Glue records and lame delegation: how broken NS delegation breaks DNS
AdminKit's DNS lookup can surface two related delegation faults: glue records and lame delegation. A parent zone's NS records sometimes point to a name server inside the domain being delegated, which glue records (RFC 8499) solve by supplying that server's address directly. Lame delegation is the case where a listed name server isn't serving the zone at all.
By AdminKit team, Netbase JSC · Updated
Key takeaways
- A parent zone delegates a domain with NS records that name servers, not addresses. When a server’s name sits inside the zone it serves, the parent also has to publish its address. That address record is glue.
- RFC 9471 says a referral must carry all available glue for in-domain name servers, or set the TC flag so the resolver retries over TCP.
- A lame delegation is a listed name server that doesn’t answer authoritatively for the zone. Resolvers usually route around it, so it shows up as slow or intermittent failures rather than an outage.
- To check, compare what the parent hands out with what each listed server says, and look for the
aaflag in each server’s answer.
How does delegation work, and where does it get circular?
DNS resolves names by walking down a hierarchy: the root zone points a resolver to a TLD’s name servers, the TLD points it to a domain’s name servers, and the domain’s name servers answer for everything under it — IANA’s root zone database is the authoritative record of which name servers (and, where needed, which glue) each TLD delegates to at the very top of that chain. RFC 1034 describes this delegation as a chain of NS records, each pointing to the name of a server responsible for the next zone down — not its IP address. That distinction is the whole problem glue records exist to solve.
If example.com delegates to a name server called ns1.example.com, a resolver needs that server’s IP address to query it — but ns1.example.com is itself a name inside the very zone the resolver hasn’t been able to reach yet. Asking “what’s the IP of ns1.example.com?” would normally mean asking example.com’s name servers, which is circular: you’d need to already be able to reach the server to find out how to reach it.
RFC 1034 §4.2.1 puts it the same way: without glue, the NS records could tell a resolver that “in order to learn a name server’s address, we should contact the server using the address we wish to learn.”
What are glue records, and when do you need them?
RFC 8499 defines glue records as the address records — A or AAAA — published at the parent zone (the .com zone, in this example) alongside the delegating NS records, giving a resolver the IP address of an in-domain name server directly, without needing to resolve it through the zone it’s delegated to. This is why registrars ask for “nameserver IP addresses” specifically when a domain’s own name servers live inside that same domain: that information becomes glue, stored at the parent, and without it the delegation would be unresolvable for any server whose name falls inside the zone being delegated. Registrars, which ICANN accredits for generic top-level domains, usually pass this to the registry as a host object: the EPP host mapping in RFC 5732 says a name server’s IP addresses are required “only as needed to produce DNS glue records”. A name server hosted under a completely separate domain (a common setup with third-party DNS providers) has no such circular dependency and needs no glue at all.
In the parent zone, a delegation with glue looks like this (documentation addresses):
; in the com zone
example.com. 172800 IN NS ns1.example.com.
example.com. 172800 IN NS ns2.example.net.
ns1.example.com. 172800 IN A 192.0.2.53
ns1.example.com. 172800 IN AAAA 2001:db8::53
Only ns1.example.com gets glue here. ns2.example.net lives in another zone, so a resolver can look up its address the normal way.
RFC 9499, which replaced RFC 8499 in 2024, sorts the name in an NS record into three relationships with the delegated zone. The older “in-bailiwick” and “out-of-bailiwick” labels still appear in tools and FAQs, but RFC 9499 calls them historic.
| Relationship | Example delegation | Name server | Glue needed? |
|---|---|---|---|
| In-domain | child.example.com | ns.child.example.com | Yes. There is no other way to find the address. |
| Sibling domain | child.example.com | ns.another.example.com (also under example.com) | Usually not, unless two zones name each other’s servers (a cycle). |
| Unrelated | child.example.com | ns.example.net | No. The resolver looks the name up separately. |
RFC 1912 §2.3 adds one practical rule: if a name server has several addresses, list all of them in the glue, or caches can end up with only some of them.
What does RFC 9471 require in a referral?
A referral is the answer a parent server gives when it doesn’t hold the data itself: “ask these servers instead”. RFC 9471 updates RFC 1034 to make the glue rules explicit. A server building a referral MUST include all available glue for in-domain name servers in the additional section, or set the TC (truncated) flag if they don’t fit, so the resolver retries over TCP. Glue for sibling-domain servers SHOULD be included, but the server may leave it out when space runs short.
The RFC exists because the failure is real: it includes an observed case where a referral carried no glue and no TC flag, and resolution failed. It also notes that it only governs server software. It doesn’t make anyone put glue into a zone; if the glue was never registered, there is nothing to return.
Here is the referral a parent server gives for www.example.com. Example output (illustrative, documentation addresses):
$ dig @192.0.2.1 www.example.com A +norecurse
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 4120
;; flags: qr; QUERY: 1, ANSWER: 0, AUTHORITY: 2, ADDITIONAL: 3
;; AUTHORITY SECTION:
example.com. 172800 IN NS ns1.example.com.
example.com. 172800 IN NS ns2.example.net.
