First, distinguish Settings, Config and connection status
Settings can come from different places
When looking for an option, first identify whether it controls app behavior, configuration behavior or an individual server. Settings commonly cover preferences such as tests, logs, sync and local services. Config organizes configuration files, while the server editor stores the address, port, authentication and transport details. All can affect a connection, but they are not interchangeable. For example, changing the local Proxy port does not change the remote server port, and changing the test method will not fix incorrect authentication details.
The connection switch on Home indicates whether Shadowrocket is attempting to establish a system network tunnel. It does not prove that every request is reaching the intended server. Global Routing determines whether requests follow the configuration, use Proxy or go Direct; the selected server may be the route used for traffic that needs a proxy. When troubleshooting, note all three states. An enabled switch alone does not confirm that the configuration, DNS resolution and remote connection are working.
How to interpret default, current and recommended values
In this guide, “default” means an initial state that can be verified in the current installation—not a value assumed from a screenshot found online. Imported configurations, restored backups, system updates and network changes can all affect behavior. A blank field may inherit a value from the system or configuration; it does not necessarily mean the option is off. Check the option description and active Config before entering anything. In particular, do not copy unverified numbers into port, timeout or interval fields.
Recommended values also depend on context. For local testing, start by keeping the method that currently works. For domain-based routing, first verify the rules and DNS path. If you do not need to use multiple devices, there is no need to enable sync just to make the settings feel complete. Change an option only when you understand the problem, know what result to expect and can revert it. More settings do not mean a more stable connection; changing several variables at once can hide the real cause.
Establish a baseline you can restore
Before making changes, note the current Config name, Global Routing mode, selected server, network type and original value of each setting you plan to change. Before editing a configuration file, save a separate copy that you can restore, and give it a clear name. Backups may contain authentication details, so keep them somewhere you control. When showing someone an issue, include only the relevant field names and a redacted configuration structure—never the full subscription URL.
Change one variable at a time, then disconnect and reconnect so the new setting applies to a new session. Existing long-lived connections may keep using the old route, and browser or app caches may preserve earlier results. Test with the same target and action under similar network conditions. If the change does not have the expected effect, first confirm that the edited configuration is active before trying something else. Avoid piling on more changes straight away.
| Location | Main purpose | Check first when troubleshooting |
|---|---|---|
| Settings | App preferences and available helper settings | Whether the option applies on this platform |
| Config | Resolution, routing and configuration sections | Whether the intended file is selected |
| Server editor | Remote address, port, authentication and transport | Whether the details match the user's existing service information |
| Home | Connection control, selected server and operating status | Whether the target request occurs after connecting |
What to keep in mind when using different platforms
The same concept may rely on different system capabilities and permissions on different platforms. iCloud and Widget depend on the relevant system features; local ports are affected by the listening address, firewall and port availability. Do not assume another device should have the same layout just because one device has a particular option. When moving settings between platforms, carry over the intended behavior and rules, then check which options are available on the other device rather than following the same screen taps mechanically.
This page explains how to use your own subscription or server details. It does not provide, sell or recommend nodes or subscriptions. A one-time client purchase is not a line-service plan; getting the app and using a remote service are separate things. If you have not completed the initial import, return to the quick-start guide to establish a minimal working connection, then come back to fine-tune settings. Leave unrelated options as they are.
DNS: first identify what resolves the domain
DNS resolution and proxy connections are separate steps
When you visit a domain, the app generally needs a usable destination address before it can establish the next connection. DNS resolves names; the proxy server transports the associated requests. They are related, but they are not the same test. A normal server latency result only shows that a particular probe received a response. A webpage may still fail if name resolution does not work. Conversely, a resolved domain does not guarantee that the remote service, certificate validation or proxy transport will succeed.
When checking DNS, first find out whether the device is using system resolution, a resolver specified in the configuration or the app's own resolution behavior. Wi-Fi and cellular networks may provide different DNS settings, and some apps maintain their own caches. After changing a DNS option in Shadowrocket, results may not change immediately if the target still uses a cached response. Verify separately that the setting was saved and that a new request actually used the new path.
