macOS Troubleshooting Guide
Clash Troubleshooting
Find where the request is getting stuck, then change the setting that addresses it.
For a first-time setup, start with the setup guide and follow the steps to add a subscription, choose a mode, connect, and verify the result. This page helps diagnose issues that have already occurred. After applying a fix, return to the guide’s verification steps and test again with an app you use regularly.
No internet: restore a direct connection first
Return to a state where traffic is not being intercepted
In the Clash client, turn off the system proxy and TUN. Then open System Settings → Network and check which network service is actually connected. A connected Wi-Fi icon only means your device is linked to an access point; it does not guarantee internet access. If your Mac is also connected through an Ethernet adapter or dock, don’t assume requests are using Wi-Fi. Temporarily unplug Ethernet or turn off Wi-Fi to test each connection separately.
In a browser, open a public website you can normally reach without a proxy, then try a second, unrelated site. Hotel and guest networks may require you to sign in through a captive portal; complete that login before re-enabling the proxy. If other devices on the same network also can’t reach either site, troubleshoot the network connection first. If only this Mac is affected, check that the active network service’s address, router, and DNS settings match the network’s requirements.
Test the system proxy and local port separately
Once direct access works, open the Clash client and leave both traffic-capture options turned off for now. Make sure the active configuration has loaded, then find the HTTP or mixed port in Settings and explicitly connect to the local proxy from the command line. The commands below use an example port and a public test domain; replace them if your port differs. The first command bypasses application-level proxy settings, while the second explicitly uses the proxy. If TUN is still enabled, the first command is not a clean direct-connection test.
curl --noproxy '*' --max-time 12 -I https://example.com/
curl --noproxy '' --proxy http://127.0.0.1:7897 --max-time 12 -I https://example.com/
If the second command can’t connect to the local port, check whether the core process is running, whether the port is correct, and whether the client exposes only a SOCKS port. If the proxy connection succeeds but the request later times out, don’t rush to change the port. In Connections or Logs, find the request by time and inspect its matched rule, outbound policy, and next error. If the request reached the core, you’ve narrowed the issue to the core or the route beyond it.
Choose the next check based on the scope of the issue
| Test result | Check first | Leave unchanged for now |
|---|---|---|
| Still fails with traffic capture off | Network login, router, existing network settings | Nodes and proxy groups |
| Direct connection works, explicit proxy fails | Listening port, core logs, outbound policy | Browser extensions |
| Explicit proxy works, app fails | System proxy, app proxy settings, bypass list | Subscriptions and nodes |
| Fails only with TUN enabled | Permissions, routes, other traffic-capture tools | App installer |
If the issue occurs only in rule mode, identify the rule that actually matched. A proxy group’s name does not tell you which outbound was used: a group named for a region may still select an unavailable node or include DIRECT. Temporarily switch the policy for one public test target and compare. Restore rule mode after testing. Global mode is not a good long-term fix, since services that should connect directly may also be routed through the proxy.
If everything works with traffic capture off but even local network resources become inaccessible with TUN enabled, test the resource using its usual method—for example, reconnect to a local share from Finder. Check whether a workplace networking tool manages routes and whether local subnets overlap. Stop at the setting that restores your work; don’t change a managed device’s network policy just to pass one test.
Restore settings one at a time, then retest
Enable only the traffic-capture method you were using when the issue occurred. Test a public website, the app that originally failed, and a resource you normally access directly. If public websites work with the system proxy enabled but the command line still fails, see the system proxy section. If multiple sites report that the server can’t be found but an app can reach a known local address, see the DNS section. Don’t change the node and DNS settings at the same time.
If your connection drops after quitting the client, return to the proxy settings for the active network service and check whether they still point to a local port. Remove only proxy entries left behind by this test; keep existing automatic settings and managed configuration intact. If it happens repeatedly, note whether the app quit normally or was force-quit. If only a force quit leaves the proxy enabled, address the abnormal exit first, then check whether a normal quit restores connectivity.
