How to move DNS to a new provider without downtime
28 October 2026 · Alex
Moving DNS providers is one of those jobs that is either completely uneventful or a genuine outage, with very little in between. The difference is almost never technical skill. It is whether you verified the new zone before delegating to it, and whether you left the old one running afterwards.
Here is the sequence that makes it uneventful.
Before you touch anything
Export the current zone. Most providers offer a BIND zone file export. If yours does not, list the records and copy them by hand — and count them, so you can check the number later.
Watch for what an export misses. Zone files do not carry the things your old provider did for you: web forwarding or URL redirects, email forwarding, dynamic DNS hostnames, and any ALIAS or CNAME-flattening record, which is provider-specific and will not translate directly. Note each one; each needs a decision, not a copy.
Write down the current TTLs. You are about to need them.
Step one, and the one people skip: lower the TTLs
At least 24 hours before the migration — longer if any TTL exceeds a day — lower the TTL on every record in the old zone to 300 seconds.
This is the step that makes rollback possible. If something is wrong after you switch, a five-minute TTL means you are five minutes from having the old answers back. An 86400 TTL means you are a day from it, and there is nothing you can do in between.
You cannot skip this and make it up later. The TTL that governs today's caching is the one that was published yesterday.
Step two: build the zone at the new provider
Import the zone file, or add the records by hand. Then check three things that commonly go wrong:
- Record count. Compare against what you counted. A quiet truncation on import is common.
- NS records at the apex. Your new provider's nameservers should be the ones listed — not the old provider's, which is what a zone file import will have carried in. Some providers replace these for you; check rather than assume.
- SOA. The new provider writes its own. That is correct, and is not something to copy across.
Do not change your delegation yet.
Step three: verify against the new nameservers directly
This is the whole point of the exercise, and it can be done before anything is live. Query the new nameservers by name — they will answer for your zone even though the world is not yet asking them:
dig www.example.com @ns01.new-provider.com
dig MX example.com @ns01.new-provider.com
dig TXT example.com @ns01.new-provider.com
Compare each answer against the old provider. Walk the whole list: every A and AAAA, every MX with its priorities, every TXT — SPF and DKIM especially, because a broken DKIM record does not fail loudly, it just quietly sends your mail to spam.
Check every nameserver the new provider assigns you, not just the first. A zone that transferred to three of four is a zone that fails for a quarter of your visitors, intermittently, in a way that is very hard to diagnose later.
Step four: switch the delegation
At your registrar — not your DNS provider — change the nameservers for the domain.
Then understand what you are waiting for. The delegation lives in the parent zone at the registry, and those NS records carry a TTL you do not control, commonly a day or two. Some resolvers will use the new nameservers within minutes. Others will keep asking the old ones for a day. Both will be happening at once, which is exactly why the next step is not optional.
Step five: leave the old zone running
Do not delete the old zone. Do not cancel the old account. Leave it serving correct records for at least a week.
During the overlap, some resolvers still ask the old nameservers. If those answers are right, nobody notices anything. If the zone is gone, those resolvers get nothing, and a portion of your traffic fails — while every test you run yourself succeeds, because your resolver moved over on day one.
This is the single most common way a DNS migration causes an outage, and it happens after the part everyone was nervous about went fine.
A week is a reasonable minimum. If the old NS TTL was 48 hours, make it two.
Step six: tidy up
Once you are confident nothing is asking the old provider:
- Put the TTLs back up. 3600 is a sensible default; 86400 for records that genuinely never change.
- Re-enable DNSSEC if you had it — see below.
- Then, and only then, delete the old zone.
DNSSEC needs its own plan
If the domain is DNSSEC-signed, you cannot simply move it. The DS record at the registry points at a key the old provider holds, and the new provider has a different one. Switch the nameservers with the old DS still published and the domain does not just break — it becomes unresolvable for every validating resolver, which is most of them.
The safe route for most people is:
- Remove the DS record at your registrar.
- Wait for the old DS TTL to expire, so no validator still expects a signature.
- Migrate as above.
- Sign at the new provider and publish the new DS.
There is a more elegant path involving both providers publishing each other's keys, but it requires cooperation between them and is not worth attempting unless both document it.
The checklist
| When | Do |
|---|---|
| A day or more before | Lower all TTLs to 300 |
| Before switching | Build the zone, verify every record against every new nameserver |
| Before switching | Remove the DS record if DNSSEC is on |
| Switch | Change nameservers at the registrar |
| For a week after | Leave the old zone serving, unchanged |
| After a week | Raise TTLs, re-sign, delete the old zone |
Nothing here is difficult. All of it is boring, and all of it is skipped by people who have a bad migration.
KumoDNS imports BIND zone files and keeps the platform's own nameservers at the apex, so an imported zone is not left delegated to the provider you are leaving. The import and export guide covers the detail.
Need somewhere to put these records?
KumoDNS hosts one zone free — no card, no time limit.