Blank, system and explicit addresses
In a configuration, dns-server = system means that the setting uses the system-provided resolver; it does not guarantee a particular level of network quality. Decide whether to keep system DNS based on the current network and results for the target. If resolution works, leave it as is rather than replacing it just because someone else's configuration lists several addresses. Before specifying a resolver, confirm that the address is reachable, understand its purpose and know where DNS requests will be sent.
A blank field in the interface should not automatically be interpreted as system; it may mean that no value overrides the configuration. Check the option's description for its exact meaning. Do not enter a full encrypted DNS URL into a field that accepts only a plain address. The protocol and input format must be supported by that setting. If the interface does not say a format is supported, do not assume it is. Invalid input may cause resolution problems without triggering an error when you save it.
[General]
dns-server = system
[Rule]
DOMAIN-SUFFIX,example.com,PROXY
FINAL,DIRECT
This example only shows that a configuration section can contain both DNS and rules; it is not a complete configuration recommended for every network. example.com is a documentation domain, and the rule's PROXY action requires a valid proxy selection. The point is to distinguish their roles: the DNS option in [General] affects resolution, while rules in [Rule] determine how requests are handled. One cannot be inferred from the other just because similar terms appear.
What to check when latency looks normal but a webpage will not load
First check whether the problem affects just one domain. Try another target you can normally reach and look in Log for resolution errors. If only one domain is affected, check local mappings, caches and the domain itself. If several domains fail, compare results across networks. Do not change the server, DNS and Global Routing at the same time; even if things start working, you will not know which change helped.
If the domain resolves to an address but the connection then fails, check the server settings, matched rules and destination port. If a certificate error appears only after the connection is established, check the system time, target hostname and relevant TLS settings. Replacing the domain with its IP address is not a universally valid test: the service may depend on the hostname, SNI or certificate matching, so results under changed test conditions may not be comparable.
How IPv6, caches and local mappings can affect results
A domain can return both IPv4 and IPv6 addresses. If only one route has a problem, some apps may work while others take longer to respond. Do not disable IPv6 just because it appears in the results. First confirm that the system network supports the relevant connection and identify the address type in the log, then compare one change at a time. If disabling an option helps, record the result and network conditions rather than treating it as a setting that works for every network.
Local mappings in [Host] can also change resolution results. Check whether the active configuration assigns a fixed address to the target, and confirm that the mapping is still needed. Changing DNS will not necessarily bypass a local mapping, and clearing an app cache does not remove a configuration entry. Before restoring anything, note the original line and temporarily disable only the entry related to the affected target. Avoid deleting other mappings that work.
Global Routing: Config, Proxy and Direct
Choose the routing mode before interpreting individual rules
The three common Global Routing modes are Config, Proxy and Direct. Config processes requests using the active configuration, for rule-based routing. Proxy routes requests handled by the app through the proxy mode; Direct provides a direct-connection comparison. These are overall routing choices—not three servers or three subscription formats.
To check whether a rule is working, first confirm that Config is selected and that the active file contains the rule. In Proxy or Direct mode, the final route alone cannot show that a configuration rule matched. A common misread is to edit a rule, leave Proxy selected and then assume that all targets going through the proxy means the rule order is wrong. Check the overall mode before changing rule content.
| Mode | Main purpose | What it does not prove |
|---|---|---|
| Config | Route requests according to the active configuration | That every request uses the proxy |
| Proxy | Compare the proxy route | That a particular configuration rule matched |
| Direct | Compare a direct connection | That the remote proxy server is working |
Rules are checked in order; put specific conditions first
DOMAIN matches a full domain, DOMAIN-SUFFIX matches a domain suffix, and DOMAIN-KEYWORD matches a keyword in a domain. Their scopes differ; do not swap them just to reduce the number of lines. For a single domain, prefer a precise condition. A suffix rule may also match subdomains, while a keyword rule has a broader scope. Consider the risk of matching unrelated domains before adding one.
