Official Guide · Troubleshooting

Shadowrocket Troubleshooting Guide

Start with the symptoms. First distinguish system connectivity, server connectivity, and request routing, then adjust the relevant settings. Change one thing at a time and keep test results you can reproduce.

For first-time setup, read the setup guide to import a subscription, select Global Routing, and make your first connection. Use this page to troubleshoot issues that arise after setup. Once resolved, return to the verification steps in the setup guide to confirm the result.

  • Have your own subscription or server details ready
  • Save a copy of your configuration before making changes
  • Change only one test condition at a time

Troubleshooting guide contents

Connection switch won’t turn on: Check the system VPN configuration first

The switch immediately turning off after you tap it on Home is different from a switch that stays on while webpages fail to load. For the first case, check whether the system accepted and started the VPN configuration. For the second, troubleshoot the server and routing. Note whether the switch turns off with a prompt, whether a system authorization dialog appears, and whether the connection remains after you leave the app. Don’t keep switching servers before confirming the system connection state. Check connection logs for server address errors; the switch color alone isn’t enough to diagnose them.

First-time authorization and other VPN configurations

The first time you connect, the system may ask to add a VPN configuration. Follow the prompts and authenticate with your device passcode or other required method. If you previously canceled authorization, return to Shadowrocket and try connecting again to see whether the prompt reappears. Then open system Settings and go to General > VPN & Device Management to check whether a Shadowrocket VPN configuration is listed. The menu location and wording may vary by system version; use the options shown on your device. Don’t confuse network settings with device management profiles.

If the device has other VPN configurations, disconnect any active VPN connections before testing Shadowrocket. Some organization management, content filtering, or security settings may restrict personal VPNs. If the device is managed, ask the administrator to confirm the policy; don’t remove organization configurations yourself. On a personal device, avoid jumping straight to “Reset Network Settings.” It affects saved Wi-Fi networks and other information, and does not replace checking the authorization prompt. First identify which connection is active; this usually preserves more useful diagnostic clues than deleting all network configurations at once.

Switch turns off, reconnects automatically, or needs a rebuilt configuration

If the switch turns off and then turns itself back on, check whether the conditions for On Demand are still met. On-demand connections can restart the tunnel when a network request occurs; this doesn’t mean the switch is broken. Note the original conditions, temporarily turn off On Demand, and control the connection manually while troubleshooting. If the system says it’s connecting but the app hasn’t refreshed, wait for the status to settle before checking again. Rapid repeated taps can overlap start and stop requests, making it hard to distinguish the original issue from a state change caused by your actions.

If authorization is complete but the connection still drops immediately, quit and reopen the app, then restart the device and test again. Only consider removing a failed Shadowrocket VPN configuration from system settings and authorizing it again after backing up your configuration and confirming that you can retrieve your subscription. Removing the system VPN configuration is not the same as deleting subscriptions, servers, or Config files from the app. Check exactly what you’re removing. Don’t clear app data just to rebuild the tunnel; that can turn a system startup issue into a configuration recovery problem.

Distinguish startup failures from server errors

Check the logs from the time of the error and distinguish configuration read failures, system tunnel startup failures, and subsequent failures connecting to a remote server. Record the exact error and sequence of events. A server domain appearing in the log doesn’t prove the server is at fault. If the system VPN cannot be established on different networks on the same device, but the same subscription works on another device you own, focus on local authorization, management policies, and configuration state. If the tunnel starts and a timeout or TLS error occurs afterward, continue to the server timeout section rather than repeatedly troubleshooting the wrong layer.

Keep one known condition fixed for every retest: the same Wi-Fi network, Config, and server. Connect manually, then open the same test page. Keep results from before and after each change so you can tell whether restarting, reauthorizing, or turning off On Demand made a difference. If you need help, summarize whether first-time authorization was completed, whether the system shows a connection, the last error before disconnection, and whether the device is managed. You don’t need to send the full configuration file.

