What to do when your DNS provider goes down

When your DNS host has an outage, your site disappears even though your servers are fine. How to confirm it, what you can do during it, and how to prepare.

· 6 min read · By the Spot Downtime team

Your servers are healthy, your deploy is fine, and your site is unreachable because the service that answers "where is example.com?" stopped answering. A DNS provider outage is one of the few failures where you cannot fix the cause yourself, so the work splits into confirming it quickly, communicating, and preparing so the next one hurts less.

Confirm it is DNS

Symptoms: browsers show DNS_PROBE_FINISHED_NXDOMAIN or "server not found", API clients fail with name resolution errors, and every monitor on the domain fails at the same instant with no HTTP status at all. Check the authoritative name servers directly:

bash
# Which name servers does the domain delegate to?
dig NS example.com +short

# Ask each of them directly, bypassing caches
dig @ns1.dnsprovider.com example.com A +time=3 +tries=1
dig @ns2.dnsprovider.com example.com A +time=3 +tries=1
  • Timeouts from all the authoritative servers mean the provider is down or unreachable.
  • SERVFAIL can mean the provider is broken, or a DNSSEC problem on your side. See the SERVFAIL guide.
  • Correct answers from the authoritative servers but failures for users points to a resolver or propagation problem, not a provider outage.

Then check the provider's status page, and compare with a public resolver:

bash
dig @1.1.1.1 example.com A
dig @8.8.8.8 example.com A
DNS lookupQuery your records from outside your network to see what resolvers return right now.

Why some users still reach you

Recursive resolvers cache answers for the record's TTL. If your A record has a 1 hour TTL, many users keep resolving the site for up to an hour after the provider fails, and each resolver's cache expires at a different moment. That is why an outage can look partial, and why the failure grows over the first hour. See DNS TTL explained.

It also means a very low TTL makes you more exposed to a provider outage, not less: caches empty faster, so the outage shows up sooner. Short TTLs are for planned changes, not a permanent setting.

What you can do during the outage

1. Tell people

Post on a status page hosted away from the failing DNS. If your status page uses a subdomain of the same domain with the same provider, it will be down too. Use a separate domain or a hosted address, and keep a social account ready. Wording that works is in incident status updates.

2. Give customers a direct route

Services that can be reached by IP address, or on an alternate domain at a different DNS provider, can be given to customers who need access now. Internal tools can use /etc/hosts overrides, which are crude but immediate:

bash
# /etc/hosts on a critical machine
203.0.113.10  api.example.com

3. Move the zone, if you can

You can recreate the zone at another provider, but the change only takes effect when the NS records at your registrar point to the new servers, and that depends on the registrar, not the failing DNS host. The delegation TTL at the registry is often 24 to 48 hours, so a full cutover is slow. Resolvers that cached the old delegation keep asking the dead servers. Treat migration during an outage as a last resort, and only if you have an export of the zone.

Prepare before it happens

  1. Keep an up-to-date zone export. Zone file or API dump, stored somewhere you can reach without the failing provider. Re-creating records from memory under pressure goes badly.
  2. Consider secondary DNS. Two independent providers serving the same zone, each listed in your NS records, mean one can fail while the other answers. This needs a provider pair that supports zone transfers, or a tool that syncs records to both via API.
  3. Check name server diversity. Name servers that all sit under one provider's network share a fate. Look at the NS set and see whether they are on genuinely separate infrastructure.
  4. Host the status page and alert channels independently. Alerts that depend on the failing domain won't arrive.
  5. Set sensible TTLs. Longer TTLs on stable records (an hour or more) give you cache cushion during an outage.

Registrar and DNS host are separate dependencies

Your registrar holds the delegation and your DNS host holds the records. An expired domain or a locked registrar account breaks things in the same way and needs a different fix. Track domain expiry too; see how to monitor SSL and domain expiry.

Monitoring it

A website monitor shows the symptom: every check fails with a resolution error, from every location, at once. A DNS monitor on the records that matter goes further and tells you when an answer disappears or changes, so you can spot trouble that is not a full outage, such as a record that was changed or removed by accident. For setup, see how to monitor DNS records.

Get new guides by email

Practical posts on monitoring, outages, error fixes and on-call, when they're published. No spam.

Unsubscribe from any email with one click. See our privacy policy.

Keep reading