Interpreting a subdomain delegation NS/glue mismatch
AdminKit's DNS lookup can surface a subdomain delegation mismatch: delegating a subdomain to its own name servers is a zone-file change, not a registrar change. When the parent zone's NS records don't match the subdomain's own servers, or glue is missing, resolution breaks in a way that's easy to misdiagnose as a registrar problem.
By AdminKit team, Netbase JSC · Updated
Key takeaways
- A subdomain is delegated by NS records inside the parent zone, not at the registrar. WHOIS/RDAP and the registrar’s control panel never show it.
- Glue is only required when a delegated name server’s name sits inside the subdomain it serves (an “in-domain” name server). That glue also goes in the parent zone.
- The parent’s NS set and the child’s own NS set can legitimately differ, but every server the parent lists must answer authoritatively for the subdomain.
- Diagnose in order: does the delegation exist, do the name servers resolve, do they answer authoritatively, do both NS sets agree.
Where does a subdomain delegation actually live?
This guide is part of AdminKit’s DNS cluster, covering delegation and glue alongside the other zone-mechanics guides linked throughout. A domain can be “delegated” in two genuinely different senses, and conflating them is the usual reason a subdomain delegation problem takes longer to find than it should. Delegating example.com itself happens at the registrar: the registrar publishes the domain’s NS records (and any glue they require) into the .com zone, and a Whois/RDAP lookup shows exactly what’s on file there, matching the delegation IANA’s root zone database publishes for the TLD itself. Delegating dev.example.com to a different set of name servers — a common pattern when a SaaS platform, a separate team, or a different DNS provider is meant to run one piece of a larger domain — is an entirely different act: it’s a set of NS records published inside example.com’s own zone file, by whoever controls that zone’s DNS, with no registrar step at all and nothing a WHOIS/RDAP lookup on the domain will ever show you. Google Cloud’s own DNS documentation walks through this identical workflow for its Cloud DNS product — creating a separate hosted zone for the subdomain and then publishing NS records for it in the parent zone — confirming it as a standard, vendor-supported pattern rather than an edge case. RFC 1034 describes both as ordinary zone cuts, but only one of them is visible outside the DNS itself.
Whole domain (example.com) | Subdomain (dev.example.com) | |
|---|---|---|
| Where the NS records go | The TLD zone (.com), via the registrar | The parent zone (example.com), via its DNS host |
| Who can change them | Registrar, on the registrant’s instruction | Whoever edits the parent zone |
| Where glue goes | Registrar’s “host records” / glue settings | Ordinary A/AAAA records in the parent zone |
| Visible in WHOIS/RDAP | Yes | No |
What do the parent and child zones look like?
A working delegation is two sets of records that point at each other. The example below delegates dev.example.com to one in-domain name server (which needs glue) and one name server under another domain (which doesn’t). Illustrative zone extracts, documentation addresses:
; parent zone: example.com (hosted by the main DNS provider)
dev.example.com. 3600 IN NS ns1.dev.example.com.
dev.example.com. 3600 IN NS ns.provider.example.net.
ns1.dev.example.com. 3600 IN A 192.0.2.53 ; glue: name is inside dev.example.com
; child zone: dev.example.com (hosted by the team or SaaS running it)
dev.example.com. 3600 IN SOA ns1.dev.example.com. hostmaster.example.com. 2026111001 7200 3600 1209600 3600
dev.example.com. 3600 IN NS ns1.dev.example.com.
dev.example.com. 3600 IN NS ns.provider.example.net.
ns1.dev.example.com. 3600 IN A 192.0.2.53
The NS records appear twice on purpose. The parent’s copy is the referral that sends resolvers to the right place. The child’s copy is the authoritative version, which resolvers use once they get there.
When is subdomain glue needed, and where does it live?
The same circular-lookup problem that makes a glue record (see glue records and lame delegation for the full background) necessary for a whole domain applies identically one level down. If dev.example.com is delegated to a name server called ns1.dev.example.com — a hostname inside the very subdomain being delegated — resolving that name server’s address would normally mean asking a zone that hasn’t been reached yet, exactly the circularity glue records exist to solve. RFC 9499, which replaced RFC 8499 as the DNS terminology reference, quotes RFC 1034’s definition: glue records are address records for name servers “below” the cut, needed so we aren’t told to contact a server “using the address we wish to learn”. The difference from apex-domain glue is entirely about where the fix goes: a whole-domain glue record is published by the registrar into the TLD zone, but a subdomain’s glue record has to be published directly in the parent zone’s own file — example.com’s zone, not the registrar’s nameserver settings — alongside the NS records that delegate the subdomain in the first place. A delegation set up by someone who only thought to check the registrar, because that’s where “glue records” usually live for a whole domain, is a common way this gets missed entirely.