;; ADDITIONAL SECTION:
ns1.example.com. 172800 IN A 192.0.2.53
ns1.example.com. 172800 IN AAAA 2001:db8::53
Read it like this: no aa flag and an empty answer section mean “not my zone”; the authority section lists the delegated servers; the additional section is the glue. If ns1.example.com appeared in the authority section with nothing in the additional section, that delegation would have missing glue.
What is a lame delegation?
A separate, and more common, DNS delegation failure is a lame delegation — RFC 8499 defines it as a name server listed in a zone’s NS records that is not actually configured to serve that zone. This happens in ordinary, unglamorous ways: a name server gets decommissioned or repurposed but the NS record pointing to it is never removed; a domain migrates DNS providers and an old provider’s server is left in the NS set; or a secondary server was simply never set up to transfer and serve the zone it was listed for. Whatever the cause, a resolver that happens to query a lamely-delegated server gets a timeout, a refusal, or an answer for an entirely different zone — the delegation points somewhere that can’t actually help.
The definition RFC 8499 quotes comes from RFC 1912 §2.8, which states the test simply: queries to a listed server about your names “should always result in an ‘authoritative’ response. If not, this is called a ‘lame delegation’.” RFC 9499 notes that the term has drifted. People now also use it for a listed server that doesn’t answer at all, one whose address isn’t reachable, and one that answers with an error or without the authoritative bit. Because of that, RFC 9499 recommends more specific wording. When you report a problem to a provider, say which of those you see.
Why does it matter even when most queries still work?
Lame delegation rarely breaks a domain outright, because resolvers typically try multiple listed name servers and move on after one fails to answer — which is exactly why lame delegations often go unnoticed for a long time. Resolver logs are one place they surface: the BIND 9 configuration reference has a dedicated lame-servers logging category for misconfigurations in remote servers that BIND discovers while querying them during resolution. The cost is intermittent: resolution becomes slower (extra round-trips to time out against the lame server before succeeding elsewhere) and less reliable, with the risk rising sharply if the working servers also have an outage, since a domain with lame delegations effectively has fewer functioning name servers than its NS record set implies.
RFC 1912 sums up the range: “At best, extra DNS traffic will result from a lame delegation. At worst, you can get unresolved hosts and bounced e-mail.” It also asks for at least two name servers per domain. A domain listing two servers, one of them lame, is running on one.
How do you check a delegation with dig?
Check both sides of the zone cut. RFC 1034 §4.2.2 asks the administrators of the parent and child zones to keep the NS and glue records on both sides “consistent”, and the checks below test exactly that.
1. Follow the delegation from the root. dig +trace makes dig resolve the name itself, starting at the root, and prints each referral along the way. Example output (illustrative, trimmed, documentation addresses):
$ dig +trace example.com SOA
...
example.com. 172800 IN NS ns1.example.com.
example.com. 172800 IN NS ns2.example.net.
;; Received 120 bytes from 192.0.2.1#53(parent-ns.example.net) in 24 ms
example.com. 3600 IN SOA ns1.example.com. hostmaster.example.com. 2026111001 7200 900 1209600 300
;; Received 98 bytes from 192.0.2.53#53(ns1.example.com) in 31 ms
The block that comes from a parent server is the delegation as the world sees it. The last block comes from whichever listed server dig picked. A lame server may not show up in the trace at all if dig happens to pick a working one, so the next steps query each server in turn.
2. Ask a parent server for the NS set. Use one of the TLD servers from the trace, with +norecurse, as in the referral example above. Note every name in the authority section and every glue address.
3. Ask each listed server for the zone’s SOA. Example output (illustrative, documentation addresses):
$ dig @192.0.2.53 example.com SOA +norecurse
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 9071
;; flags: qr aa; QUERY: 1, ANSWER: 1, AUTHORITY: 0, ADDITIONAL: 1
$ dig @198.51.100.53 example.com SOA +norecurse
;; ->>HEADER<<- opcode: QUERY, status: REFUSED, id: 33817
;; flags: qr; QUERY: 1, ANSWER: 0, AUTHORITY: 0, ADDITIONAL: 1
The first server is healthy: aa means it is an authority for the name, as RFC 1035 §4.1.1 defines the bit. The second answers but refuses the zone, which is a lame delegation. A timeout, or a NOERROR answer without aa, would also count. Compare the SOA serial across healthy servers too; see SOA records explained for what that number tells you.
4. Ask each server for its own NS set (dig @server example.com NS +norecurse) and compare it with the parent’s list from step 2. They should match.
| Symptom | Likely cause | Fix |
|---|---|---|
One listed server times out or returns REFUSED | Old provider or decommissioned server still in the NS set | Remove it at the registrar and in the zone’s own NS records |
Server answers without the aa flag | It isn’t loading the zone (secondary never set up, or a resolver at that address) | Configure it as a primary or secondary for the zone, or remove it |
| Parent lists an in-domain server with no glue | Host object has no address at the registry | Add the IP addresses for that name server at the registrar |
| Glue address differs from the server’s own A record | Server moved; glue was never updated | Update the host object’s addresses at the registrar |
| Parent and child NS sets differ | Change made on one side only | Make both lists the same |
How do you add or update glue at your registrar?