GEOIP and IP-CIDR match address-related conditions; IP-CIDR6 matches IPv6 ranges. FINAL handles requests not matched by earlier conditions, so put it at the end. When domain and address rules are mixed, both resolution results and rule order affect what you observe. The presence of a domain in the list alone does not prove that its request will be handled as expected.
[Rule]
DOMAIN,example.com,DIRECT
DOMAIN-SUFFIX,example.com,PROXY
GEOIP,CN,DIRECT
FINAL,PROXY
In this example, the full domain example.com matches DIRECT first. Its subdomains can then match the suffix rule and use PROXY. GEOIP and FINAL handle requests that did not match earlier rules. This example illustrates rule order; it is not intended to replace an existing configuration. PROXY, DIRECT and REJECT are handling actions. REJECT blocks the relevant request and is not a general connection fix.
What the sections in a Config file do
[General] organizes global behavior, [Rule] organizes routing conditions, [Host] stores local name mappings, and [URL Rewrite] handles request rewriting. A rule written in the wrong section will not be categorized automatically just because it looks like a rule. Keep section headings when editing and check that neighboring lines belong to the same type. Avoid copying only a few middle lines and leaving out the configuration context they depend on.
A subscription is mainly for importing and updating your own server details. A configuration file may also contain rules and behavior settings, so the two are not interchangeable. Successfully updating the server list does not mean the active Config was replaced; changing Config does not necessarily change the selected server. If traffic behaves differently after an update, check each object separately instead of repeatedly using Subscribe in an attempt to fix a rules issue.
How to verify changes and when to revert
Make a new request to a specific target and check whether its actual domain, destination address, handling action and selected server match your expectations. Keep the test focused: a webpage may request resources from several domains, so one request going Direct does not mean every connection for the page does. Identify the resource that actually failed before changing a rule. Avoid masking other issues with an overly broad rule.
After changing rules, keep a record of the difference: which line changed, what request behavior you expected and what condition actually matched. Do not reorder the entire rule set without evidence. For syntax details, see rule keywords and matching scope. For an overview of sections, see configuration file structure. To revert, switch back to the saved configuration and reconnect to test it.
On Demand: when the system triggers a connection
On Demand versus the manual connection switch
On Demand lets the system trigger a VPN connection when specified conditions are met. It answers “when should a connection be attempted?”, not “which domains should use the proxy once connected?” For the latter, check Global Routing and Config. Confusing the two can lead to expecting a Wi-Fi condition to change the domain rules for that network. These are separate configuration responsibilities.
Before enabling On Demand, confirm that a manual connection works. If the server address, authentication or DNS is wrong, On Demand will only repeat the same failure when its conditions are met; it will not fill in missing details. First connect manually on a stable network, verify access to a target and check the logs. Then add a trigger condition. This makes it easier to tell whether a later problem is caused by automatic triggering or by a connection that was never working in the first place.
Choosing Wi-Fi and cellular conditions
Before setting network conditions, define what you need: connect on a particular known Wi-Fi network, connect after leaving it, or handle cellular data differently. Avoid stacking conditions that overlap or override one another. If the interface lets you match a Wi-Fi name, check the actual network name. Identical names do not guarantee the same physical network; public networks may share a name, so the name should not be treated as authentication.
Switching between cellular and Wi-Fi can briefly interrupt the connection, change addresses and require sessions to be re-established. To check On Demand behavior, wait for the system to finish switching networks, then make a new request instead of watching the icon during the transition. Existing long-lived connections may drop, and background apps may not retry immediately. Test the trigger with a repeatable foreground action and note the network type before and after switching.
Why a connection may come back after you turn it off
If the system tries to reconnect when a new request occurs after you disconnect manually, first check whether the On Demand conditions are still met. Temporarily disable the relevant trigger before testing the manual switch. Do not immediately delete every VPN configuration. Device management requirements or other network settings may also be present. Confirm what each item is for before removing it, and follow your organization's network policy on managed devices.
