When a website stops loading, everyone checks the web server first. Is the machine up? Is the application crashing? Those are sensible questions, but a surprising number of outages have a different cause: the servers are fine and visitors simply can't find them. DNS is the first thing every request touches, so when it fails, everything behind it looks dead.
Why DNS failures look like total outages
Before a browser can send a single byte to your site, it needs your domain's IP address. If that lookup fails, there's no second chance and no partial page. The error says "server not found", your monitoring tool reports downtime, and yet nothing is actually wrong with your hosting.
The most famous example is the October 2016 attack on Dyn, a large DNS provider. Sites like Twitter, Netflix and Reddit were unreachable for some users for hours, even though their own infrastructure was healthy. They all depended on the same DNS company, and that was enough.
The usual suspects
- An expired domain. The registry stops publishing it, and it can vanish from DNS within hours. A status of
clientHoldorserverHoldin a WHOIS lookup is the giveaway. - Wrong nameservers after a migration. The registrar still lists the old provider, whose zone is empty or was deleted.
- A deleted or edited record. One wrong click in a control panel is enough. Typos in an IP address are depressingly common.
- A DNSSEC failure. Validating resolvers refuse to answer, while non-validating ones carry on, which makes the outage look random. See our DNSSEC guide.
- A provider outage. A single DNS host with a bad day takes every domain on it down.
- A TTL that was too long. You fixed the record in minutes, but caches kept the broken one for hours.
- A third-party dependency. If your
wwwis a CNAME to a CDN or SaaS hostname, their DNS problems become yours.
Prevention that actually works
Use more than one DNS provider
This is the single biggest improvement for the effort. With a secondary provider, you publish the same zone in two places and list nameservers from both. If one provider disappears, resolvers move on to the other. It does cost a little, and you must keep the two in sync, but few measures buy more resilience.
Protect the domain registration
Turn on auto-renew, keep a working payment method, and put the expiry date in a calendar that more than one person sees. Enable the registrar lock so the domain can't be transferred or changed without a deliberate step. Make sure the contact email still reaches a person, because renewal warnings go there.
Pick TTLs for the record's job
Records you might need to change in an emergency, such as the address of your main site, deserve a TTL of five to fifteen minutes. Records that never change can be a day or more. The reasoning is in our guide to TTL values.
Keep a copy of your zone
Export your records somewhere outside your DNS provider every time something changes. When an account is locked or a zone deleted, you rebuild in minutes instead of reconstructing from memory.
Review changes before they go live
Treat DNS like code. Have a second person look at edits to important records, and write down what you changed and why. Most outages caused by DNS come from changes, not from attacks.
Monitor from outside
An uptime check running from the same network as your server proves little. Check from several regions and, ideally, check DNS separately from HTTP so you know which layer failed. Our DNS checker is handy for a manual look from dozens of countries at once.
A five-minute diagnosis when the site is "down"
Work from the outside in.
- Does the name resolve anywhere? Run a public lookup from a few resolvers. If all say NXDOMAIN or time out, the problem is upstream of caches.
- Is the registration healthy? Check the domain's expiry date and status.
- Do the nameservers match? Compare what the registry lists with what you expect:
dig NS example.com +short dig +trace example.com
- Do your authoritative servers answer? Query them directly with
dig example.com @ns1.yourprovider.com. If they respond correctly while public resolvers don't, you're looking at caching or DNSSEC. - Is the answer right? Compare the IP returned with the address your server really has.
- Only then look at the server. If DNS is correct everywhere, move on to HTTP, TLS and the application.
Handy shortcut: if the answer differs between locations, the issue is almost certainly caching, a recent change or DNSSEC. If it's identical and wrong everywhere, you have a bad record.
Write down what happened
After the fire is out, spend ten minutes on a short note: what changed, how you found it, how long it took, and what would have caught it sooner. It doesn't need to be formal. A paragraph per incident is enough, and a folder full of them becomes the best DNS runbook a small team can have. Patterns appear quickly, because the same two or three causes tend to repeat.
What to do during an incident
Fix the record, then be patient with the clock. Lowering the TTL now won't help the caches that already stored the old answer, because they only look at the new TTL after the old one expires. If end users are affected, tell them that clearing the local cache (see how to flush DNS) may help, and keep checking from several locations until they all agree. For individual visitors seeing browser errors, our guide to DNS error fixes is worth sharing.