This guide is for cases where you’ve imported your provider’s configuration into Shadowrocket, but a node times out or pages won’t load. Follow this order: keep test conditions consistent, identify where the failure occurs, check protocol settings, verify device time, and retry an actual request. This helps distinguish an unreachable network, authentication failure, TLS verification error, and incorrect rule selection—without changing several settings at once and losing useful clues.
1. Distinguish a failed latency test from a failed connection
A latency reading describes the result of one test; it doesn’t prove that all traffic will work. A TCP connection test mainly checks whether a connection can be made to a specified address and port. An HTTP request through a proxy also depends on authentication, transport negotiation, and the test server’s response. These tests use different methods and paths, so their results aren’t directly comparable. A single millisecond reading is not enough to judge connection quality.
A request can fail at any of these stages. For example, a port may accept a connection even when the password is wrong, so a basic TCP test may still return a value. Conversely, a test server may be unavailable, causing the test to time out while other sites remain accessible. For UDP-based protocols such as Hysteria2 or WireGuard, probing the TCP port with the same number is especially unlikely to tell you whether the actual protocol works.
- Only one node fails: Check that node’s configuration, the port provided by your service provider, and the service status first.
- Every node in the same subscription fails: Check whether the subscription needs updating, your account status, and the current network. Verify each separately.
- Results differ between Wi-Fi and cellular: Keep the configuration unchanged and check network restrictions, DNS, and UDP reachability first.
- Latency returns a value, but pages won’t load: Check authentication, TLS, Global Routing, and the policy applied to the request. Don’t stop at the latency result.
2. Choose one node and keep test conditions consistent
Before changing anything, note the node name, network type, test method, and original error. Don’t update the subscription, switch nodes, change DNS, and replace rules all at once. Change one thing at a time and retest against the same target to identify what made a difference. The steps below use Home, Config, and Settings on iPhone / iPad; on other platforms, use the corresponding settings.
-
Check the basic network connection
In Settings → On Demand, check whether automatic connection is enabled. You can turn it off temporarily while troubleshooting, but note the original setting first. Disconnect from Home, then confirm that your regular network can open a site known to be directly accessible. If a network sign-in page appears, sign in first.
-
Select a specific test node
Return to Home, select the node you’re testing, and note its name. If the current policy uses automatic selection, select a specific node manually for troubleshooting so successive requests don’t go through different nodes.
-
Repeat the latency test
On Home, run the app’s latency test for the same node and note the test method. Run it three times, recording each result or timeout. Don’t treat the best of the three results as proof of a stable connection.
-
Retest on another network
Keep the node and test method unchanged, switch between available Wi-Fi and cellular networks, and test again. While connected, try loading a web page and note the time and error so you can match it to the relevant log entries later.
If the regular network itself isn’t working, fix the device’s network connection first; repeatedly editing the proxy password won’t help diagnose that. If only one network fails, don’t delete the node right away. A successful request using the same configuration on another network is a useful check that the configuration is basically correct.
3. Check the address, port, authentication, and transport settings
Open the node’s edit screen from Home and compare each field with the original details provided by your service provider. Start with Type, server address, and Port; then check the password or UUID; finally, check transport and TLS settings. One incorrect field can prevent a connection. Similar names, matching port numbers, or a successfully scanned QR code don’t guarantee that every field is correct.
| Check | What to verify | What this does—and doesn’t—tell you |
|---|---|---|
| Address and Port | Check the server domain or IP address, port, and any unintended spaces. | If the original configuration specifies 8443, don’t change it to 443 just because TLS is enabled. |
| Shadowsocks | Check the encryption method and password. Any plugin settings required by the server must also match. | An accessible port doesn’t mean decryption and authentication will succeed. |
| VMess / VLESS | Check the UUID, transport type, and any related security settings included in the original configuration. | Copying only the address and UUID while missing the transport settings is not enough. |
| Trojan | Check the password, TLS, Server Name, and any server-specific requirements. | Even with the correct password, a certificate name mismatch can cause the connection to fail. |
| WebSocket | If the original configuration uses WebSocket, check Path and Host. | A mismatch in Path capitalization, slashes, or Host can prevent the handshake. |
| Hysteria2 / WireGuard | Check the authentication or key settings for the protocol, and confirm that the network allows the required UDP traffic. | A TCP port probe does not replace a test of the UDP-based protocol. |
If the node came from a subscription, first check whether your service provider changed the port, password, or transport settings. Update your own subscription, then review the fields in the refreshed entry. Continuing to edit an old entry can hide the fact that the configuration has changed. A successful subscription update only confirms that the list was retrieved and parsed; it doesn’t mean every node is online or that your account still has available usage.
