For Shadowrocket users on iPhone, iPad, Mac, Windows, Android, and Linux who see a connection marked as active and a latency value, but webpages still won’t load. First distinguish DNS failures from connection failures, then check the active config, Global Routing, and DNS responses. Change one setting at a time. The steps use Home, Config, and Settings on iPhone / iPad as a reference; on other platforms, look for the equivalent feature rather than assuming the menus are identical.
Identify where the failure occurs
A latency test in Shadowrocket only shows that one type of probe received a response at that moment. It doesn’t prove that every website is reachable. Depending on the test, it may check a server port or contact a fixed test target. A browser request also depends on DNS resolution, rule selection, the destination connection, and the TLS handshake. Success at one stage doesn’t mean the stages that follow are working.
This flow helps narrow down the problem; it doesn’t mean every protocol handles each step in exactly the same order. For example, DOMAIN-SUFFIX can match a domain directly, while IP-CIDR or GEOIP checks may require the destination IP. Some proxy requests may also pass the domain to a remote server for resolution. So neither “DNS always happens locally” nor “using a proxy means local DNS isn’t needed” is universally true.
Different ports can have different reachability. Being able to browse websites doesn’t mean UDP 53 is reachable on every DNS server. Likewise, a resolver returning a record doesn’t mean the HTTPS service at that address is reachable. Don’t change all your DNS settings based on a single latency result.
Compare results step by step
Keep a copy of your current config and note the affected domain, network, Global Routing mode, and time of the error. Use the same device and domain throughout testing, changing only one condition at a time. These steps don’t require adding a subscription. If you need to verify an existing subscription, get the relevant details from your service provider.
-
Check the basic network connection
Turn off the connection in Home and visit a website that normally works without a proxy. On public Wi-Fi, sign in through the captive portal first if needed. If the internet is still completely unavailable with the connection off, troubleshoot the system network before changing Config.
-
Check the active config
Open Config, confirm which config file is selected, then view its contents and check [General]. Save a copy before editing. Changing a file that isn’t selected won’t affect the current connection.
-
Check the routing mode
Go to Home → Global Routing and note whether the current mode is Config, Proxy, Direct, or Scene. Reproduce the issue in Config, then temporarily compare with Proxy. Restore the original mode afterward; don’t treat a temporary test mode as the fix.
-
Check automatic connection behavior
If the connection turns back on after you disable it, check Settings → On Demand. Note the original setting and temporarily disable the relevant automatic connection behavior while troubleshooting. Otherwise, your “connection off” test may still be running through the tunnel.
-
Test on another network
Switch between Wi-Fi and cellular, keeping the same config and destination domain. After saving a config change, reconnect and reload the test page so you don’t mistake a result from an existing connection for one produced by the new setting.
If the issue occurs only on one Wi-Fi network, first check DNS reachability, router filtering, and sign-in status on that network. If Config fails but Proxy works, investigate rule selection and the selected outbound. This narrows down the cause but doesn’t, on its own, prove that the DNS setting is wrong. Scene may apply a preset; don’t treat it as a fixed global mode unless you’ve checked what the scene does.
Check DNS responses
On the same Mac, Windows PC, or Linux device with the relevant tools installed, use nslookup to inspect the response status. The first command queries the resolver used by the system; the second explicitly queries a public resolver for comparison. These addresses are examples for diagnosis only. Reachability depends on the current network, so they aren’t suitable for every network.
nslookup example.com
nslookup example.com 1.1.1.1
These commands don’t read the Shadowrocket config directly. System routing, tunnel handling, and DNS interception can all affect the path a query takes, so also note whether the connection is enabled and the current Global Routing mode. There’s no need to install additional tools on iPhone / iPad for this step. First compare the network and config, then assess the available diagnostic results.
Error: NXDOMAIN
Cause: The resolver reports that the domain doesn’t exist. The domain may be misspelled, its records may have been removed, or the resolver may be filtering the response.
Fix: Check the full domain name and query it using another reachable resolver. Confirm the records with the domain or service administrator. Don’t mistake “no record” for a simple timeout.
Error: SERVFAIL
Cause: The resolver couldn’t process the query. Possible causes include an upstream failure, an authoritative server issue, or DNSSEC validation failure.
Fix: Compare the result with another resolver. If only one domain keeps failing while others work, investigate that domain’s DNS chain; there’s no need to reset the entire client right away.
Error: DNS request timed out.
Cause: This timeout message may appear in nslookup. It means no response arrived within the allotted time, not that the queried domain doesn’t exist.
Fix: Try another network first, then check whether the specified resolver and port are reachable. Repeatedly increasing retries along the same path may not help.
These are possible DNS tool statuses or outputs, not claims about fixed dialogs that Shadowrocket always displays. Even after receiving an A or AAAA record, check whether the address is expected and where the subsequent connection fails. Opening an HTTPS site by IP isn’t a complete substitute for testing by domain: certificate names, SNI, and virtual-host routing still depend on the domain name.
