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.
When a DNS change still shows the old value after the record's TTL has passed, one of the authoritative nameservers for the zone may still be serving the old copy. RFC 1034 says the TTL "describes how long a RR can be cached before it should be discarded", and then adds that this limit "does not apply to authoritative data in zones; it is also timed out, but by the refreshing policies for the zone." A secondary nameserver is refreshed by those policies, not by the TTL of the record you changed.
The refresh values are in the zone's SOA record. RFC 1035 defines REFRESH as "A 32 bit time interval before the zone should be refreshed." RFC 1034 says that whenever a new zone is loaded in a secondary, "the secondary waits REFRESH seconds before checking with the primary for a new serial." The check is on the SERIAL field, which RFC 1034 says "is always advanced whenever any change is made to the zone." A secondary whose serial equals the primary's serial concludes that "no changes have occurred".
What the symptom means
There are two separate waits. The first is the resolver's cache, which lasts for the TTL of the record when it was cached. The second is the secondary nameserver's copy of the zone, which lasts until the secondary next checks the primary. Waiting for the TTL does nothing for the second one.
A query to an authoritative server shows the zone as that server holds it. If the primary has the new value and a secondary has the old one, a lookup that reaches the secondary returns the old value, whatever the TTL says. The only way to tell is to ask every server in the delegation.
RFC 1035 and RFC 1034 give no single "propagation time". The SOA REFRESH, RETRY and EXPIRE fields set the secondary's polling, and RFC 1034 calls these "the minimum acceptable polling intervals".
A name error is cached separately. RFC 2308 sets that cache time from "the minimum of the MINIMUM field of the SOA record and the TTL of the SOA itself." Waiting for that number does not flush an old A or MX record.
Likely causes
| Cause | What you see | How to confirm |
|---|---|---|
| A secondary has not refreshed yet | The primary returns the new value and the secondary returns the old one, and its SOA serial is lower | Compare the SOA serial on every authoritative server |
| The serial was never advanced on the primary | Every server has the same serial, but the primary returns the new value and a secondary returns the old one | Equal serials with different data |
| The edit never reached the primary the delegation names | Every authoritative server returns the old value | dig +trace ends at servers that still hold the old record |
| The resolver still holds the old record | Every authoritative server returns the new value, and your resolver returns the old one with a TTL counting down | dig @ an authoritative server, then dig with no @ |
Prove the cause before changing anything
All of these are read-only. The manual's form is dig @server name type. +noall +answer prints "the name, record type, TTL, and value". +trace follows the delegation from the root. The default type is A.
1. List every authoritative nameserver for the zone.
dig +noall +answer example.com NS
Write down every name. The query uses the servers in /etc/resolv.conf unless you pass @server.
2. Query each server in that list for the SOA record, and compare.
dig +noall +answer @ns1.example.net example.com SOA
dig +noall +answer @ns2.example.net example.com SOA
The SOA data runs MNAME, RNAME, SERIAL, REFRESH, RETRY, EXPIRE, MINIMUM. Compare the third value, SERIAL, across all servers. RFC 1035 says serial numbers "should be compared using sequence space arithmetic", so a larger number is not always newer near the 32-bit boundary. A secondary with a lower serial than the primary has not refreshed yet. Read REFRESH on the same line, in seconds, as the interval it waits between checks.
3. Query the record you changed on each server.
dig +noall +answer @ns1.example.net www.example.com A
dig +noall +answer @ns2.example.net www.example.com A
Use the type you changed. Servers that return different values are the cause, and the server with the old value is the one to fix. If all of them return the new value, the old value is in a resolver's cache.
4. Read the record through your resolver.
dig +noall +answer www.example.com A
With no @server, a resolver that answers from its cache shows a TTL lower than the one you published, and a first query shows the full TTL. RFC 2308 section 5 says this for a cached negative answer, whose TTL "decrements in a similar manner to a normal cached answer". The wait for the old address in a resolver is the remaining TTL, and the SOA MINIMUM is not that wait.
5. See which servers the world asks.
dig +trace www.example.com A
If the trace ends at different servers from the ones you edited, you changed a zone those servers do not serve.
Fix
Which fix applies depends on step 2 and step 3.
