DEV Community

upordown.online
upordown.online

Posted on

Changing DNS providers? A 12-point checklist so your site and email survive the move

Uptime monitoring from outside your own infrastructure that alerts you the minute your site goes down — upordown.online

Changing DNS providers feels like a configuration change. In practice it's a migration with a long tail: once you update the nameservers at your registrar, resolvers around the world switch over gradually, over hours to a couple of days, and for that whole window some visitors see the old zone and some see the new one.

If both zones say the same thing, nobody notices. If the new zone is missing one record, some fraction of your users and mail servers start failing, and the failure comes and goes depending on which resolver they hit. That's a horrible thing to debug at night.

Here's the checklist I'd use.

Before you touch the nameservers

1. Export the complete current zone

Don't rebuild the zone from memory or from what the dashboard's main page shows. Export it. Most providers offer a BIND-format zone file export. If yours doesn't, list every record via their API. The records people forget are always the ones nobody looks at: _dmarc, DKIM selectors like google._domainkey, _acme-challenge, SRV records, and verification TXT records for services set up years ago.

2. Query the old zone directly, too

Exports can miss records (some providers hide "proxied" or provider-managed records). Ask the current authoritative nameservers directly for the names you care about:

NS=$(dig +short NS example.com | head -1)
for name in @ www mail _dmarc google._domainkey; do
  for type in A AAAA CNAME MX TXT; do
    dig +short @"$NS" "$name.example.com" "$type" | sed "s/^/$name $type: /"
  done
done
Enter fullscreen mode Exit fullscreen mode

Fix the loop for the apex (@ becomes the bare domain) and add every name you know about. Save the output: it's your before picture.

3. Watch out for records that only work on the old provider

Some providers support record types or features that aren't standard DNS: apex CNAME flattening, ALIAS or ANAME records, geo or latency routing, "proxied" records that point at the provider's own CDN. These don't translate one-to-one. Decide what each one becomes on the new provider before the move, not during it.

4. Lower the TTLs a day or two ahead

Lower TTLs on the records you might need to change quickly (A, AAAA, CNAME, MX) to 300 seconds, at least one full old-TTL period before the move. The NS records themselves are often cached for a day or more by resolvers and the parent TLD, and you can't speed that up much, which is exactly why the next step matters.

5. Build the new zone completely, then compare it

Create every record in the new provider. Then query the new nameservers directly (they're authoritative for your domain even before the switch, as long as you ask them by name) and diff the output against your before picture:

dig +short @new-ns1.provider.net www.example.com A
Enter fullscreen mode Exit fullscreen mode

The two zones should be identical except for the NS records and anything you've deliberately changed.

6. Handle DNSSEC before anything else

If DNSSEC is enabled, this is the step that causes multi-day outages. The parent zone holds a DS record pointing at your old provider's signing key. Switch nameservers to a provider that signs with a different key and every validating resolver will reject your domain as bogus.

The safe order is: disable DNSSEC at the registrar (remove the DS record), wait for the DS record's TTL to expire, move nameservers, confirm everything resolves, then enable DNSSEC on the new provider and add its DS record. Some providers support a multi-signer migration without disabling DNSSEC, but only do that if you've followed their documented procedure.

The scariest DNS cutover moment is “is the site up for everyone else?” upordown.online answers that from outside your network the moment the nameservers change.

The switch itself

7. Change nameservers at the registrar, and only there

The nameserver change happens at your registrar, not at either DNS provider. Double-check the spelling of every nameserver. One typo means some fraction of resolvers asks a server that doesn't know your domain.

8. Don't delete the old zone yet

Keep the old zone fully intact at the old provider for at least a week. Resolvers with cached NS records will keep asking the old nameservers during the transition. If the old zone is gone, those users get errors.

After the switch

9. Check from many resolvers, not just your own

Your laptop's resolver cached the old answer and will tell you nothing useful. Check through several public resolvers and from several regions:

for r in 1.1.1.1 8.8.8.8 9.9.9.9 208.67.222.222; do
  echo "$r: $(dig +short @$r NS example.com | tr '\n' ' ')"
done
Enter fullscreen mode Exit fullscreen mode

10. Test email explicitly

Send mail to and from the domain, and check that SPF, DKIM and DMARC pass in the received message headers (Authentication-Results). Email is where missed TXT records show up, and it fails quietly: messages land in spam rather than bouncing.

11. Check everything that verifies domain ownership

Certificate renewal (especially with DNS-01 challenges), search console verification, SaaS domain verifications, and custom domains on hosted services. Each one depends on a record you probably created once and forgot about.

12. Monitor the records for a week

Keep a check running that resolves your key records and compares them with what you expect, from several networks. Problems from a DNS move often appear days later, when a cached record finally expires. A DNS check alongside your HTTP checks, like the ones upordown.online runs, catches that before your users do.

Related reading

What record did you forget the last time you moved DNS? For me it has always been a DKIM selector.

Top comments (0)