Node timeouts: distinguish health checks from real requests
Keep the test target and timing consistent
Latency tests rely on a configured probe URL. A timeout means only that this probe did not finish within its time limit; it does not mean every website is unavailable. Note the test time, network, proxy group, and node actually selected. Keep the same probe URL, test two other available nodes manually, and make a real request to the same public website. This helps distinguish a single bad node, a failing probe target, and an issue affecting all outbound traffic.
If real requests work and only the probe times out, check whether the group’s probe URL is reachable, expects a particular status code, or redirects. Fix the probe target or its success criteria before substantially increasing the timeout. A threshold that’s too long slows failover; one that’s too short may mistake a brief network wobble for an outage. An occasional high reading alone is not a reason to reconfigure everything.
Use the logs to find where the request stops
Look through the log entries just before the failure time. If the error says the node hostname can’t be resolved, check resolution of the node address, not rules for the website’s domain. If the hostname resolves but the connection to the remote address times out, the problem is more likely along the network path between your device and that address. If a connection is established but the handshake fails, check whether the subscription configuration is out of date, the client core supports the configured protocol fields, and the system date and time zone are correct.
Ignoring certificate errors is not a fix. Check the system clock, then use parameters confirmed by the configuration provider, or update the subscription while keeping a copy of the old configuration. If the log explicitly reports an unsupported field, check the current client’s compatibility notes. Don’t delete fields one by one just to get the client to start. A successful syntax check means the configuration can be read; it doesn’t mean the remote server will accept the connection.
Compare one node, others in its group, and another network
If one node times out, manually select another node in the same group and start a new request that isn’t already in progress. If the whole group fails but another group in the same configuration works, check whether the failing group shares a probe URL, upstream, or outbound dependency. The displayed selection in an automatic group may be stale or based on an earlier probe; use the actual outbound shown in the connection record.
If every node times out on your current Wi-Fi and direct access is confirmed to work, test on a phone hotspot if permitted. Keep the node, mode, and target the same. A successful hotspot test points to a difference involving the access network, but doesn’t by itself prove whether the cause is the router, address family, or a remote restriction. Run the same test on another device on the original network to see whether the issue follows the network or affects only this Mac.
If both networks fail and you have an older configuration that previously worked, switch to it on the same device for comparison. An outdated configuration can’t prove that the client is at fault. If multiple configurations fail, note whether they stop at the same stage. For example, a refused local connection and a timeout to a remote host are entirely different issues, even if both appear as connection errors.
Avoid misreading automatic failover
An active video, download, or long-lived connection may not migrate immediately when the group selection changes. Close the test connection after switching, then make a new request in the same app so a failure on the old node isn’t attributed to the new one. If the client lets you inspect connections by rule, check the target address or request time to confirm the outbound actually changed; don’t rely only on the selected style in the group list.
If selecting a working node manually restores access but the automatic group later switches back to a failing node, temporarily use a manual group to stay connected. Then check the automatic group’s probe URL, interval, and members before restoring automatic selection. When reporting the issue, keep only the failure stage and a redacted group name; don’t include the subscription URL or node credentials.
Subscription failures: separate download, parsing, and activation
Keep the configuration that currently works
A failed update doesn’t mean the active configuration has stopped working. Check when the client last updated successfully and whether current requests still work. Don’t delete the active configuration or clear all subscriptions yet. If the client supports exporting, back up the working configuration to a protected location on your device. It may contain private connection details and should not be shared as a routine log file.
Note whether the failure happens after you click Update, while the content is being parsed, or after you activate the new configuration. First check whether the client can retrieve the content; next check its format and structure; only then investigate policy references and core startup. If the UI only says “Import failed,” open the logs and find the earliest specific error at the same time, rather than relying on the final generic message.