If a secondary's serial is lower than the primary's and the zone was changed correctly, no change is needed. The secondary takes the new zone at its next successful check, normally within the REFRESH interval, and the old value then goes. RFC 1034 says "If this check cannot be completed, new checks are started every RETRY seconds." Ask again after those intervals.
If every server has the same serial but different data, the serial was not advanced, and the secondary will not fetch the change on its own. If every authoritative server returns the old value, the edit never reached the primary. Each of those is a change to the zone, covered below. If every authoritative server returns the new value, wait out the TTL that step 4 shows. No zone change helps a resolver's cache.
The serial was not advanced on the primary
First, save the SOA record as the primary serves it, and export the zone from the primary or your DNS host into a file named zone-before-serial.txt:
dig +noall +answer @ns1.example.net example.com SOA > soa-before.txt
Disruptive step: the next step raises the SOA serial of the live zone on the primary. Every secondary will fetch the whole zone at its next check, and anything in the zone that you have edited and not yet checked becomes live on all of them. The serial cannot be put back. A serial lowered again is treated as older than the copies already fetched, so the secondaries would not take it. Check the exported zone data in zone-before-serial.txt against your own change before you start.
Advance SERIAL in the zone on the primary, by the method your DNS host or server uses, so that it is larger than the value in soa-before.txt. Then repeat step 2.
To undo it: the old serial cannot be restored. If the zone data itself is wrong, put the data saved in zone-before-serial.txt back and advance the serial again, so the secondaries fetch the restored data.
Example (illustrative): www.example.com was 203.0.113.10 with a TTL of 3600. It is changed to 203.0.113.20 on ns1.example.net. Two hours later dig +noall +answer @ns1.example.net www.example.com A returns 203.0.113.20, and the same query to ns2.example.net returns 203.0.113.10. The SOA query shows serial 2026112101 on both. The serial was not advanced, so ns2.example.net never fetched the change. The administrator saves the SOA, advances the serial on ns1.example.net, and after the next refresh both servers return 203.0.113.20.
The edit never reached the primary
First, save the record as the primary serves it now:
dig +noall +answer @ns1.example.net www.example.com A > record-before.txt
Risk: the next step changes the live record on the primary, so every resolver will cache the new value for its TTL. The wrong address or target replaces the old one for that time.
Edit the record on the primary the trace ends at, and advance the SERIAL in the same change, then query the server again.
To undo it: set the record back to the value in record-before.txt and advance the serial again, so the secondaries fetch the restored data.
Confirm the fix
- The SOA serial is the same on every authoritative server.
- Each server returns the new value for the record.
- A resolver that still shows the old value shows a TTL no higher than the TTL that was in force when it cached the record. When that TTL reaches zero, a new query returns the new value.
What to watch afterwards
- Check the SOA serial on every authoritative server after each zone change, not only the record.
- RFC 1034 says "If a change can be anticipated, the TTL can be reduced prior to the change to minimize inconsistency during the change, and then increased back to its former value following the change." That lowers what resolvers keep. It does nothing for a secondary.
- A cached NXDOMAIN uses the negative TTL. A cached address uses the address record's TTL.
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
- 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
- Troubleshootingdig NXDOMAIN for a subdomain that should existdig returns NXDOMAIN for a subdomain you expected. Read the status without +short, ask the authoritative server, and check a cached name error before you add the record.Read article
Risks and limits
- RFC 1035 also describes an older use of MINIMUM as a lower bound on TTLs exported from a zone. RFC 2308 deprecates the "minimum TTL value of all RRs in a zone" meaning and assigns MINIMUM to negative answers.
- Resolvers you do not operate will not drop a cached record because you ask them to. The TTL is the mechanism RFC 1034 and RFC 1035 give you.
- These RFCs describe how a secondary polls. They do not describe how a given DNS host triggers a refresh; your DNS host's own documentation covers that.
- The dig manual does not define "propagation". The queries above are how you see authority and cache.
Sources
- RFC 1034: Domain Names - Concepts and Facilities — IETF, November 1987, accessed 22 November 2026
- RFC 1035: Domain Names - Implementation and Specification — IETF, November 1987, accessed 22 November 2026
- RFC 2308: Negative Caching of DNS Queries (DNS NCACHE) — IETF, March 1998, accessed 22 November 2026
- BIND 9 manual pages: dig — ISC, BIND 9.20.22 manual, no date shown, accessed 22 November 2026
This article was originally published on .