When investigating a connection that keeps coming back, note three times: when you disconnect, when the network changes and when an app makes another request. If reconnection follows the same condition each time, check that condition. If there is no consistent pattern, check the logs and system status. An icon disappearing and reappearing alone cannot tell you whether On Demand, a change in the active configuration or delayed status updates caused it. Reproduce the behavior under controlled conditions.
Default and recommended behavior: keep it understandable before adding automation
This guide does not assume a fixed initial On Demand switch state. Initial setup, restored data and system VPN settings can all affect the current state, so check the device's actual interface. Start with a manual connection that works, add only one necessary condition and test it before expanding the setup. If you are not familiar with how conditions interact, being able to turn the connection on and off manually is a better troubleshooting baseline than a set of automatic rules you cannot explain.
Test cases where the connection should not happen as well as those where it should. Confirming that a trigger works in one situation is not enough to prove the setup is correct. Also connect to a network where it should not trigger, disconnect manually and check whether a new request reconnects. If behavior is unexpected, simplify the setup to one condition. Do not try to control connection triggers by adding a reject action at the end of the rules; that combines two separate concerns.
Sleep, background activity and battery usage
Background scheduling by the system, network request frequency and server reconnection behavior can all affect connection activity. If battery use seems higher after enabling On Demand, compare actual requests and reconnection logs over the same usage period rather than inferring the cause from the switch alone. Repeated failures and retries deserve attention: they may indicate invalid settings or an unreachable network. Fix the cause before trying to reduce unnecessary retries.
When resuming use, first verify foreground access, then test the usual lock, wake and network-switch sequence. The system does not guarantee that every app will resume its session immediately on wake, so distinguish a restored tunnel from an app that has retried its request. If manual connections always work but a particular trigger sequence fails, note the network conditions, triggering action and log summary, then continue troubleshooting with Troubleshooting.
Ping and Connectivity Test: understand what each test checks
A latency reading does not mean the full service connection succeeded
Ping, connection probes and full webpage access may test different stages. One test may only check whether a host or port is reachable; another may complete a handshake or request a specific target. Follow the description of the current test method in the interface. Interpret its timing in that context. Do not rank results from different methods and use them to judge the experience of every app.
Traditional Ping is commonly associated with ICMP, but a server that does not respond to ICMP may still have a reachable proxy port. Likewise, a reachable port does not prove that protocol authentication is correct. If you can choose a test method, first decide whether you want to check a host, a port or a request through the proxy, then compare results using the same method. A timeout does not always mean the server is out of service.
Choosing a test target, timeout and concurrency
If a test requires a target, choose one with a clear purpose that you can reach repeatedly and that matches the expected input format. Testing a full URL is not necessarily the same as entering a hostname; do not treat them as interchangeable based on guesswork. A target service may limit requests, change its response or become temporarily unavailable, so one failed test is only a clue. Cross-check it with an independent test before blaming every server.
A timeout that is too short may mark a slow but successful request as a failure; one that is too long delays detection of an unreachable target. There is no fixed value that works for every network. Start with the current value if it works, repeat the same test on the same network, and adjust based on normal completion time and how long you can reasonably wait. If results vary widely, investigate network changes rather than continually increasing the timeout to hide the problem.
Batch tests are also affected by concurrency, device load and competition for network resources. Testing many entries at once can produce different results from testing one at a time. If you only need to check whether the current server works, test it individually and then try an actual request. If batch results fail but an individual test works, reduce test interference and try again before changing each server's protocol settings.
| Observed result | What it may indicate | Next step |
|---|---|---|
| Host probe failed | The probe did not receive the expected response | Check the port and test the actual protocol connection |
| Port reachable, authentication failed | A listening service was reached | Check the protocol, authentication and transport settings |
| Test passed, webpage failed | The test path does not cover every path used by the webpage | Check DNS, rules and the target request |
| Failure only after switching networks | The session or network conditions changed | Reconnect and check the trigger state |
Check protocol settings field by field; do not infer them from the protocol name
Your server details may use Shadowsocks, VMess, VLESS, Trojan, Hysteria2, WireGuard, HTTP or SOCKS5. Each protocol requires different fields; the same address and port do not make protocols interchangeable. First verify the protocol, then check the authentication details it actually requires, followed by transport, TLS and hostname settings. For missing details, refer to your own service information.