Check the response status and update route
Copy the subscription URL for your client from the provider’s account page. When pasting it, check for leading or trailing spaces and line breaks; don’t edit query parameters yourself. If the response says you’re unauthorized or access is forbidden, check whether the subscription expired or was reset and whether the provider restricts subscription types. If requests are being rate-limited, stop refreshing and try again after the limit clears. Repeated clicks make the original error harder to identify.
If the error says the hostname can’t be resolved or the connection timed out, check which network the client uses for updates. A website opening in your browser only proves the browser can reach it; the client may use a different proxy route for background downloads. Check whether the client has a proxy setting for subscription updates, then compare direct updates with updates through a known-working proxy. If updating depends on a configuration that no longer works, use another working configuration or an allowed network to restore the update route.
Even a successful status code may return a login page, access notice, or provider error instead of the expected configuration. Don’t submit your subscription URL to an online parser. Use the client’s error details or the provider’s support documentation to identify the response type; inspect its contents locally if needed. For certificate errors, check the clock and URL or contact the provider—don’t disable certificate verification.
Use the error to find formatting, indentation, or reference issues
Links from the same provider may return a full configuration, a list of nodes, or a format for another client. A full configuration can usually be activated on its own; a node list or rule list may need to be referenced by an existing configuration. If import fails because top-level items are missing, first confirm what kind of output you received. Don’t paste node-only data at the end of an existing rules file just to get one import to work.
If the error includes a line number, check that line and the few lines above it. Common causes include misaligned list items, tabs copied into the file, an unclosed string, or content after a colon being parsed differently than expected. If you use custom overrides, disable them in a test copy first. If the original subscription loads but the merged configuration fails, fix the override rather than replacing a valid subscription.
If syntax validation passes but the core reports a missing proxy group, make sure each rule reference exactly matches the current configuration, including spaces and punctuation. A subscription update may have renamed a group while a local override still uses the old name. Fix one reference in a copy and reload it. Don’t change every unknown policy to DIRECT: the configuration might pass validation, but its traffic routing will have changed.
Verify the new configuration with a small set of tests
After importing the new configuration, select it as active and check that all proxy groups are present. Test two public targets that previously used different policies. If the connection fails only after activation, switch back to the known-working configuration and save the failure stage and relevant logs. When changing the automatic update schedule, avoid critical work hours. First complete a manual update successfully, then enable scheduled updates. For more examples, see How to Troubleshoot Invalid or Unparseable Subscription Links.
Slow speeds: separate connection delays from throughput
Find out whether the slowdown is before a page opens or during transfer
A blank page that takes a long time to open, a page that loads content gradually, and a download that stays slow throughout are different symptoms. For the first, check DNS resolution, connection setup, and the handshake. For the second, find out whether the delay affects the main page or assets such as images and scripts. For the third, measure sustained transfer speed rather than page-load time. Before testing, pause your own large file syncs and system downloads so background activity doesn’t use different amounts of bandwidth in each run.
Choose a public, stable test target and keep the device and network the same. First measure a direct-connection baseline, then test different nodes in the same proxy group. Change only one variable per run and alternate between the two routes to reduce the effect of network fluctuations. For download tests, use the same source and file; don’t compare speeds across different sites. For page-load comparisons, avoid results skewed by cached content.
Break down the timing of an HTTP request
curl --noproxy '*' --max-time 20 --silent --show-error \
--output /dev/null \
--write-out 'dns=%{time_namelookup}\nconnect=%{time_connect}\ntls=%{time_appconnect}\nfirst=%{time_starttransfer}\ntotal=%{time_total}\n' \
https://example.com/
These values are measured in seconds and accumulate from the start of the request. For a direct request, subtract the DNS lookup time from the connection time to estimate connection setup; subtract the connection time from the handshake time to estimate the handshake. The command discards the response body, so it’s useful for comparing wait times on small requests, not for measuring bandwidth. For a true direct test, turn off TUN first; a proxy-bypass option alone won’t bypass a virtual network interface.