RFC 9499 sorts every delegated name server into one of three relationships with the zone, and RFC 9471 sets the glue rules for each. Applied to dev.example.com, delegated from example.com:
| Name server | RFC 9499 type | Glue in the parent zone? |
|---|---|---|
ns1.dev.example.com | In-domain (inside the delegated subdomain) | Required. RFC 9471: referrals must include it, or set the truncation flag |
ns.ops.example.com | Sibling domain (inside the parent, outside the subdomain) | Usually not required, but servers commonly include it; required if two siblings depend on each other |
ns.provider.example.net | Unrelated (under another domain) | Not needed; resolvers look the name up separately |
The sibling case is where older advice (“glue for anything inside the parent domain”) and the current wording differ. A resolver can normally resolve ns.ops.example.com by following a separate referral to ops.example.com. The exception RFC 9471 calls out is a cycle: if dev uses a server in ops, and ops uses a server in dev, only glue in the parent breaks the loop.
How do you see the parent’s view and the child’s view?
AdminKit’s DNS lookup tool asks a public recursive resolver, so for NS it shows what the resolver ended up trusting, which is normally the child’s authoritative set. To compare that with the parent’s referral, ask a parent server directly with recursion off. Illustrative commands and trimmed output, documentation addresses:
# 1. Who serves the parent zone?
dig NS example.com +short
# 2. Ask a parent server for the referral (no recursion)
dig @ns1.example.com dev.example.com NS +norec
;; flags: qr; QUERY: 1, ANSWER: 0, AUTHORITY: 2, ADDITIONAL: 2
;; AUTHORITY SECTION:
dev.example.com. 3600 IN NS ns1.dev.example.com.
dev.example.com. 3600 IN NS ns.provider.example.net.
;; ADDITIONAL SECTION:
ns1.dev.example.com. 3600 IN A 192.0.2.53
# 3. Ask each delegated server for the child's own NS set
dig @192.0.2.53 dev.example.com NS +norec
;; flags: qr aa; QUERY: 1, ANSWER: 2, AUTHORITY: 0, ADDITIONAL: 1
A referral has records in the AUTHORITY section, no aa flag, and glue in ADDITIONAL. A correct answer from a child server has the aa (authoritative answer) flag and the NS set in the ANSWER section. Compare the two lists.
The decision tree
Work through these in order rather than guessing which layer is broken:
- Does the parent zone even show a delegation for the subdomain? Query the subdomain’s NS records with AdminKit’s DNS lookup tool. No NS records at all means the subdomain was never delegated in the parent zone — there’s nothing further down the chain to diagnose yet.
- Do the delegated name servers resolve to an address? If one of them has a hostname inside the subdomain and doesn’t resolve, that’s a missing glue record at the parent zone — the resolver can’t even reach the server to ask it anything, regardless of whether that server is correctly configured. If it’s a sibling name (elsewhere in the parent domain) that doesn’t resolve, check that name’s own A/AAAA record in the zone that holds it.
- If the name servers resolve, do they actually answer authoritatively for the subdomain? Query one of the delegated servers directly for the subdomain’s own records. A timeout, refusal, or non-authoritative answer here is a lame delegation, not a glue problem — see glue records and lame delegation explained for how to confirm and fix that specific failure.
- Does the subdomain’s own NS record set match what the parent delegates? A child zone listing a different (or larger) NS set than the parent isn’t automatically wrong — resolvers generally trust the child’s own answer once the referral succeeds — but if the parent’s list includes a server the child zone no longer recognizes as authoritative for itself, that’s a stale delegation left over from a provider migration, worth cleaning up even when it isn’t currently causing outright failures.
Why does this look like a registrar problem until it isn’t?
The most common dead end in troubleshooting a broken subdomain delegation is spending time in the registrar’s control panel, where the whole-domain NS records and any apex-level glue are both correct and entirely irrelevant to the problem. Because a subdomain delegation lives one level deeper, in the parent zone’s own DNS records, confirming it means querying DNS directly rather than checking what the registrar has on file — the registrar genuinely has no visibility into delegations made inside a zone it merely delegates the apex of. Amazon Route 53’s guide to routing traffic for subdomains shows the same split: to delegate a subdomain to its own hosted zone, you create an NS record named for the subdomain in the parent domain’s hosted zone, listing the subdomain zone’s name servers.
How do you move a subdomain to new name servers safely?