For connections that use UDP or QUIC, network support for that transport can affect the result. A successful TCP port test does not prove that the UDP path works. If only a particular protocol fails on a specific network, compare the same configuration on another network and check the logs for handshake or timeout details. For background on transport, see Hysteria2 and QUIC explained; this does not mean either will be faster on every network.
System time and repeatable tests
A significant system time discrepancy can affect certificate validity checks or authentication that depends on time synchronization. Check that the system clock and time zone match reality, and confirm that automatic time setting works. Do not respond to a certificate error by turning off verification. Check the target hostname, TLS settings and system time to identify the cause. A connection that works only when verification is skipped does not prove the original configuration was correct.
Keep a short test record with the network type, test method, target, result and any changed settings—never authentication secrets. Run a small number of repeat tests against the same server, then verify with the app you actually need. If you still get a timeout, follow the server timeout troubleshooting steps. Do not choose a server based solely on the lowest latency or treat one failure as permanent.
Log and Data: identify which stage is failing
Reproduce the issue once, then check the relevant log
Logs help break a request into stages you can inspect; collecting as much text as possible is not the goal. Note the time, pause unrelated activity and reproduce the problem once with a specific target before checking the corresponding record. With several apps open, background sync, page resources and automatic retries can clutter the log and make unrelated errors look relevant. Narrowing the test is usually more useful than collecting more data.
Fields shown may vary by platform and log level; refer to the current interface. Use a consistent reading order: Is there a target request? What is its name or address? Which route handled it? Did it fail during resolution, connection, authentication or transport? Classify the stage first, then look at the details. Do not search for a familiar English word without checking its context.
No request in the log? First check whether traffic entered
If you cannot find a related request after reproducing the issue, first check whether logging is enabled, whether its display range includes the action and whether the timestamp is correct. Then check the system connection status and active network interface. Some requests may be satisfied from cache or use a route outside the current view. Repeatedly changing remote authentication details is unlikely to help before there is evidence the request reached that stage.
A webpage is made up of multiple requests, not just one. If the page loads but images do not, check the domain for a failed image request rather than only the main domain in the address bar. An app's login may also use an API separate from its home page. Record only the steps needed to reproduce one action—for example, open a page and tap Login once—then use the timestamp to find the corresponding connections and narrow down the target.
Distinguishing resolution, handshake and remote closure errors
A resolution failure means the expected name lookup result was not obtained; check DNS and local mappings first. A connection timeout may mean that the target is unreachable, the port is blocked or the transport is restricted; interpret it in light of the test method. For authentication failures, check the protocol and your existing credentials. A connection closed by the remote side may involve the target service or a protocol mismatch. The word “closed” alone does not identify which side is responsible.
A log error may show an intermediate failure that a later retry recovered from. Check whether the user's action ultimately succeeded, whether failures keep repeating and whether new requests stop at the same stage. Handle a temporary error differently from a consistent, reproducible failure. If the connection works, do not change working settings just to remove every log entry. Logs are not a scorecard that has to be free of messages.
Data statistics are not the same as remote service billing
Interpret Data statistics according to their displayed scope, which may depend on the current session, device records or interface filters. They can help show whether data is moving, but do not automatically replace usage records from a remote service. Client and remote statistics may use different time ranges, count protocol overhead differently or measure at different points. A difference between the numbers is not automatically an error. To compare them, align the start time and observation period first.
Before resetting statistics, check whether you need the current data for troubleshooting. Resetting Data does not fix DNS or clear server state; clearing logs does not undo configuration. Treat “clear the display,” “reconnect” and “restore settings” as separate actions, and verify the symptom after each one. This helps avoid assuming a network issue is permanently fixed just because something was cleared.