Connected but no internet: Troubleshoot routing and rules layer by layer

A VPN status of Connected means the tunnel is established; it doesn’t confirm that the selected server works or that every request is routed through a proxy. First turn off Shadowrocket and visit a page that normally opens directly to confirm that Wi-Fi or cellular data is working. If public Wi-Fi requires a sign-in, complete it first. Keep the tunnel off while retrying if the sign-in page won’t load. Until the underlying network works, changing rules, DNS, or protocol settings usually adds variables without solving the problem.

Run a comparison test with Global Routing

Once the underlying network is working, keep the same server and test URL, then compare the three Global Routing options: Config, Proxy, and Direct. Direct tests the current direct connection; Proxy tests the proxy connection; Config applies the rules in the active configuration. This is a temporary diagnostic test, not a recommendation to use one option permanently. Restore your usual routing behavior when you’re done.

Test resultCheck firstNext step
Direct works; Proxy failsServer, protocol, and proxy-side DNSTest a fixed server and check for handshake errors
Proxy works; Config failsMatching rules, policy names, and fallbackCheck which rule the request actually matches
Still fails with the tunnel offUnderlying network or destination serviceRestore regular network access first
Only one app failsApp requests, cache, and reused connectionsReconnect the affected app

Use the same destination for all three tests, and make a new request after each switch. Existing long-lived connections may continue using the previous route, and browser cache can make a page appear to “still work.” Close and reopen the relevant page. If needed, fully quit the test app and check the connection log again. Don’t change the routing mode and server at the same time; even if the issue goes away, you won’t know which change fixed it.

Check the active file, rule order, and FINAL

In Config, confirm which file is currently active, not just whether a file appears in the list. Rules are usually matched from top to bottom, so a broad rule near the top can shadow a more specific one below it. The following is a syntax example only: example.com is a documentation domain, not a service for testing proxy performance. PROXY must refer to an actually available proxy policy. Don’t overwrite your only everyday configuration with this snippet; make a copy first.

[Rule]
DOMAIN-SUFFIX,example.com,PROXY
GEOIP,CN,DIRECT
FINAL,DIRECT

This snippet sends domain suffix matches to PROXY, then processes the geographic matching rule, with everything else handled by FINAL,DIRECT. If you want unmatched connections to use a proxy too, the fallback should reflect that goal, for example FINAL,PROXY. These options serve different purposes; only one can be the final effective rule. GEOIP depends on the destination address and related data; it doesn’t determine location from a webpage’s language or site name. For a complete example of routing selected websites, see DOMAIN-SUFFIX and FINAL explained.

Check actual requests when only part of a page fails

If a webpage’s main domain loads but images, sign-in, or video fail, check the resource domains rather than only the address bar. A page may request several domains, each routed by a different policy. Use connection logs to find the failed request, then check its rule and destination policy. Don’t send every related domain to REJECT or PROXY just because one resource failed. Verify the specific request first to keep your rules clear and predictable.

If Proxy and Direct both produce inconsistent results, temporarily use a minimal test configuration, disable processing that isn’t needed for this test, and keep the original file. The goal is to confirm, in order, that the underlying network works, the proxy path works, and the correct rule is matched—not to add complexity and hope for an occasional fix. Once the minimal configuration works, restore the original rules one at a time and retest the same request after each change. The first step that brings the problem back is where to focus next.

Server timeouts and handshake failures: Identify where the connection stops

A timeout means an operation didn’t finish within the allowed time. It doesn’t by itself mean the password is wrong or that the entire server is offline. First identify whether the failure occurs during DNS resolution, transport connection, TLS handshake, or proxy protocol authentication. Choose an existing server, record the full error and time, then run Connectivity Test and try loading an actual webpage. These tests may use different destinations and methods, so a passing test only confirms that the specific check succeeded; it doesn’t guarantee every request will work.

