A latency result only shows that part of the tested path responded. It does not confirm that DNS resolution, rule-based routing, and the full website connection are working. This guide is for users who already have a provider, nodes, and a configuration. Steps focus on Home, Config, and Settings on iPhone and iPad; on Mac, Windows, Android, and Linux, check the equivalent settings in the same order. First distinguish DNS errors from connection timeouts, then isolate the cause with one-variable-at-a-time tests, and finally restore your usual setup.
1. Normal latency doesn’t mean the webpage loaded
A node test in Shadowrocket is not the same as loading a webpage in a browser. The test may only check a connection to the server port or request a specific test URL. Loading a webpage also requires resolving the domain, selecting an exit route, establishing a TCP or QUIC connection, and completing the TLS handshake. A latency reading can be returned even if the request fails at any later stage.
This flow helps narrow down the failure; it does not mean every connection resolves DNS locally first. Some proxied requests can send the domain to the remote server for resolution, while some rules need an IP address before they can be evaluated. A browser’s encrypted DNS may also use a separate resolution path. So if changing the system DNS makes no difference, that does not rule out DNS issues or prove that the node is unavailable.
These port numbers identify service types; they are not recommendations to change your node’s port to one of them. Node, DNS service, and website ports belong to different connections. Use the actual settings in your existing configuration.
2. Check the exact error before changing DNS
Record the full browser message, target domain, network, and time of the failure. The messages below are errors you may see in a browser or command line; they are not necessarily logs shown by Shadowrocket. Wording varies across systems, so keep the original English error code when diagnosing the problem.
Error: ERR_NAME_NOT_RESOLVED
What it means and what to try: The browser did not get a usable DNS result. Possible causes include the upstream DNS service, a typo in the domain, or the browser’s own resolution settings. Check the domain, then compare results in another browser and on another network.
Error: DNS_PROBE_FINISHED_NXDOMAIN
What it means and what to try: The browser’s DNS check reports that the domain does not exist, but a cached or incorrect response is still possible. Confirm that the domain should exist, then query it on another network and compare the results.
Error: curl: (6) Could not resolve host
What it means and what to try: curl could not resolve the specified hostname. Check the input and system resolution results on the same computer; don’t assume this means the proxy server refused the connection.
Error: ERR_CONNECTION_TIMED_OUT
What it means and what to try: The connection did not complete in time; this error is not specific to DNS. First check whether the target IP was resolved, then check the matched rule, exit route, and website port’s reachability.
If the page reports a certificate name mismatch or an incorrect device time, check the system date, public Wi-Fi sign-in status, and certificate warning first. Don’t disable certificate verification to “fix DNS.” If the main page loads but images, sign-in, or video fail, record the domains of the failed resources separately instead of testing only the main domain in the address bar.
3. Set up a reproducible test in Home, Config, and Settings
Save a copy of your current configuration before you begin, and note the node in use and the Global Routing setting. Change one thing at a time and test the same full URL each time. These steps help narrow down the cause; you don’t need to clear your existing rules or reinstall the app first.
-
Record the current setup
In Home, note the selected node and Global Routing setting. In Config, confirm which configuration is enabled. Also record whether you’re on Wi-Fi or Cellular, the browser error, and the failing domain so you can compare results after switching settings.
-
Pause automatic connections
Go to Settings → On Demand and temporarily disable any enabled connection conditions, then return to Home to control the connection toggle. Otherwise, testing while disconnected may trigger an automatic reconnection, making the comparison unreliable.
-
Check the underlying network
Disconnect and open a website that should be reachable directly to confirm that Wi-Fi sign-in is complete. Then compare Wi-Fi and Cellular. Don’t use a site that requires a specific exit route to judge whether the underlying network is working.
-
Compare routing modes
Reconnect, then test the same URL with Config and Proxy under Home → Global Routing. Use Direct only for destinations that should be accessed directly. If you use Scene, note its settings first, then switch to a clearly defined mode for testing.
-
Restore your original settings
Record which combination worked, then restore your original Global Routing and On Demand settings. Keep the configuration copy and make further changes only to the issue you’ve isolated. Don’t treat a temporary Proxy test as the final fix.
Takeaway: Use comparisons that isolate each layer
If Proxy opens the site but Config fails with the same node, check the active configuration’s rules, DNS settings, and policy selection first. That result alone does not prove there is a DNS issue.
4. Check the DNS resolver, transport, and active configuration
In Config, find the active configuration and check its DNS settings in the editor. If you’re editing text, look for [General] and dns-server. This setting specifies DNS resolvers; don’t enter a subscription URL, node address, or full website URL as a standard DNS server address. The layout in the visual editor may vary by platform, so check the interface you’re actually using.
| Setting to check | What it means and how to check it | Common pitfall |
|---|---|---|
dns-server | Check which resolver the active configuration specifies and whether it still uses system DNS settings. | Changing Wi-Fi DNS without checking the in-app configuration. |
| Traditional DNS | Typically uses UDP / TCP port 53; the current resolution path must be able to reach the service. | Assuming the request will reach the server just because its address is correct. |
| DNS over HTTPS | Uses HTTPS to carry DNS queries; the endpoint address and format must match the configuration syntax supported by the app. | Using a regular webpage URL as a DoH endpoint, or overlooking the need to resolve the endpoint’s own domain. |
| A / AAAA | These are IPv4 and IPv6 address records, respectively. Check whether the relevant network path is actually reachable. | Assuming that an AAAA record means the current IPv6 route is working. |
| Browser encrypted DNS | Check whether the browser specifies its own resolver, and compare its results with system DNS. | Treating results as the same test when the browser and app use different resolution paths. |
Don’t change DNS, disable IPv6, switch nodes, and edit rules all at once. First compare results using a resolver already confirmed to be reachable on the current network, then test other transport methods. If a DoH endpoint’s domain needs to be resolved before the endpoint can be reached, and the resolution path it depends on is already broken, you can end up in a loop where the DNS added to fix DNS cannot connect either.
