Total blackout and partial failure are different bugs
Answer one question before you change anything: does every site fail, or only some of them. The two symptoms almost never share a cause.
A total blackout usually lives on your own machine. The system proxy points at a port nothing is listening on, or a leftover virtual adapter is still holding the route. Partial failure is where nodes and rules finally matter. People who skip this question end up cycling through fifteen nodes to fix something that was never on the remote side.
Glance at the log page first
If not a single line is being written, the core is not running and none of the steps below will tell you anything.
Three cuts, each one halves the search space
- 1
Switch to direct, open an ordinary site
In direct mode nothing is proxied. If a normal page loads, your machine and the client are both alive and the fault is somewhere in the proxy path. If it does not load, stop here and fix the underlying network connection, because the client is not involved.
- 2
Switch to global, pick any node
If it works in global but not in rule mode, your rules are sending that traffic somewhere unexpected. Look the domain up on the rules page to see which line caught it. If global fails too, try two or three nodes in different regions before concluding anything, because a single dead node is extremely common.
- 3
Change networks and repeat
When every node fails, check whether the subscription has expired or been suspended. If it looks fine, tether to a phone and try once. Working on the tether tells you the original network is interfering with this kind of traffic, and no amount of node switching will fix that.
Writing down the symptom beats restarting
Once traffic comes back, spend ten seconds confirming it properly. The connections page should be scrolling, the log page should be free of repeating errors, and local sites should still go direct after you return to rule mode. Testing a single website proves very little, because a cached page will happily load through a broken proxy.
- You can name the step where it stops working, not just say it is broken
- You know whether the symptom changes on a different network, which alone separates local faults from route faults
- Every other proxy application is closed before you draw a conclusion
Common questions
Every node shows a timeout. What do I check first?
Check whether the latency test target is reachable from your network at all. When the test address itself is blocked, perfectly good nodes all turn red. Point the test at something you actually use and run it again. That separates a dead node from a dead test far faster than swapping nodes.
My browser works but nothing else does. Is the node at fault?
No. That pattern means the proxy path is fine and those applications simply do not read the system proxy setting. They were never proxied. Either configure them individually or use TUN mode, which does not depend on applications cooperating.