Check the address, port, and protocol type one by one

On the server edit screen, check Address and Port. Make sure the address doesn’t include spaces, non-English punctuation, a webpage title, or a full subscription URL. Enter a hostname or IP address as required by the selected type; don’t add a protocol prefix without a reason. The port must match the server’s listening port; different ports on the same server may run entirely different services. Choose Type based on your service provider’s information or your own server configuration—don’t guess the protocol from the port number.

Failure stageCommon checksHow to verify
DNS resolutionAddress spelling, DNS reachabilityConfirm the server hostname resolves
Transport connectionPort, listening service, firewallCompare different underlying networks
TLS handshakeSNI, certificate, system timeCheck the server’s TLS settings
Protocol authenticationPassword, user ID, encryption methodCompare each field with current, valid details

Shadowsocks requires a matching encryption method and password. For Trojan, check the password and TLS-related settings. For VMess, VLESS, and other types, use the server details to verify the user ID and transport settings. If you use a transport such as WebSocket, Path, Host, and the server-side routing must also match. Just because the interface offers a field doesn’t mean it needs to be filled in. Don’t copy missing values from another server. The same protocol can use different transport combinations.

TLS, SNI, and system time

For TLS errors, first check that the device time is set automatically, then verify the relationship between the server’s required SNI and its certificate. The connection address and SNI can differ, but their values depend on the actual deployment; don’t enter a familiar domain at random. Check the server configuration for expired certificates, certificate chain issues, or name mismatches. Don’t treat Allow Insecure as a general fix. It changes certificate verification and may hide the actual name or certificate problem.

If an old connection still works but a new one fails, the old connection may not have performed a new handshake; that doesn’t rule out a certificate change. Disconnect and reconnect once, then compare the logs to avoid cached state affecting the result. For an explanation of Trojan fields, see Address, Port, Password, and SNI explained. If you manage the server, check its logs for the same time period to see whether the request arrived and where it was rejected.

Test on another network; don’t treat a timeout as the only clue

Keep the server and settings unchanged, then test on both Wi-Fi and cellular data. If only one network fails, check its reachability, sign-in requirements, transport restrictions, and IPv6 support. If both fail, check the server listener, account status, and whether the server details have changed. If other existing servers in the same subscription work, that only shows the client can connect; it doesn’t prove every setting on the failing server is correct.

UDP-based and TCP-based protocols can behave differently on the same network. Success with one protocol doesn’t imply another should also work, and you shouldn’t repeatedly change MTU without evidence. When reporting the issue, share the protocol type, failure stage, results on both networks, and the names of the fields you checked. Don’t share passwords or full user IDs. Once the connection completes reliably, return to Config and test your traffic; avoid troubleshooting handshakes and routing in the same test cycle.

Subscription add or update failed: Check retrieval and parsing

A subscription update involves downloading the content and parsing it. A request that can’t complete, a server error page, and a successful download with an unrecognized format each require different troubleshooting. First make sure you’re updating a Subscribe entry, not entering the subscription link as a single server address in Add Server. Get subscription links from your service provider. Treat links as passwords: don’t post them in public discussions, image captions, or unrelated online testing tools.

Keep the existing entry, then check the link

If the old entries still connect, don’t delete the entire subscription group first. Note the subscription label, update status, and exact error. Back up your current configuration, then try updating the existing subscription once. When copying the link, check whether a chat app truncated its parameters, added quotation marks or explanatory text, or included a line break. Query parameters containing question marks and separators are usually part of the full URL; don’t remove them to shorten the link. If the credentials were reset, an old-looking link may no longer work.

If your provider offers several import formats, choose content intended for Shadowrocket and make sure you have a subscription URL, not an account webpage. Successfully opening a login page doesn’t mean that page contains subscription data. If the response is a login form, verification page, or regular HTML, the client may report a parsing error or fail to add servers. Check the correct endpoint and permissions rather than re-encoding and importing the webpage content.