To test an explicit HTTP proxy, replace the bypass option in the first command with --noproxy '' --proxy http://127.0.0.1:7897, using the port shown in your client. In this case, connection time may mainly reflect the connection to your local proxy, and DNS lookup time may not represent the remote hostname lookup. Compare end-to-end time to first byte instead, then use the logs to see whether the delay is in setting up the proxy or waiting for the upstream response.
Narrow down whether the cause is a node, rule, or client load
If both direct and proxied connections slow down, first check whether Wi-Fi retransmissions, router congestion, or background uploads are happening at the same time. Move closer to the access point or test over Ethernet. If direct access is stable and downloads are slow on only one node, try another node in the same group with the same file. A fast node probe and a slow download aren’t contradictory: a small probe doesn’t measure sustained throughput.
If only one site is slow, use your browser’s developer tools Network panel to find the request taking longest, then look for its domain in the Clash client’s connection records. The main page and its assets may match different policies. Identify the request that’s actually waiting before changing anything; don’t switch the entire browser to global mode just because images load slowly. An existing request may also remain on its original route, so compare with a new request.
If the Mac heats up or even scrolling becomes sluggish only when TUN is enabled, check Activity Monitor to see whether the client UI, core, or another app is consistently using CPU. Verbose logs and continuously refreshing connection lists can add overhead. Restore the usual log level, close unnecessary debug views, and repeat the same task. A brief CPU spike alone doesn’t show that there’s an ongoing performance problem.
Restore a policy you can explain, then check stability
If an automatic group probes too frequently, small fluctuations may trigger repeated node switches. During troubleshooting, pin a stable node and check whether a meeting or download completes without interruption before changing the probe interval or group members. If only one site slows down repeatedly at a particular time, note the time, resource, and outbound, then test again. Fix reproducible slow requests first; you don’t need to find one fastest node for every situation.
DNS issues: verify where lookups are actually handled
Distinguish failed DNS lookups from connections to an address that won’t respond
If your browser says it can’t find the server, check the spelling of the domain, then see whether other public domains fail too. A lookup failure for one domain may mean its records changed or the domain doesn’t exist; simultaneous failures for multiple public domains point to the resolver or network path. If an address was returned but the page still times out, don’t assume it’s a DNS issue—check whether the address is reachable and the handshake completes.
scutil --dns
dscacheutil -q host -a name example.com
dig +time=2 +tries=1 example.com A
dig +time=2 +tries=1 example.com AAAA
scutil --dns shows which resolvers the system currently has, including entries scoped to a domain or interface; it doesn’t perform a lookup for a particular domain. dscacheutil shows the system’s host lookup result, which may come from cache. dig displays a DNS response, but on macOS it may not select a server using the same system resolver routing as an app. If the results differ, understand the different lookup paths before clearing caches.
Compare the system, browser, and core
If a command resolves the domain but only the browser fails, check whether the browser has its own Secure DNS setting or an extension that changes how connections are made. Comparing another browser or a clean browser profile is more useful than resetting the entire Mac. Conversely, if the browser works but system lookups fail, that doesn’t prove the system’s resolver path has recovered.
When a request is forwarded over HTTP or SOCKS, the client core may resolve the domain instead of the app that made the request. In an explicit SOCKS test, socks5:// and socks5h:// resolve hostnames differently with curl: curl resolves the name locally with the first, while the proxy receives the hostname with the second. Use the same port and target when comparing them, or you can’t attribute the difference to the lookup location.
TUN and core DNS are separate settings. Even with DNS interception configured, confirm the request actually follows a path where it can be intercepted. Apps using encrypted DNS may bypass the usual DNS query path, so the toggle state alone doesn’t prove anything. In fake-IP mode, a virtual address is not itself a leak or an error; check whether the core can map the subsequent connection back to the correct domain.
Keep LAN and private domain lookups working
If public websites work but company intranet sites or local device names don’t, first turn off the new traffic-capture setting and check whether the existing network can resolve those names. Workplace networks may use dedicated DNS servers for specific domain suffixes; sending every lookup to a public resolver can break intranet access. .local names are often resolved over local multicast, so a query to a public DNS server can’t tell you whether a device is online.
