For users with a connection profile who can't open a webpage or find that only some sites fail. First compare the same domain with the connection on and off. Then use the error, DNS results, and matched rule to determine whether the issue is with DNS, the proxy or direct route, or the subsequent HTTPS handshake.
Latency tests look normal, so why won't webpages load?
A latency result in Home only means that the test received a response under the conditions at that moment. It doesn't mean the browser's domain resolved, or that the browser used the same route. Even if the test checks a URL, it verifies only that test target. The actual webpage may also load scripts, images, or sign-in pages from other domains, and any one of those failing can leave the page loading indefinitely.
- Use consistent test targets. In the same browser, retry one URL that fails and one that usually works. Record the full failing hostname, such as
www.example.com, not just the root domain; their DNS records and matched rules may differ. - Establish a baseline with the connection off. On the same network, turn off the connection in
Homeand try again. If it reconnects automatically, checkSettings → On Demand. On devices where this setting can be changed, you can temporarily disable it for the comparison, then restore the original setting after testing. - Change only one variable at a time. Turn the connection back on over the same network, keeping the original profile and server, and try the same URL. To compare Wi-Fi and cellular, run a separate test; don't change the network, DNS, and
Global Routingat the same time.
Tell DNS errors apart from connection timeouts and certificate issues
These are error codes a browser may display, not messages from the Shadowrocket interface. Note the exact error and when it appeared, then check diagnostic records for the same hostname at the same time. If all you see is a generic “can't open webpage” message, don't assume DNS is the cause.
Error: ERR_NAME_NOT_RESOLVED
Cause: The browser couldn't resolve the current hostname to a usable address. The name may be misspelled, the record may not exist, or the DNS resolution path may have failed. What to do: Check the full hostname in the address bar, compare the result with the connection off, and check which resolver the active profile uses.
Error: ERR_CONNECTION_TIMED_OUT
Cause: The connection wasn't established before the timeout. The error code alone doesn't show which part timed out. What to do: If diagnostic records show the destination IP, check the matched policy and destination port. Standard HTTPS webpages usually use port 443.
Error: ERR_CERT_DATE_INVALID
Cause: The device clock is incorrect, or the site's certificate is outside its validity period. What to do: Check the device's date and time, then try the same hostname on another network. Don't bypass certificate warnings or start by changing DNS.
Check the DNS settings in the active Config
In Config, confirm which profile is active, then check the [General] section in that profile—not a different copy with the same name in the list. If dns-server is set to system, the system-provided DNS is used. The resolver the system uses may change when you switch networks. If the profile specifies a server address or DoH URL, make sure the current network and route can reach that DNS service.
Bottom line: before changing DNS, distinguish between an error response and no response
If you get NXDOMAIN, first check for a misspelled hostname or one that exists only on an intranet. If the query times out, first check whether the device can reach its current resolver. These symptoms call for different troubleshooting steps; there's no need to change several settings at once.
NXDOMAINvs.SERVFAIL: The former means the resolver returned that the name doesn't exist. The latter means the resolver couldn't complete the query, possibly because of an upstream issue or validation failure; it doesn't mean the domain doesn't exist.- DoH URL: Check the hostname and path required by the DNS service you're using. Being able to reach port 443 doesn't mean the URL returned a DNS answer successfully. A hostname resolution issue, TLS validation failure, or incorrect request path can all cause the query to fail.
- Static
[Host]mapping: If the active profile already maps this hostname, check that the address is still valid. An outdated mapping may give the app an IP address that doesn't connect to the right server; changing the upstream DNS won't fix the mapping. - Test and restore: Keep the original profile unchanged, and change just one DNS setting in a test copy. Reconnect and retry the same two URLs. If the results don't change, restore the original profile and check the routing. When troubleshooting an intranet, keep its existing DNS settings.
If only some domains fail, check Global Routing and your rules
Once the destination domain resolves to an address, check where the connection is routed. Note the current mode in Home → Global Routing. In Config mode, rules in the active profile determine the route. Switching to Proxy or Direct changes where traffic goes, so access returning after a switch doesn't prove the original DNS settings were wrong. If you were using a Scene, note which scene was active when comparing Wi-Fi and cellular to avoid mistaking a scene change for a DNS difference.
In Config mode, check the first matching rule from the top of [Rule]. Don't just check whether the file contains a rule that looks right. This snippet illustrates rule order; it isn't a ruleset to replace your existing configuration with:
[Rule]
DOMAIN-SUFFIX,example.com,PROXY
FINAL,DIRECT
www.example.comcan match the first rule; a similar-looking name with a different suffix won't. If no other rule matches,FINALdetermines the route. If the homepage and sign-in request use different policies, check the matched rule for each hostname separately.- If the profile uses
IP-CIDRorGEOIP, also check the destination address actually in use. A domain can resolve to different IPs on different networks, which can change the result of IP-based matching. - If an
Aquery for a domain returns a result but anAAAAquery doesn't, that alone doesn't indicate a DNS failure. If both address types are returned but the connection log shows timeouts for only one, check the corresponding IPv4 or IPv6 route separately.
Common DNS troubleshooting questions
A site won't load on Wi-Fi but works on cellular. What should I check first?
Reconnect to the original Wi-Fi network and open the same public webpage with the Shadowrocket connection off. If it still fails, check whether the Wi-Fi network requires a sign-in and whether its DNS is working. If it fails only with the connection on, check the resolver and routing in the active profile. Use the same URL for both tests.
On Mac, dig returns an answer. Why won't the browser load the site?
dig example.com A only shows that the resolver path used by that command returned an IPv4 result; the browser or proxy may use a different path. Replace the example with the hostname that's actually failing, query AAAA if needed, then check whether the browser has its own secure DNS enabled and compare the results with Shadowrocket's connection records.
Why does the error persist briefly after switching back to the original DNS?
The system, browser, and upstream resolver may all cache results, including results that a name doesn't exist. Restore the original profile, reconnect, and retry the same URL; reconnecting doesn't guarantee that every cache is cleared. If DNS records have just changed, wait for the relevant cache to expire before comparing results, and avoid repeatedly switching settings in the meantime.
Latency still looks normal. What should I include when reporting the issue?
Include the device OS, network type, time of the issue, failing public hostname, browser error, Global Routing mode, and results with the connection on and off. If diagnostic records are available, note the DNS status and matched policy too. Before sharing, remove sensitive information such as subscription URLs, access tokens, and intranet hostnames.