Distinguish network, permission, and format errors

If the error says the domain couldn’t be resolved, see the DNS section. For a timeout, first determine how the subscription URL is being accessed. If a rule sends the subscription domain through a proxy that no longer works, you can end up in a loop: the failed proxy prevents updates, and the failed update prevents you from replacing the proxy. If the underlying network can reach the URL, temporarily use Direct for the update or choose another existing working connection. Don’t bypass an HTTPS error by turning off certificate checks.

HTTP status codes can point you in the right direction, but they don’t replace the server’s explanation. A 401 or 403 is often related to authentication, permissions, or access policies; a 404 may mean the path changed; and a 429 usually means you should stop retrying and wait for the allowed interval. The server may also wrap an error in a normal response, so check the response body type too. Repeatedly tapping Update won’t fix invalid credentials and may trigger more request limits. Save the result of one attempt, then check which link is currently valid in your account.

Update succeeded, but expected servers are missing

After an update completes, check the actual entries in the subscription group rather than only the total server count. The server may have removed old entries, renamed them, or returned the same list. Filters may also hide some entries. Manually added servers and subscription servers may belong to different groups. Check their source before making changes so you don’t mistake a manual entry for an update leftover. If names are duplicated, distinguish entries by protocol, address, and subscription source—not by label alone.

Updating a subscription is different from updating a Config file. A successful server list update doesn’t mean the active routing file, rule references, or policy selection now match your expectations. After updating, confirm that the selected server is still available. For multiple subscriptions, see Groups, sorting, and removing duplicate servers. Organize your sources before making bulk changes. Update one group at a time to identify which source returned unexpected content.

When reporting the issue to your service provider, include the time of failure, error type, whether the domain resolves, the access route used, and the general circumstances of the last successful update. If you need to share the full link, use a private channel agreed upon by both parties. Remove URL query parameters and user IDs from public logs. After updating, test one server before cleaning up old entries so you can restore the previous configuration if the new list has a problem.

Slow connection: Measure latency, throughput, and page load separately

“Slow” can mean a long wait before a response, a low sustained transfer rate, or a page where some resources never finish loading. Connectivity Test results and file download throughput are different measurements. A server with a fast connection test may still download slowly because of egress load, destination limits, or packet loss. First describe whether the delay is at the start or continues throughout the transfer. Then decide whether to check DNS, the server path, or the app itself. Don’t repeatedly switch servers based on a single test score.

Set up a repeatable comparison

Keep the device, network, server, and test content fixed. Pause large file syncs and system downloads. First check the regular connection with the tunnel off, then test the same destination under Proxy. If the destination doesn’t normally work directly, choose an appropriate test target that lets you check the underlying network; don’t treat a failed direct connection as a speed result. Change only the server or the network in each round. Look at the trend across several tests rather than judging by a single peak.

Wi-Fi signal bars don’t tell you everything about link quality. If performance improves near the access point, other devices are slow at the same time, or restarting the router helps, check the local network first. On cellular data, consider your location, signal changes, data allowance, and Low Data Mode. Speed tests use data and may compete with video playback for bandwidth. Don’t run multiple speed tests at once; the results would reflect competing traffic, not the capacity of a single connection.

Distinguish initial delay from sustained transfer

If the first page load is slow but refreshes quickly, DNS, TLS setup, or connection reuse may be involved. If downloads remain slow throughout, check packet loss, server load, and destination limits. If downloads start quickly and slow as the device heats up, also check temperature and background activity. Keep the same server and compare two different destinations: if only one is slow, focus on its route or server-side limits. If all are slow, the shared transport path is a more likely place to investigate.

A webpage can load scripts, images, APIs, and media through many separate requests. Receiving the main document first doesn’t mean the page has finished loading. Check connection logs for the request that took the longest and see which policy handled it. If a resource domain was routed through a failed path, the page may appear to “load but feel slow.” This is more likely a rules or DNS issue; switching to a server with higher throughput may not help. Fix the failed resource route before evaluating overall page speed.