Understand the General settings
Check DNS settings in the [General] section of the active config. Don’t append a second [General] section to the end of the file, and don’t leave conflicting settings with the same name. Find the existing entry and change its value. Support for extended options can vary by platform and build, so check the in-app documentation for your version.
| Setting or record | What it does | Troubleshooting notes |
|---|---|---|
dns-server |
Specifies the DNS resolvers used by the config; system means use the system DNS. |
System DNS can change with the Wi-Fi network, cellular connection, or network configuration; it isn’t a fixed public address. |
ipv6 |
Controls IPv6 support in the config and is relevant when troubleshooting AAAA responses and IPv6 connections. | This isn’t a DNS server address, and it doesn’t disable IPv6 on every network interface on the device. |
A / AAAA |
These refer to IPv4 and IPv6 address records, respectively; they aren’t General setting keys. | Receiving both records doesn’t mean both network paths work. Test each connection separately. |
To establish a baseline using system DNS, temporarily change the existing dns-server entry to the form shown below. This is a minimal snippet, not a complete config. Don’t use it to overwrite existing rules or other General entries.
[General]
dns-server = system
If system resolution is clearly failing and you’ve confirmed that a specific resolver is reachable on the current network, replace just this entry for comparison. Test each of the two addresses below first; don’t assume that listing two addresses provides reliable automatic failover:
[General]
dns-server = 1.1.1.1, 8.8.8.8
Check rules and IPv6
If the domain resolves but the webpage still won’t load, focus on rules, the outbound, and the IPv6 path. In Config mode, rules are checked from top to bottom, and the first match determines the policy. PROXY, DIRECT, and REJECT mean proxy, direct connection, and reject, respectively. A successful DNS response doesn’t bypass REJECT or automatically send an unreachable DIRECT destination through a proxy.
[Rule]
DOMAIN-SUFFIX,example.com,PROXY
IP-CIDR,192.168.0.0/16,DIRECT,no-resolve
FINAL,PROXY
- DOMAIN-SUFFIX: The example matches example.com and its subdomains. Check whether another rule matches the affected domain first, especially a REJECT rule or one that routes to a different outbound.
- IP-CIDR: The example sends requests directly when the known destination address falls within 192.168.0.0/16.
no-resolvemeans this rule won’t trigger an additional domain lookup for matching; it doesn’t disable DNS for the entire client. - GEOIP: Matches based on IP geolocation, not page content or a domain suffix. Changes in DNS results may affect which rules match.
- FINAL: Provides a fallback when no earlier rule matches and should be placed at the end of the rules. It can’t fix an invalid DNS response or a server-side authentication failure.
If an AAAA record is returned but the current network has incomplete IPv6 routing, some websites may take a long time to load. Temporarily test ipv6 = false in a copy of the config, save it, reconnect, and test the same domain again. If the behavior changes, investigate the network and IPv6 routing. This isn’t a recommendation to disable IPv6 permanently, and one successful page load doesn’t confirm the root cause.
Likewise, incorrect connection settings for Shadowsocks, VMess, VLESS, Trojan, Hysteria2, or WireGuard won’t be fixed by changing DNS. If resolution succeeds but the outbound keeps failing, check the existing service’s address, port, authentication, and transport requirements instead of adding more resolvers.
Common symptoms and fixes
Latency is normal, so why won’t any websites load?
First check whether the basic network works, then confirm which file is actually selected in Config. If multiple unrelated domains all fail to resolve, check whether dns-server is reachable. If they all return addresses but connections fail, check Global Routing, the outbound, and TLS. Don’t assume every failure is a DNS issue.
Only one website won’t load. Should I change DNS?
Look up the domain and compare the result with another resolver. If both return a valid address, check whether DOMAIN-SUFFIX or REJECT matches and whether the destination service is working. A problem with one site isn’t a reason to reset DNS globally.
Why am I still seeing the old result after changing DNS?
The app, operating system, and recursive resolvers may all cache records. Confirm that you saved the active config, reconnect, and fully close the test app before trying again. DNS record TTLs still apply; reconnecting doesn’t necessarily clear every cache.
What if Wi-Fi fails but cellular works?
Keep the same config and check whether Wi-Fi requires sign-in, whether the router filters DNS requests, and whether the specified resolver’s port is reachable. If the problem occurs only with system DNS, compare the DNS servers assigned by that Wi-Fi network instead of changing service authentication details.
Will updating a subscription fix DNS?
Not necessarily. A subscription update usually refreshes service entries; it doesn’t necessarily rewrite the active Config’s General settings. Get subscription links from your service provider. If the update itself times out, first determine whether the problem is the link, DNS resolution, or the connection path.
After troubleshooting, restore any Global Routing and On Demand settings changed temporarily for comparison. Keep only changes that have been verified repeatedly. When contacting your service provider, include the network type, affected domain, DNS response status, and the stage where the connection fails. Hide passwords, keys, and subscription tokens in the config first. A one-time client purchase isn’t a service plan, and downloading the client again doesn’t provide a new connection service.
- Record the dns-server value before and after the change and note the network used. Confirm that the result is reproducible.
- Record whether there was no DNS response, an address was returned but the connection failed, or the connection reached TLS and failed there. Avoid vague reports such as “it doesn’t work.”
- Keep a recoverable copy of the config. If the comparison doesn’t help, undo the change rather than accumulating settings you don’t understand.