01

A reachable IP with no working domain

Run one test first. Ping a public address such as 1.1.1.1 and watch it reply, then open any website and watch it fail. When both are true, the fault sits in name resolution, and everything below is aimed at that layer.

Next, find out who resolves. With TUN running, or with DNS override enabled in the config, the client takes over resolution and editing the DNS servers on your network adapter changes nothing. With both switched off, the opposite holds: whatever you adjust inside the client has no effect on what the system resolves. Getting this backwards wastes a lot of time.

Read the symptom before changing anything
Resolution faults and dead nodes look almost identical from the browser. The ping test is what separates them.

SponsoredWhere does the subscription link come from?Our partner provider gives you 1 GB of high-speed Hong Kong data at signup — import it in one click.Get high-speed nodes
02

Three faults that cause most of it

  1. 1

    Stale results in a cache

    After a node switch or a config change, the system and the browser can still be serving the old answer, which reads as a fix that did not take. On Windows run ipconfig /flushdns. On macOS run sudo killall -HUP mDNSResponder. The browser keeps its own layer on top of that, so close it completely and open it again.

  2. 2

    Leftover fake addresses

    If the program exited abnormally, routes or cached entries pointing at the 198.18.0.0/16 range can survive in the system. The symptom is specific: a handful of domains resolve to an address nothing can reach while everything else behaves. Start the client again so it takes the layer back, then flush once more.

  3. 3

    Internal names sent outside

    Company domains and local device names get treated as public domains and shipped off to an external resolver, which of course has never heard of them. Point those suffixes at an internal nameserver in the DNS section of the config, or add them to the list that skips fake addresses so they resolve to their real ones.

03

Confirm the answer actually changed

Do not stop at the page loading. Query the same domain again from a terminal and read the address it returns. If a domain resolves to an obviously wrong region, the query took a route you did not expect, and the log will tell you which nameserver answered it.

  • The same domain queried twice returns the same address and no longer starts with 198.18
  • Internal hosts and public domains both open, rather than one at the expense of the other
  • It still works after a full browser restart, which rules out a leftover cache
04

Common questions

I switched nodes and the domain still resolves to the old address. What is going on?

Almost always a cache. The answer is stored separately by the system and by the browser, and switching nodes clears neither. Run ipconfig /flushdns on Windows or sudo killall -HUP mDNSResponder on macOS, then quit the browser fully and open it again. If it survives that, restart the client so it re-establishes the resolution layer.

Closing Clash Verge left me with no internet at all. Why?

The program did not hand the layer back when it exited, so the system is still pointed at DNS settings or routes that no longer lead anywhere. Start the client and quit it normally once, which usually restores it. You can also set the adapter DNS back to automatic yourself and then flush the cache.