Protocol, UDP, and MTU considerations

Different protocols, transport layers, and encryption methods suit different network conditions, but that doesn’t mean one protocol is always faster. UDP-based connections may be restricted on some networks, while TCP-based connections can be affected by packet loss and retransmissions. Compare settings supported by your server. Don’t change the client’s Type without changing the server configuration; a protocol mismatch is a connection failure, not a performance difference.

Consider MTU, path fragmentation, and intermediate network equipment only when small pages work but large responses or some connections stall. First rule out destination issues, inconsistent rules, and ordinary packet loss. Then make changes based on evidence from the server or network diagnostics. No single MTU value works for every connection. Blindly lowering it may add overhead; raising it may cause new stalls. Record the original value before changing it, use the same large-response destination for testing, and keep the change only if the improvement is stable.

Choose connections based on stability and how you use them; there’s no need to test the entire server list frequently. Batch testing creates extra connections and can interfere with active traffic, especially on mobile data. Once you’ve identified the issue, keep an existing connection that suits the current network, restore your everyday Config, and check that requests are routed as expected. If only background apps are slow while foreground browsing works, also check background scheduling and network changes in the battery section before attributing the issue to server performance.

DNS resolution issues: Distinguish server domains from destination domains

DNS issues can occur at two different points: the client may need to resolve the server’s domain first, and business requests may then need to resolve the destination website’s domain. If the first lookup fails, the proxy connection can’t be established. If the second fails, the VPN may show as connected while specific websites remain inaccessible. Check which name failed in the log; don’t assume every domain error comes from the same DNS setting. A successful lookup only means an address was returned, not that the service at that address is reachable.

Find out which path DNS requests use

Check DNS-related settings in the active Config. Distinguish between system-provided resolvers, manually specified resolvers, and DNS services accessed over a particular route. Specifying a DNS address doesn’t mean the current network can reach it. Encrypted DNS services also need a connection and may require an initial lookup of their own domain. If the DNS service depends on a proxy that hasn’t connected yet, check for a startup dependency loop rather than adding more resolvers.

The log messages “name does not exist” and “request timed out” point to different checks. The first may indicate a misspelled domain, a genuinely missing record, or a server response issue. The second may mean the resolver is unreachable or the request path is failing. Record the queried name and response type, then compare results on different underlying networks. A resolver that works at home may not be reachable on public Wi-Fi or cellular data.

The effects of cache, IPv4, and IPv6

After changing DNS, the app, system, or existing connections may still use a cached result. Disconnect and reconnect the test connection, then close and reopen the test page. If needed, reopen the target app. Don’t decide that a setting didn’t work based on one refresh in the address bar. If a name has both IPv4 and IPv6 addresses, different networks may choose different routes, making one device work while another fails or changing the result when you switch Wi-Fi networks.

Test IPv6-related settings against your actual network; don’t disable IPv6 for every kind of issue. Continue investigating only if logs or comparison tests show that the IPv6 path fails while IPv4 works. After a change, test both a site that previously failed and one that worked to make sure you haven’t disrupted other traffic. Visiting a website by IP address is not a universal comparison: HTTPS certificate names, virtual hosting, and app APIs may require the correct domain.

Why domain and IP rules can produce different results

DOMAIN and DOMAIN-SUFFIX use domain information in the request. IP-CIDR matches address ranges, while GEOIP depends on addresses and related data. A website can use multiple addresses and may return different ones on different networks, so a single DNS result shouldn’t be treated as permanently equivalent to that website. A shared address range in a rule may also affect unrelated services. For website routing, first check whether an exact domain condition will work.

