Negative caching
What is Negative caching?
Negative caching lets a resolver remember that a DNS query returned no answer -- NXDOMAIN or an empty NOERROR -- for a TTL set by the zone's SOA record, per RFC 2308. AdminKit's TTL guide explains how that same SOA minimum controls DNS propagation delay.
By AdminKit team, Netbase JSC · Updated
Negative caching is a resolver’s practice of remembering that a DNS query returned no answer — either NXDOMAIN (the name doesn’t exist) or NOERROR with an empty answer (the name exists but not for the requested record type) — so it can return that same negative result without re-querying authoritative servers. RFC 2308 (a Proposed Standard per its IETF Datatracker entry) defines how long a negative answer may be cached: the TTL comes from the minimum value in the zone’s SOA record, defined by RFC 1035, not the TTL of the (nonexistent) record itself — Cloudflare’s own DNS documentation labels this same SOA field “Minimum TTL: the TTL for caching negative responses,” citing RFC 2308 directly. This matters in practice because a misconfigured or not-yet-propagated DNS record can keep failing for resolvers that cached the earlier NXDOMAIN, even after the correct record is published, until that negative-cache TTL expires. Resolver operators can also cap that lifetime themselves: the Unbound resolver’s configuration reference provides a cache-max-negative-ttl setting that limits how long NXDOMAIN and NODATA answers stay cached, defaulting to 3600 seconds. See the NXDOMAIN entry for what that specific response code means and why some networks intercept it.
AdminKit groups negative caching under its DNS cluster — browse the AdminKit glossary index for other related terms.
Frequently asked questions
What two kinds of DNS answers does negative caching apply to?
NXDOMAIN (the name doesn't exist) and NOERROR with an empty answer (the name exists but not for the requested record type) -- both are 'no answer' results a resolver can cache and reuse.
Where does the TTL for a negative-cached answer come from?
The minimum value in the queried zone's SOA record, defined by RFC 1035 -- not the TTL of the record being queried, since that record doesn't exist (or doesn't exist for that type).
Why can a fixed DNS record still fail after it's corrected?
Because resolvers that already cached the earlier NXDOMAIN keep returning it until that negative-cache TTL expires, even though the correct record has already been published.
What RFC standardizes negative caching?
RFC 2308, which defines how long a negative answer may be cached and ties that duration to the zone's SOA minimum field rather than any TTL on the nonexistent record itself.