- Before changing anything: Save the original configuration and note its DNS settings so you can revert if needed.
- After making a change: Save it and confirm that the edited configuration is active. Reconnect before testing again.
- If results differ: Close and reopen the browser, then test the same domain again. The browser, system, and app may each retain a cache.
- If only IPv6 fails: First check whether the address returned in the AAAA record is reachable. Then try a temporary IPv4 comparison; disabling IPv6 long-term is not a general fix.
5. Tell DNS failures apart from incorrect Config routing
Config under Global Routing handles connections according to the active configuration’s rules. Proxy sends applicable traffic through the proxy route, Direct uses the direct route, and Scene selects a handling mode based on its settings. These are important variables to check, but switching to Proxy does not guarantee that all DNS requests will automatically use remote DNS. The result also depends on the DNS settings and how the connection is handled.
Find the failed request in the connection log and check the target domain, visible destination IP, matched rule, and final route. DOMAIN-SUFFIX matches domain suffixes; IP-CIDR matches IP ranges; GEOIP uses the geographic location of an IP; FINAL is the fallback when no earlier rule matches. An IP-based decision may require a DNS result or use a destination IP already present in the request.
[Rule]
DOMAIN-SUFFIX,example.com,DIRECT
IP-CIDR,192.168.0.0/16,DIRECT,no-resolve
GEOIP,CN,DIRECT
FINAL,PROXY
These examples illustrate rule syntax; they are not recommendations to replace your existing configuration. example.com is a sample domain; 192.168.0.0/16 is a private IP range; no-resolve means not to actively resolve the domain just for this rule. Before copying a rule, understand its order in the existing ruleset and how the active configuration defines its policy names.
- Proxy works, Config fails: Check for an earlier DIRECT match, an incorrect policy, or a reject rule, then check the DNS resolution path.
- Both fail: Check whether the node server’s domain resolves, whether the authentication details are valid, and whether your existing service allows access to the destination.
- Only one website fails: Compare the rules matched by the main domain and resource domains, then check the website’s status. Don’t replace the entire DNS configuration right away.
- An IP is available but the connection still times out: Focus on exit-route reachability, the port, TLS, or website restrictions. Having an IP does not prove that resolution was correct, but the issue is no longer simply “no DNS result.”
6. Use lookup results on desktop as supporting evidence—not IP access as a substitute for testing
On Mac, Windows, or Linux, you can use built-in system tools to inspect DNS results. The commands below query a sample domain; for a real issue, replace it with the domain that is failing. If the relevant tool isn’t installed, skip that check—there’s no need to install extra software just for troubleshooting.
nslookup example.com
nslookup -type=AAAA example.com
curl -I --connect-timeout 10 https://example.com/
nslookup can show whether a query returned an address, NXDOMAIN, or a timeout; curl can then test an HTTPS request. If the server does not support HEAD requests, curl -I may return 405, which is different from a DNS resolution failure. Keep in mind that these tools may use a different resolution path from the browser. A successful command only confirms that the tool’s request worked this time.
Takeaway: Record DNS and connection results separately
Record three details—whether an A / AAAA record was returned, whether the request timed out, and the HTTPS response status. This is more useful for diagnosis than simply noting whether the site loaded. A certificate error after replacing an HTTPS domain with its IP address is not evidence of a DNS problem.
HTTPS sites often rely on the domain for SNI and certificate validation, and accessing an IP directly may reach a different virtual host. So whether a browser can open the IP cannot fully prove that DNS is working or that the original website is at fault. If both resolution and the connection work on another network, first check DNS reachability, router settings, and sign-in status on the original network.
7. Keep the fix from being overwritten by a configuration update
Updating a node subscription and updating a configuration are not the same thing. Refreshing a node subscription usually syncs the server list or parameters; it does not fix DNS. Updating a remote configuration may change DNS and rules. Before updating, confirm what is being updated and save any local changes you need to keep.
- If the node server uses a domain name, make sure that domain also resolves. Otherwise, the connection can fail before it reaches the destination website.
- If your provider has notified you of changed node parameters, update them through your existing subscription. Don’t guess the server address, port, or authentication details.
- Keep a working copy of the configuration and note the DNS settings, rules, and network conditions involved in the fix. Don’t include subscription keys in public records.
- After testing, restore any required conditions under Settings → On Demand and check Global Routing in Home again.
- When reporting an issue, share the redacted error text, network type, and comparison results. Hide full subscription URLs, passwords, UUIDs, and private keys.
This guide covers using and troubleshooting subscriptions and configurations you already have. Shadowrocket is a closed-source commercial app; the app and network service are separate, and buying the app does not include a service plan. The iPhone / iPad version is available on the App Store for a one-time $2.99 purchase from Shadow Launch Technology Limited; system requirements are listed on the App Store page. The Mac (Apple silicon / Intel), Windows, Android, and Linux versions are free.