[Rule]
DOMAIN,example.com,PROXY
DOMAIN-SUFFIX,example.org,DIRECT
IP-CIDR,192.0.2.0/24,DIRECT,no-resolve
FINAL,PROXY

The 192.0.2.0/24 range in this example is reserved for documentation, not a usable server. no-resolve avoids triggering an extra DNS lookup just to match that IP rule; it doesn’t disable all DNS. Understand the effect before adding it to a real file, and don’t overwrite your everyday configuration with a copied snippet. If a request already has a destination IP, matching may differ from a request that only has a domain. Check connection logs rather than relying on the rule text alone.

If local device names fail, distinguish ordinary DNS from local network name discovery and specific service discovery mechanisms. Printers, storage devices, and home devices may use different discovery methods. Being able to reach an IP address doesn’t mean name discovery will work too. For troubleshooting, first restore the DNS path that works with your existing configuration, then re-enable custom settings one at a time. Record the failed name, network type, DNS path actually used, and error so the issue can be reproduced after an update or network change.

After fixing the issue, test one domain that previously failed and one that worked, then check the server’s own domain again. This helps ensure you haven’t fixed DNS for business requests only to break the next proxy connection. If the error has changed from a resolution failure to a connection timeout, continue to the server section instead of changing DNS that is already working.

Battery drain or overheating: Check traffic and repeated reconnects

An app’s percentage on the system battery page is a relative figure, not a direct measure of how much extra battery it used. Overall device usage, screen brightness, cellular signal, video playback, and background sync can all change the percentage. First compare battery drain, temperature, and network activity under similar conditions, then check whether Shadowrocket is involved. Deleting configurations just because the percentage increased may erase useful evidence without addressing the app that is continuously transferring data.

Compare the same workload under similar conditions

Choose times when device temperature and network signal are similar, then compare an idle connection with the same everyday activity. Don’t play video in one test and leave the screen locked in the other. Avoid comparing results while the system is finishing an update, syncing photos, or downloading many apps. Note whether the screen is on, whether the device is charging, and the general level of network activity. This helps distinguish encryption and forwarding load from network conditions or increased usage by an app.

The system may attribute some network extension activity to the client, but that doesn’t mean the client initiated every request. A background app that repeatedly syncs, retries a failed API, or streams media can make use of the proxy for those connections. Check connection logs from the same time to find recurring destinations and failure patterns, then investigate the app responsible for that activity. Don’t infer the request’s source from its displayed app name alone, and don’t block unfamiliar system domains just to lower the reported battery percentage.

Rule out repeated reconnects and batch testing

Weak signal, frequent switching between Wi-Fi and cellular data, and unreachable servers can all trigger repeated connection attempts. Repeated TLS handshakes, DNS lookups, and authentication may use more power and generate more heat than a stable connection. Check the logs for rapid sequences of connects, disconnects, and retries, then test on one stable network. If the issue mainly occurs while commuting between networks, investigate the reconnect pattern before changing all protocol settings.

Check the On Demand conditions too. Conditions that change frequently can make the connection start and stop as the network changes, though turning off on-demand connections permanently may not suit your needs. During diagnosis, temporarily disable automatic conditions and control the connection manually. If the issue goes away, restore conditions one by one. Also stop unnecessary full-list Connectivity Test runs; batch tests can create extra background requests and make actual traffic harder to assess.

Use logs and additional processing only when needed

Detailed logs are useful for reproducing an issue briefly, but keeping extensive logging enabled for a long time can add processing and write overhead. Note when the issue occurs, capture enough detail to diagnose it, then restore your usual logging level. Don’t keep every detail indefinitely while waiting for a rare problem. If your configuration includes extra rewrite or script processing, disable items one at a time in a copy and compare results. Use the same request and conditions to judge the difference; the number of processing items alone proves nothing.

If a configuration feature requires decryption inspection, make sure you understand its scope, certificate requirements, and source before deciding to use it. Ordinary connections, subscription updates, and basic routing troubleshooting don’t require adding a certificate. If the device heats up, stop unnecessary processing and high-load activity, then compare again after it cools down. The system may change device performance at higher temperatures; running repeated speed tests at that point adds load and makes results harder to compare.

