When DNS stops working
The most convincing impersonation of a total internet outage. Your connection is perfect, every packet would arrive, and nothing knows where to send them.
The most convincing impersonation of a total internet outage. Your connection is perfect, every packet would arrive, and nothing knows where to send them.
Every connection you make starts with a question: What address does this name have? That question goes to a resolver, and if the resolver does not answer, nothing else happens. Your connection is perfect. Your router is fine. Your ISP is delivering packets flawlessly. And nothing loads, because nothing knows where to send anything.
This is the single most common cause of the experience people describe as "the whole internet is down", and it is also one of the fastest things to fix once identified.
Open https://1.1.1.1. That is a numeric address and needs no resolution.
If a page appears, your connectivity is fine and DNS is broken. If nothing appears, you have a genuine connectivity problem and DNS is not your issue.
With a terminal, the more precise version:
nslookup example.com
nslookup example.com 1.1.1.1
The first uses your configured resolver, the second bypasses it. If the second works and the first does not, you have isolated the fault exactly.
ISP resolvers are unglamorous infrastructure and they do fail. When they do, every customer of that ISP experiences a total internet outage while their connection is entirely healthy.
Public resolvers fail too, and the blast radius is larger because they are configured on routers worldwide. A 1.1.1.1 or 8.8.8.8 incident produces a sudden global flood of unrelated outage reports from people who have no idea their router is even using one.
Answers are cached at every level: Your browser, your operating system, your router, your ISP's resolver. When a site moves, an old answer somewhere in that chain keeps sending you to an address that no longer serves it.
Negative answers are cached too, which is the part that surprises people. If you tried a domain just before its records became valid, you may hold the "does not exist" answer for hours. Clearing caches is the fix, and you can only clear the ones you control.
If a domain's authoritative servers are unreachable, nothing can resolve it. This affects everyone and there is nothing you can do - And it is the mechanism behind the most complete outages there are. When Meta withdrew its routes in 2021, its nameservers disappeared from the internet and every Meta hostname stopped resolving globally.
Ad blockers at the router level, parental controls, corporate DNS policies and some ISP-level blocking all work by returning false answers. A blocked domain typically returns NXDOMAIN or an address pointing at a block page, which is indistinguishable from a real failure unless you know the filtering is there.
VPNs push their own resolvers. A misconfigured client can leave you with a resolver that is unreachable over the tunnel, or one that answers with internal addresses. Disconnecting is an instant test.
| Resolver | Addresses | Notes |
|---|---|---|
| Cloudflare | 1.1.1.1 / 1.0.0.1 | Fast; states it does not log identifying data |
| 8.8.8.8 / 8.8.4.4 | Very widely deployed and extremely reliable | |
| Quad9 | 9.9.9.9 | Blocks known-malicious domains |
Set it on the router and every device on the network benefits. Set it on one device to test without affecting anyone else. Step-by-step instructions for every platform, and how to choose between them.
Do not configure only one. A single resolver with no fallback means its outage is your outage. Configure two from different operators.
Worth saying plainly: Your resolver can see every domain you visit, even though it cannot see the pages. That is a meaningful amount of information, and it is why resolver privacy policies matter and why DNS over HTTPS exists.
DoH encrypts the query so your network cannot read or tamper with it. Browsers increasingly enable it by default, which has a practical side effect worth knowing: It bypasses router-level ad blocking and parental controls, because those work by intercepting exactly the queries DoH hides. If your Pi-hole suddenly stopped working, this is usually why.
DNS gets blamed for a lot that is not its fault. If nslookup returns an address and the site still will not load, resolution worked and the problem is further along - A connection failure, a certificate error, or the site being genuinely down. The connection errors guide covers what happens after the name resolves.
Open an IP address directly, such as https://1.1.1.1. If that loads and domain names do not, your connection is fine and DNS is the fault. If neither loads, the problem is connectivity rather than name resolution.
Cloudflare (1.1.1.1), Google (8.8.8.8) and Quad9 (9.9.9.9) are all fast and reliable. The important part is configuring two from different operators, so that one resolver having an outage does not become your outage.
Because it bypasses whatever was wrong with the old one - A stale cache, a resolver that is down, or filtering that was returning false answers. It does not fix the site itself; it fixes the step where your device learns the site's address.
Yes, and it is the most common cause of that experience. When a resolver stops answering, no name resolves at all while your connection stays perfectly healthy - Every site, every app, every game fails simultaneously.
Exact commands for flushing the DNS cache on every major platform, plus the browser and router caches people forget about.
What DNS_PROBE_FINISHED_NXDOMAIN means, the difference between a genuinely non-existent domain and a resolver problem, and how to tell them apart.
A layered diagnostic for working out whether the fault is your device, your network, your ISP, DNS, or the website itself.
How to diagnose a router or Wi-Fi problem from the status lights, and the difference between a line fault, a Wi-Fi fault and congestion.