Log detail and retention
More detailed logs can increase storage and processing costs and expose more private information. Increase the detail only when you need to reproduce an issue, then restore the original setting. If the log is large, shorten the reproduction steps and filter by target and time instead of leaving the most detailed level enabled. Check the current interface to see whether the version offers log levels, retention options or a clear action. Do not guess at configuration keys that may not exist.
When asking for help, summarize the platform, network type, Global Routing mode, action that caused the problem, failure stage and single-variable tests already performed. A relevant redacted excerpt is usually more useful than a full-screen capture. If you cannot yet classify the issue, look for a similar symptom in Troubleshooting. For configuration sections, return to the rules section of this guide and check which object is active.
iCloud sync, backups and Widget
First distinguish sync, export and remote updates
iCloud sync relies on the relevant Apple system, account and permissions; it is not a universal sync method available on every platform. Sync keeps supported data in the appropriate state across devices. An export backup preserves a copy from a particular point in time. Subscribe updates information from a subscription source you already use. These actions may affect different data; they are not interchangeable just because each can bring data back.
Before enabling sync, keep a restorable copy on the current device and check what data the other device is using. If the two sides already differ, decide which copy should be the baseline before syncing. Do not make major edits on both devices and repeatedly toggle sync, expecting the system to determine which changes are correct. Check the interface prompts and actual results to understand how conflicts, deletions and replacements are handled.
Do not guess what sync includes
Check the current sync description to see whether it covers servers, configurations, selection state or specific preferences, then verify the scope with a small, recognizable test. Do not assume that seeing a configuration name means credentials, logs and local ports are all synced. Test first with a name change that contains no secrets, then arrange the full migration once you know what is included. Check device-local settings on each device individually.
If synced data does not appear right away, confirm both devices use the intended account, the system allows the app to use iCloud, the network is available and cloud storage is in good standing. Give both apps time to sync, then check the result. Repeatedly signing out or deleting the app adds variables and may affect local copies, so it is not a good first step. If the system does not show a completion notice, waiting alone does not prove that all data has synced.
Saving and restoring backup files
Backups may contain information that can access remote services, so handle them as sensitive files. Store them somewhere you control and label each file with the device, Config and purpose it belongs to. Do not put authentication details in a filename. When transferring a file to another device you own, confirm the recipient device and storage location. Remove temporary copies afterward if appropriate, rather than leaving backups scattered in multiple locations indefinitely.
Before restoring, save the current state. Import the backup, then check the server list, active Config, Global Routing mode and connection permissions individually. A successful restore notice confirms that the import completed; it does not mean the remote service details are still valid. Manually test one known-good entry before restoring On Demand and other automatic behavior. This prevents old settings from being hidden by automatic retries.
Widget is a shortcut, not a full diagnostic interface
Available Widget options depend on the features provided by the system and app. A Widget can give quick access to supported information or actions, but updates may depend on the system's refresh schedule. If it does not reflect a recent connection change, open Shadowrocket and check Home and the actual request. Do not rely on desktop Widget text alone to determine whether the tunnel is working, and avoid repeated taps that trigger conflicting actions.
After adding a Widget, compare it with a manual on/off operation and check what information it displays and what tapping it does. Interaction may differ between locked and unlocked states; follow system prompts to grant access. Confirm connection permissions in the app or system as needed. A Widget cannot bypass system permissions. If the expected Widget does not appear, first check whether the system supports that form.
After moving to a new device, verify each item instead of just checking the list
A list that looks the same on a new device does not mean the runtime environment is the same. Network conditions, system VPN permissions, local network access and battery settings can all affect connection behavior. Check in this order: manual connection, DNS, rule matching, test method, local ports, then On Demand and Widget. Do the same when moving between devices that support iCloud; data migration does not replace functional testing.