Some configurations let you choose a resolver for specific domains or exclude domains from fake-IP handling, but these settings do different things: the first chooses where to send a query; the second changes the type of address returned. Check the documentation for the core you’re using to identify the relevant fields. Make the smallest change possible to a specific domain suffix, and only after confirming that this is where the problem lies. Ask your network administrator to confirm managed private-DNS rules.
Check the expected DNS route before calling it a leak, then retest
A test page showing a particular DNS provider doesn’t automatically mean there’s a DNS leak. First determine which resolver the app was expected to use, then compare the result with the browser settings, current mode, and core logs. After a change, make a new request in the same app and test both a public domain and one that should resolve locally. If only connections to IPv6 addresses fail, check IPv6 connectivity before disabling it system-wide.
System proxy not working: check the network service and app
Check the active proxy settings for the network service in use
An enabled toggle in the client doesn’t mean every app is using the current system proxy. Open System Settings → Network, select the network service currently carrying traffic, then check Web Proxy, Secure Web Proxy, and SOCKS Proxy under Details → Proxies. Labels may vary slightly across macOS versions. After switching Wi-Fi, a dock, or user accounts, confirm the settings belong to the service you’re actually using.
networksetup -listallnetworkservices
scutil --proxy
networksetup -getwebproxy 'Wi-Fi'
networksetup -getsecurewebproxy 'Wi-Fi'
Replace the service name in the last two commands with one shown in the first command’s output. Compare the enabled state, host, and port, and check for an automatic proxy configuration or bypass list. These commands show settings; they don’t confirm that anything is listening on the port. If the system points to 127.0.0.1, the proxy is on this Mac. If the client exits or the core isn’t running, apps that use this setting won’t be able to connect while that local proxy entry remains.
Make sure the system points to a port that is listening
In Terminal on macOS, run lsof -nP -iTCP:7897 -sTCP:LISTEN to see which process is listening on the example port. Replace it with the port shown in your client. If another process is using the port, identify it and its purpose; don’t force-quit an unfamiliar service. After changing the port, reapply the system proxy and verify it with an explicit proxy command instead of relying on the number shown in Settings.
This is a minimal mihomo port snippet for identifying the fields, not a complete configuration. A mixed port accepts both HTTP and SOCKS connections. If you only need local access, keep LAN access disabled; you don’t need to expose a LAN listener for a local browser to use the proxy. The client UI may override ports in its runtime configuration, so rely on the actual listening port and system proxy value after making changes.
mixed-port: 7897
allow-lan: false
Test the browser, terminal, and bypass settings separately
Browsers often use the system proxy, but extensions, managed browser policies, and custom proxy settings can change that. Make a new request to a public website, then find it in the client’s connection records by time. If the browser made a request but the core has no record of it, check the proxy entry point and bypass settings first. If the core has a record with a DIRECT outbound, traffic capture is working; check the matched rule instead of toggling the proxy again.
Command-line tools don’t always use the macOS system proxy. Some read environment variables from the current process; others use their own configuration files. To avoid saving the wrong port in a startup file, test with an option that applies to a single command. If a command succeeds when explicitly given a proxy, the issue is the source of proxy settings in that app. If it fails in just one terminal window, compare its environment without sharing variable values that may contain credentials.
A bypass list keeps certain hosts, such as local services, off the system proxy. If the target matches a bypass entry, it’s expected that the client won’t show a connection for it. Don’t clear the whole list; first check whether a single host or a domain pattern is matching unintentionally, and keep a record of the original settings. Managed proxy auto-configuration and bypass rules need a different approach than those on a personal device. For browser-versus-terminal differences, see Troubleshooting the System Proxy by App.