Glue lives at the registry, so you change it through your registrar, not your DNS provider. Registrars label the feature in different ways, such as “child name servers”, “host records” or “register a name server”, but underneath it is usually an EPP host object.
- Create the host first. Register
ns1.example.comwith its IPv4 and IPv6 addresses. RFC 5732 requires the parent domain (example.com) to be known to the registry before the host object can be created. - Then point the domain at it. Set the domain’s name servers to include
ns1.example.com. Run the referral check above to confirm the glue appears. - Changing an address: bring up the server on the new address, update the host object’s addresses, and keep the old address serving the zone while cached copies expire. RFC 1912 §2.8 recommends keeping the old server answering during that window to avoid a lame delegation.
- Retiring a name server: remove it from the domain’s NS set first. RFC 5732 says a host object SHOULD NOT be deleted while a domain still uses it.
Pull your domain’s current NS record set with AdminKit’s DNS lookup tool, then query each listed server directly for your zone and confirm it answers authoritatively — a server that times out, refuses, or answers for the wrong zone is lame. AdminKit’s Whois/RDAP lookup tool is a useful cross-check if the NS records themselves look unexpected, since it shows what your registrar has on file, which should match. The fix is simply removing or correcting the stale NS entry at your registrar or DNS provider so every listed server is one that’s actually configured and reachable for your zone.
Related checks
Delegation correctness and zone-transfer exposure are two distinct ways a domain’s DNS can be misconfigured without an obvious symptom day to day — see our guide on what is a DNS zone transfer (AXFR) for the transfer-exposure side, and DNS propagation and TTL explained for how caching behavior can make a delegation fix take time to be visible everywhere even after it’s corrected. The same glue and lameness questions apply one level down when a subdomain is delegated to its own name servers — see interpreting a subdomain delegation NS/glue mismatch for that decision tree.
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 does 'delegation' mean in DNS?
Delegation is how DNS distributes authority down the domain hierarchy: the root zone delegates each top-level domain (like .com) to that TLD's name servers via NS records, and each TLD in turn delegates individual domains to the name servers their owner specifies. RFC 1034 describes this as the mechanism that lets a resolver walk from the root down to the authoritative servers for any domain without any single server needing to know every domain in existence.
Why do glue records exist at all?
They solve a circular lookup problem specific to in-domain name servers. If example.com delegates to ns1.example.com, a resolver trying to find ns1.example.com's IP address would normally ask example.com's name servers — but that's exactly the information it's trying to obtain in the first place. RFC 8499 defines glue records as the address (A/AAAA) records published at the *parent* zone alongside the NS records, giving the resolver the IP address it needs to actually reach ns1.example.com without needing to resolve it first. Without glue, this specific delegation pattern would be unresolvable.
Do I always need glue records for my domain?
Only if at least one of your domain's name servers has a hostname inside your own domain (an 'in-domain' or 'in-bailiwick' name server, per RFC 8499's terminology) or a sibling domain under the same parent. If all of your name servers are hosted under a completely separate domain — a typical setup when using a third-party DNS provider — there's no circular dependency to solve, and glue records aren't required for that delegation to resolve.
What is a lame delegation, exactly?
RFC 8499 defines a lame delegation as existing when a name server is listed in a zone's NS records but is not actually, correctly serving that zone — often because the server was decommissioned, renamed, or simply never configured as a primary or secondary for it. Resolvers that query a lamely-delegated server get no authoritative answer, a timeout, or a response for the wrong zone, depending on what (if anything) is actually running at that address.
How do I check my own domain for lame delegation?
Query each name server listed in your domain's NS records directly and confirm it returns an authoritative answer for your zone — a correctly configured server responds with the authority (`aa`) flag set and your actual records; a lamely-delegated one typically times out, refuses, or returns a non-authoritative or unrelated answer. AdminKit's DNS lookup tool shows you the current NS record set as the starting point for testing each server individually.
Sources
- RFC 8499 — DNS Terminology (glue records, lame delegation)
- RFC 1034 — Domain Names: Concepts and Facilities
- RFC 1035 — Domain Names: Implementation and Specification
- IANA — Root Zone Database
- ICANN — accredits the registrars responsible for a domain's delegation and glue records
- RFC 9471 — DNS Glue Requirements in Referral Responses
- RFC 9499 — DNS Terminology (obsoletes RFC 8499)
- RFC 1912 — Common DNS Operational and Configuration Errors
- RFC 5732 — Extensible Provisioning Protocol (EPP) Host Mapping
- BIND 9 Administrator Reference Manual -- Configuration Reference
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.