If an issue appears on both devices after syncing, stop editing and save the current state of each one. Then identify which item changed most recently. When restoring one device, consider whether sync could propagate the restored state again. Keep settings local if you do not need them across devices. For data that does need syncing, maintain separate backups so convenience and recovery are handled independently.
Proxy ports, local access and recovery order
A local listening port is not the same as a server port
The port in the server editor is a remote service setting and must match the server details you already have. A local Proxy port is a listening address on the current device, for apps that support manual proxy settings. Changing the local port to match the remote one does not link them automatically, and changing the local listener will not fix remote authentication errors. Before editing a port, confirm which type you are changing.
HTTP and SOCKS5 are different proxy connection methods. Apps using them must select the same type as the listening service. Entering an address and port with the wrong type may cause the connection to close immediately or requests to fail. Check the app's own documentation to see whether it supports the proxy type. An app without manual proxy settings will not automatically use a local port just because one is available on the system.
The listening address determines who can connect
A loopback address is for access from the same device only, such as the IPv4 address 127.0.0.1. It is not the address of another device on the local network. If another device uses its own loopback address, the request stays on that device instead of reaching the one running Shadowrocket. For local apps, prefer a listener restricted to the device and keep local network sharing disabled unless you need it.
If you need local network access, first confirm that the current platform supports the required listening capability. Then check the listening address, device's local network address, firewall and network isolation settings. Listening on a broader interface expands who can connect; being on the same Wi-Fi is not access control. Do not expose the service to a public network for a temporary test. Restore the original listening scope afterward and check which apps still use the port.
Choosing a port and checking whether it is in use
This guide does not specify a fixed default port for every platform. Start by checking the port that is actually listening, then enter that same value in the app that uses it. If you need to change it, choose a valid, unused port and check system permission requirements. Saving a port does not prove that a listener has started. Check the service status or logs, then test it with a specific request from the same device.
If a port is already in use, identify what is using it instead of terminating an unknown system process. Disable a duplicate listener you do not need or assign another port to one service, then update every app that references it. If you change the port in Shadowrocket but leave the browser's manual proxy setting unchanged, the browser will keep using the old port. It may look like Shadowrocket cannot connect when the caller is actually still pointing to the old value.
| Symptom | Check first | Avoid this misstep |
|---|---|---|
| An app on this device cannot connect | Listener status, address, port and proxy type | Rewriting remote authentication details |
| Another device cannot connect | Platform support, interface scope, firewall and network isolation | Turning off all system security controls |
| The browser stops working after disconnecting | Whether the app making the request still has manual proxy settings | Repeatedly re-importing the subscription |
| Only some targets are reachable | The active rules, DNS and failed request | Changing the local listening port at random |
Start recovery with the smallest change
If the problem started after a specific change, revert that one setting and reconnect to test it. Do not immediately reset everything. If you made several recent changes, temporarily disable the relevant automatic triggers, switch back to a saved Config, check the selected server, then review DNS. Only then check the local listener and settings in apps that use it. Restore one layer at a time according to its dependencies; do not combine clearing caches, deleting data and reinstalling into a single step.
Before resetting, confirm that a usable backup is saved, that you can still retrieve your own subscription or server details, and that you have recorded system-level manual proxy and VPN settings. Reinstalling is not guaranteed to clear all system network state or restore invalid remote details. If you need to get the app again, choose your platform on the downloads page and follow its installation guidance. Do not guess the installation package path here.
Post-recovery verification order
First confirm that a specific target loads. Then check whether rules under Config match as expected, compare Wi-Fi and cellular, and only then restore On Demand, sync and Widget. Keep a result after each step and return to the last working state if something fails. This verifies basic connectivity separately from automation and avoids losing diagnostic clues by restoring everything at once.
If the problem persists, gather evidence in this order: resolution, connection, authentication, rules and the calling app or entry point. Use the necessary redacted fields to consult Troubleshooting. The goal is not to fill in every field, but to make every change purposeful, repeatable and reversible. Leave other settings at their verified values and consult the relevant section when a new need arises.