returns the zone to Insecure, and wait for its TTL. Then test with
+dnssecfrom a validating resolver.
If both providers cooperate, RFC 8901 describes multi-signer arrangements where each provider signs, and RFC 6781 section 4.3.5.1 describes the losing operator pre-publishing the new operator's keys, both avoiding the insecure gap. Option B replaces the DS in one step: unless the old and new nameservers serve each other's keys during the change, there is a period in which the DS matches only one provider's keys, and resolvers that reach the other provider get SERVFAIL. Plan an overlap. Some registries support automated DS updates from CDS and CDNSKEY records (RFC 7344); check yours before relying on it.
Confirm the fix
dig example.com A +dnssecagainst a validating resolver returns the address with no SERVFAIL.- For a signed zone, ask with
+adflag. The dig manual describes it as requesting that the server return whether "all of the answer and authority sections have been validated as secure", which dig shows as the AD flag in the response header. dig +trace example.com DSshows the DS you intended, or none for Option A.- Queries from two or three different validating resolvers agree. Allow for caches.
What to watch afterwards
- Signature expiry on the new provider's zone, so validation keeps passing.
- Registrar notices about DS changes, and who can make them. The page titled "Who controls your domain: the registrant, the registrar account and transfer protections" explains the access you need.
- Any further provider change: put the safe order into the runbook.
Read next
Related articles you may want next.
- TroubleshootingDNS Changed but Users Still Reach the Old ApplicationDistinguish authoritative DNS, resolver caches and existing connections so an application migration has a measurable cutover and return plan.Read article
- HostingWho controls your domain: the registrant, the registrar account and transfer protectionsWho legally holds your domain, who holds the registrar login, and how .au and gTLD transfers work, with checks to run before a dispute or hijack.Read article
- Backup & RecoveryAuthoritative DNS when the primary is gone: serial, expiry, and who is still allowed to answerHow SOA expiry, zone serial numbers and a secondary's saved copy decide whether DNS survives the loss of the primary. RFC 1034 and BIND 9.20.30.Read article
Risks and limits
- These commands have not been run or lab-tested. They are taken from the RFCs and BIND manual pages cited in Sources.
- The RFCs do not describe a particular registrar's screen. Procedures, field names and propagation times differ between registries and registrars.
- Removing the DS lowers protection while it is absent. Do it in a planned window and restore it promptly.
- Zones using algorithm or key rollovers at the same time as a provider change need more careful planning than shown here.
Sources
- RFC 4033: DNS Security Introduction and Requirements — IETF, March 2005, accessed 18 November 2026
- RFC 4034: Resource Records for the DNS Security Extensions — IETF, March 2005, accessed 18 November 2026
- RFC 4035: Protocol Modifications for the DNS Security Extensions — IETF, March 2005, accessed 18 November 2026
- RFC 6840: Clarifications and Implementation Notes for DNSSEC — IETF, February 2013, accessed 18 November 2026
- RFC 6781: DNSSEC Operational Practices, Version 2 — IETF, December 2012, accessed 18 November 2026
- RFC 7344: Automating DNSSEC Delegation Trust Maintenance — IETF, September 2014, accessed 18 November 2026
- RFC 8901: Multi-Signer DNSSEC Models — IETF, September 2020, accessed 18 November 2026
- BIND 9 manual pages: dig, delv, dnssec-dsfromkey — Internet Systems Consortium, no version or date shown, accessed 18 November 2026
This article was originally published on .