To verify a fix, observe the device long enough to cover normal usage rather than checking the battery percentage for a few minutes. System statistics cover a time window, so changes may not immediately affect older records. If the device still heats up with the proxy off, investigate background system tasks, signal, and device condition. If only one Config causes the issue, keep a minimal working copy and restore items one at a time to find the cause. For quick answers, see the FAQ. Avoid changing power, network, and rules all at once.

Issues after an update: Separate app, system, and configuration changes

“It stopped working after an update” needs one clarification first: what was updated—Shadowrocket, the device system, the subscription list, or a remote Config file? These changes may happen around the same time but affect different layers. Recall the network, server, and file used for the last successful connection, then check the current state of each. Don’t assume the app caused every error just because it was updated recently. Server settings, removed subscription entries, or a network change may also be involved.

Save the current state, then run a minimal regression test

Before replacing files or deleting data, back up the current Config and save the log excerpt that shows the issue. A subscription and a configuration file may not contain the same information. Before restoring, confirm that server details, rule files, and additional references have all been saved. Backups may contain passwords, subscription credentials, or user IDs; store them somewhere you control. An exported file is a backup, not something to share publicly.

Run a minimal regression test with an existing server: turn off automatic conditions, connect manually, and visit a fixed destination under Proxy. If that works, switch to Config and check the rules. If the system switch won’t stay on, return to the first section. This sequence helps identify whether the system tunnel, proxy connection, or routing file is affected. Restore one item at a time so you can interpret the results.

Configuration syntax and changed references

If a newly imported or updated Config reports an error, check the section, line, and referenced policy named in the message. Common problems include commas replaced with non-English punctuation, rules split across lines, renamed policy groups, and explanatory text copied into the file. Fix clear syntax errors first rather than deleting every unfamiliar option. Snippets from different sources may use the same rule keywords, but that doesn’t guarantee they can be combined into a complete configuration for your needs.

After a subscription update, a selected server may have been renamed or removed while a policy still refers to its old name. Check whether the currently selected server and the policy matched by the rule are available, then decide whether to select another server or update the reference. If an old Config copy works but the new file doesn’t, compare rules, DNS, and additional processing section by section. Don’t immediately overwrite your only copy to restore service; you’ll lose the ability to identify what changed.

Permissions and connection status after a system update

If the system asks for authorization after an update, confirm it again in the current device settings and check VPN status and On Demand conditions. Restart the device once, then test on a fixed network. Don’t repeatedly reset network settings, reinstall the app, and delete configurations. System cleanup and sync tasks may continue after an update. If the main symptoms are battery drain or speed changes, wait for those activities to settle and follow the relevant section to test again; don’t confuse them with a clear handshake error.

If you need to download the app again, use the download page for all platforms. Get the iPhone / iPad version from the App Store for a one-time $2.99 purchase. The Mac (Apple silicon and Intel), Windows, Android, and Linux versions are free; installers are available through the site’s download paths on the download page. For iPhone / iPad, check the system requirements listed on the App Store page. Verify the developer, Shadow Launch Technology Limited, and app ID 932747118 instead of relying on the name alone.

If a desktop or Android installation is blocked by the system, first confirm the file came from this site, then follow the installation approval guide: on Windows, confirm the source and select “More info” then “Run anyway” in SmartScreen; on macOS, check “Open Anyway” under Privacy & Security; on Android, follow the system prompt to allow installation from the current source; on Linux, install according to the package format and make AppImage executable. Don’t disable system protection globally to handle a single installer.