4. Use the exact error message to locate the failure
Check the connection history or logs for the request you just made, and keep the original English error rather than noting only that it “doesn’t work.” The snippets below may appear in network library errors; exact wording and prefixes vary by protocol and platform. The absence of a particular message doesn’t prove that the corresponding stage is working.
Error: i/o timeout
What it means and what to do: A read or write operation did not complete before its timeout. Possible causes include packet loss, an unreachable port, or an unresponsive server. Check the address and stage shown in the log, then retry the same configuration on another network. This message alone doesn’t prove that the password is wrong.
Error: connection refused
What it means and what to do: The connection was actively refused. This often means no service is listening on the target port or a device policy is blocking it. Check the server address and Port, then ask your service provider whether that port is still in use.
Error: no such host
What it means and what to do: The domain name could not be resolved to a usable address. The cause may be a typo, DNS settings, or the domain’s status. Check the server domain first, then check your current network and DNS settings in Config. Don’t replace the domain with a guessed IP address.
Error: certificate has expired or is not yet valid
What it means and what to do: The certificate’s validity period doesn’t match the current time. The device clock may be wrong, or the server certificate may have expired. Correct the device time first. If the error persists, ask your service provider to check the certificate rather than disabling certificate verification.
For authentication errors, recheck the password, UUID, or key. If the log shows that a proxy tunnel was established and only certain sites fail afterward, check the target site, rules, and DNS. Logs may contain server addresses, subscription credentials, or authentication details. Before sharing them, keep only the necessary error context and redact sensitive information.
5. Check device time and TLS—don’t skip certificate verification
TLS certificates have start and expiration dates. If the device date is wrong, a certificate may appear not yet valid or expired. Some authentication methods also rely on a time window. So if the port seems reachable but the handshake fails, check the device clock instead of repeatedly changing node settings.
- iPhone / iPad: Open system Settings → General → Date & Time and check Set Automatically. If device management restrictions prevent you from changing this setting, contact your device administrator.
- Mac: Open System Settings → General → Date & Time. Check that date and time are set automatically and that the system can sync the time.
- Windows / Android / Linux: In your system’s date and time settings, check automatic synchronization and verify the actual date and time. If synchronization fails despite the setting being enabled, resolve the device’s network connection first.
- Retest after correcting the time: Disconnect and reconnect, then retry with the same node and target. If the certificate is still reported as expired or the name doesn’t match, ask your service provider to investigate.
A different time zone display doesn’t necessarily mean the system clock is wrong; what matters is whether the actual time is accurate. Don’t work around an error by disabling TLS, skipping certificate verification, or changing Server Name at random. These steps weaken security and may temporarily hide a server-side configuration problem. If the device time is correct and the certificate has expired, the client can’t renew it for the server.
6. Latency is normal, but pages fail: check Global Routing
A node test usually doesn’t cover the full path of an actual app request. Global Routing on Home determines how traffic is handled: Proxy sends traffic through a proxy, Direct connects directly, Config applies the rules in the current configuration, and Scene uses the settings for the selected scene. Don’t compare results from these different modes as if they were the same.
- Note the current Global Routing mode and selected node. If a request works in Proxy but fails in Config, check which rule matches and where its policy points.
- In Config, confirm which configuration file is active. Check whether an earlier
DOMAIN-SUFFIX,IP-CIDR, orGEOIPrule sends the target to another policy. Rules are evaluated from top to bottom, and the first matching rule is applied. - Check the final
FINALfallback. It handles only requests that haven’t matched an earlier rule; it cannot override a rule that already matched.FINAL,DIRECTandFINAL,PROXYsend traffic out differently. - If only domain-based sites fail, check DNS resolution as well. If only one target fails, compare it with other targets instead of resetting the entire configuration.
When troubleshooting is complete, restore the original routing mode and Settings → On Demand setting. Automatic connection only determines whether a connection starts under the specified conditions; it won’t fix a wrong password, unreachable port, or expired certificate. Keep a minimal record of reproducible test results—it’s more useful for future maintenance than leaving temporary diagnostic settings enabled.
7. Still timing out? Gather details for investigation
If the connection fails on both types of network, the settings match the original configuration, and the device time is correct, give your service provider a clear record of how to reproduce the problem. The client can’t determine whether the server is under maintenance, an account is restricted, or an upstream link is having issues; those require information from the server side.
- Your device’s operating system, network type, and whether the problem is continuous or intermittent.
- The node protocol and a redacted node identifier. Don’t share passwords, UUIDs, private keys, or full subscription links.
- The latency test method, all three results, and whether actual web requests fail too.
- When the error occurred, the Global Routing mode, the exact error message, and the individual checks you’ve already completed.