The stale delegations in step 4 of the decision tree are typically left over from a provider move done in the wrong order. Build the new side first and remove the old side last:
- Create the subdomain’s zone at the new provider and confirm each new server answers with the
aaflag, using step 3 of the commands above. - Change the NS records (and glue, if any) in the parent zone to the new servers.
- Keep the old servers answering until the old NS records’ TTL has expired, because resolvers keep using cached referrals until then; DNS propagation and TTL explains the timing.
- Update the NS set inside the child zone to match, then remove the zone from the old provider.
Next step
Run AdminKit’s DNS lookup tool against the subdomain in question, checking its NS record set first, then resolving each listed name server’s own A/AAAA record — that sequence is what the decision tree above walks through, and it’s the fastest way to see which step the delegation actually breaks at.
This guide is general information, not a security audit of your specific network — see our disclaimer for what AdminKit's guidance does and does not commit to.
Frequently asked questions
What's the difference between delegating a whole domain and delegating a subdomain?
A whole domain's delegation is set at the registrar, which publishes NS records (and any needed glue) into the TLD's zone — that's what a WHOIS/RDAP lookup shows you. A subdomain's delegation is set entirely inside the parent domain's own zone file, as NS records for the subdomain name published by whoever controls the parent zone's DNS, with no registrar involvement and nothing visible in WHOIS/RDAP at all. Troubleshooting the wrong one — checking the registrar when the problem is a zone-file delegation, or vice versa — is the most common reason this takes longer to fix than it should.
Do I need a glue record for a delegated subdomain?
It is required if at least one of the name servers you're delegating the subdomain to has a hostname inside the subdomain itself, the same in-bailiwick condition that requires glue at the registrar level for a whole domain. A name server elsewhere in the parent domain (a sibling, such as ns.ops.example.com) usually doesn't strictly need glue under RFC 9471, though DNS servers commonly include it, and it becomes necessary when two subdomains depend on each other's name servers. If `dev.example.com` is delegated to `ns1.dev.example.com`, resolving that name server's address would normally require asking the zone it's delegated to answer for — a circular dependency — so an A/AAAA glue record for it has to be published directly in the parent zone (`example.com`'s own zone file) alongside the NS delegation. If the subdomain is delegated to name servers hosted under a completely separate domain, no glue is needed for that delegation.
Where does subdomain glue actually get configured, if not at the registrar?
In whichever DNS provider's control panel or zone file hosts the parent domain itself — the same place that holds the NS records delegating the subdomain. This is a different location from registrar-level glue for the whole domain, which is why checking a registrar's nameserver glue settings for a subdomain problem finds nothing: that registrar has no visibility into, or control over, delegations made inside the parent zone.
The parent zone's NS records for my subdomain don't match what the subdomain's own servers list — is that always broken?
Not automatically. Resolvers follow a referral from the parent and then generally trust what the child zone's own servers answer with for everything other than the initial handoff, so a child listing a slightly different or additional NS set than the parent isn't inherently invalid. It becomes a real problem when the servers the parent actually hands out don't answer authoritatively for the subdomain at all (a lame delegation) or when the mismatch reflects a stale decommissioned server still listed somewhere in the chain.
How do I tell whether the problem is a missing glue record versus a lame delegation?
A missing glue record for an in-bailiwick name server typically shows up as the name server itself failing to resolve — the resolver can't get an address for it at all, so it never even gets to ask it anything. A lame delegation is different: the name server's address resolves fine, a connection succeeds, but the server doesn't answer authoritatively for the zone it was delegated. The first is a resolution problem one level before the second; see glue records and lame delegation explained for the general-case background both of these inherit from.
Sources
- RFC 1034 — Domain Names: Concepts and Facilities (zones, delegation, in-bailiwick name servers)
- RFC 1035 — Domain Names: Implementation and Specification (NS and glue record mechanics)
- RFC 8499 — DNS Terminology (glue record, lame delegation, in-bailiwick)
- IANA — Root Zone Database
- Google Cloud — Cloud DNS documentation (delegating a subdomain via NS records)
- RFC 9499 — DNS Terminology (obsoletes RFC 8499; in-domain, sibling and unrelated name servers)
- RFC 9471 — DNS Glue Requirements in Referral Responses
- Amazon Route 53 Developer Guide -- Routing traffic for subdomains
Want AdminKit to watch this for you?
Free checks tell you the state today. Protective DNS and monitoring tell you when it changes.
Talk to an expert
Huy Nguyen (David), Founder & CEO, Netbase JSC
Ask whether this fix applies to your setup, or get a second pair of eyes on your DNS and mail records.