Start from the network you had before
- Turn the proxy off and check a site you normally reach directly. If the base network is down, fix Wi-Fi, mobile data or the router first.
- Check the plan and the subscription. Look at expiry and remaining quota in the provider panel, and confirm the last update produced no error. A profile name in the client does not mean the plan is active.
- Keep one client running. Multiple proxies, VPNs or browser proxy extensions interfere with each other. Quit the others before testing. If you changed system proxy settings, restore them per that client's documentation.
- Change the node, then change the target. Switch to a different node and test several sites. If only one site fails, the cause may be its region policy, your account, an outage or a rule — not every line.
Four outcomes that place the fault
| Outcome | Fault is most likely in | First action |
|---|---|---|
| Sites also fail with the client off | Local network | Fix connectivity; do not touch the client yet |
| Subscription update fails | Fetch or format | Handle the update path; check status codes and content type |
| Every node times out | Account, network or core | Check the plan, the base network and the core log |
| One application fails, others work | System proxy coverage, TUN, or that app's own networking | Check the local port, the capture method and the app's proxy support |
Classify the symptom before changing anything. Without a category, every change is a guess and every recovery is unattributable.
System proxy works, other applications do not
Browsers generally honour the system proxy. Games, command-line tools and some store applications often do not. Confirm the client's local port and that the system proxy is actually enabled; only then consider TUN for broader capture.
Before enabling TUN, record the existing DNS settings, routes and permission state. If TUN fails, disable it first and restore plain networking, rather than layering more changes on top.
Every site shows a name-resolution error
- Compare several different domains, and confirm the system clock is correct.
- Do not add several public resolvers at once; that makes the comparison unreadable.
- If resolution works with the client off and fails with it on, examine the client's DNS mode, TUN DNS capture and any configuration override.
- Record the original values before restoring defaults, item by item.
One site only: do not reset the system
Test other sites on the same node, then the same site on another node. If only the target site fails, the candidates are rule matching, exit region, account risk control, the site's own outage, or an IPv6 path difference. Keep the error code and the matched rule — that is far more diagnostic than “I tried many nodes”.
Record it in a form someone can use
- Time with time zone.
- Device OS and client version.
- The exact error text.
- Which applications worked and which did not.
- What changed after switching node or network, one variable at a time.
Mask the subscription URL, account token and personal identifiers in any screenshot. Never publish a password “just for diagnosis”.
Final step: verify you can get out
- Turn off TUN, the system proxy and the VPN switch.
- Quit the client completely.
- Confirm ordinary sites load again.
- If they do not, check for a leftover
127.0.0.1proxy value with an old port, a system VPN still set to on-demand, or another client running in the background.
Resetting the entire network stack is not step one: it changes several variables at once and can affect unrelated software.
Scope
This checklist applies to a client that reports connected while a browser or application cannot reach sites. Button names follow each client's version. The layering — base network, subscription, node, capture, DNS, target — is what stays stable across versions.