Restore data and report a reproducible issue

Reinstall only after backing up and running minimal tests, not as the first step. On iPhone / iPad, check your purchased items with the original purchase account. After restoring the app, confirm whether your subscriptions and Config files have returned; store purchase history is not an app configuration backup. Before migrating, read iCloud sync, configuration export, and restore steps. When reporting an issue, include the system environment, what was updated, the exact error, and the shortest steps to reproduce it. Don’t speculate about a version or send complete files containing credentials.

iPad-specific checks: Network changes, local networks, and configuration migration

iPad and iPhone use similar connection logic, but their network sources, background behavior, and usage scenarios may differ. iPads often connect to home Wi-Fi, school or office networks, or a phone hotspot. A subscription that works on a phone doesn’t mean the iPad has the same network path. Check the iPad’s network source, system VPN status, selected server, and Config, then compare the two devices. Matching subscription labels alone don’t prove that their settings are identical.

Wi-Fi, hotspots, and web sign-in

When an iPad connects to a phone hotspot, the phone’s proxy status doesn’t prove that the iPad has established its own proxy connection. Check the Shadowrocket switch and actual request path on the iPad itself. Keep the iPad’s server and Config unchanged, then test on both regular Wi-Fi and the hotspot. If only one network fails, check its sign-in requirements, DNS, and protocol reachability. A brief interruption while switching hotspots doesn’t necessarily mean the subscription has failed.

School, hotel, and office Wi-Fi may require a web sign-in, acceptance of terms, or device verification. Disconnect the tunnel and complete the network’s sign-in requirements before reconnecting. Authentication may expire after you leave the network, lock the device, or reach the end of a session, so “it worked yesterday” doesn’t prove that you’re signed in today. A managed iPad may restrict personal VPNs or network settings. Ask the device administrator to check the policy; don’t remove management configurations that aren’t yours.

Screen lock, Split View, and stale connections after resuming

If a page fails after you unlock the device, first check whether the system VPN disconnected or the app is holding on to a stale connection. Check the system status, then close the failed page and make a new request. If the new request works, the issue may be an old connection that didn’t recover, not a failure of all servers. On Demand can reconnect under specified conditions, but it doesn’t guarantee that every background app can keep running. Check system background scheduling separately from the client connection state; don’t judge it only by whether the app’s screen refreshes in the background.

With Split View or multiple windows, several apps may download, stream video, or sync documents at once. When troubleshooting speed, pause transfers in other windows and keep one foreground destination fixed. External displays, charging, and high brightness can also affect heat; keep these conditions similar when comparing battery use. If only one app fails after resuming, reconnect that app first and compare with other apps. You don’t need to immediately rebuild the system VPN.

Local network devices, screen mirroring, and discovery services

Printers, home storage, and screen-mirroring devices usually depend on local network access; some also use broadcasts or service discovery. First confirm you can find the device with the tunnel off, then check what changes when connected. Make sure both devices are on a network that allows them to communicate, and check whether Wi-Fi client isolation is enabled. Devices using the same Wi-Fi name may still be blocked from communicating. If needed, check local network access permissions in system settings; the required permissions depend on the app making the request.

For local network routing, you can use an IP-CIDR rule that matches your network, but check the address range first and don’t add an overly broad subnet. If domain and direct IP access behave differently, also check name resolution or discovery. If screen-mirroring discovery works but playback fails, the problem may be with the subsequent media connection rather than discovery. Identify the failing stage before changing rules; a DIRECT rule doesn’t replace checks for discovery protocols or network permissions.

Check each setting after migrating from iPhone

After migrating a Config, check the active file, selected server, subscription status, and whether all referenced files are present. Confirm that iCloud sync has finished by checking the actual files and entries; don’t assume it’s complete just because both devices use the same account. System VPN authorization must be confirmed separately on the iPad, and subscription links may have changed. Connect manually and verify the connection on the iPad before restoring On Demand. Don’t apply every automatic condition from your phone to the tablet without checking whether it suits the iPad’s use.

If you still can’t isolate the issue, write a brief report with the device system, network source, whether it’s managed, Global Routing setting, exact error, result on another network, and whether a minimal configuration works. Before trying more changes, review the quick-start guide to confirm the basics, or look for the issue in the FAQ. Reproducible conditions are more useful than repeatedly importing, reinstalling, or changing several settings at once.