Know when TUN is actually needed
If an app can use the system proxy as required, you don’t need to enable TUN for it. For apps that don’t use the system proxy, check for a proxy setting in the app itself; consider virtual network interface capture only if there isn’t one. When testing TUN, disable the system proxy for that test and make sure you know which entry point is active. Then check DNS, local network access, and everyday apps. This makes it clear which setting caused any change in behavior.
Clash crashes: distinguish the UI, core, and permissions
Note where the app exits before trying to reinstall
An app that quits as soon as it opens, a working UI with no running core, and an app that exits only after switching configurations need different fixes. Note the last visible action and whether a system prompt appeared before it closed. If you’re unsure whether it’s still running in the background, check the relevant processes in Activity Monitor. A closed window doesn’t necessarily mean the app quit, and a frozen app doesn’t necessarily mean the core stopped listening.
If quitting the app disconnects you, restore connectivity using the no-internet section before investigating the crash. In macOS Console, find the latest report in the Crash Reports area by time and process, or check the logs provided by the client. The UI process and proxy core are separate, so identifying which process the report belongs to is more useful than looking at the window alone. Note the triggering action and a summary of the report; you don’t need to upload the entire file.
Check the Mac model, install location, and system alerts
In About This Mac, check whether it lists a chip or processor, then choose the installer for Apple silicon or Intel accordingly. If you need to reinstall or switch clients, check the recommended Clash Plus on the macOS downloads page first, then choose an option compatible with your system. If you use Clash Verge Rev or FlClash, match the installer to your Mac’s architecture too; don’t rely on the app icon to identify it.
Make sure the app has been copied to the Applications folder. Opening it directly from a disk image is not the same as running an installed app. If macOS explicitly says the app is incompatible with your system, check its supported versions. If the alert refers to a security or device-management restriction, follow the system’s instructions to verify the app’s source and permissions. Don’t disable system security to see whether it will launch, and don’t confuse a security alert with a process crash.
Use minimal changes to rule out configuration and startup conflicts
If the issue started after importing a configuration or enabling TUN, turn that setting off if you can open the UI, then restore the last configuration that worked. To prevent it from starting automatically, check any login items you added under System Settings → General → Login Items & Extensions. After restarting, open only one Clash client manually. Once it stays open, import the configuration and enable traffic capture only as needed.
If the UI opens but the core won’t start, check the core error first. Invalid configuration structure, a port conflict, or missing service authorization can all prevent startup; reinstalling the UI may not help. If the system proxy works but TUN fails, check the client’s service installation or authorization prompt and make sure another tool isn’t also capturing virtual network traffic. For managed devices, ask the administrator to confirm permissions.
If the UI starts quitting immediately after an update, check the client’s support documentation to see whether you can back up and temporarily disable user configuration before launching it. Only consider moving the client’s configuration directory for testing after confirming which app it belongs to, fully quitting the app, and making a backup. Don’t delete system-wide Library files in bulk or mistake another client’s directory for the source of the issue.
Restore features one at a time and record the results
If the app starts in a clean state, first verify the UI and core, then import a configuration in a known-supported format. Leave auto-start and TUN off until connection records look normal, then re-enable the features you need one at a time until the issue returns. If it still crashes in a clean state, record the Mac architecture, system version, process name, and steps to reproduce, then contact the relevant project’s support channel. Don’t attach subscription files or unredacted logs.
Mobile troubleshooting: permissions, background activity, and network changes
First check whether the system established a VPN connection
Mobile devices don’t use the same proxy entry point as macOS. Open the client in the foreground, approve the operating system’s VPN configuration prompt, and check in Settings which app owns the active connection. A VPN icon in the status bar only means some VPN connection exists; it doesn’t prove this client created it or that traffic is using a working policy. Test a public website, then check the matching request in the client.
If your phone also has a work VPN, another proxy tool, or a filtering app that uses a local VPN, check whether it’s occupying the system’s VPN slot. Pause only tools you’re authorized to change; don’t remove or override a work-managed connection. If you can compare on a personal device, different behavior with the same configuration on a personal and managed device points to device policy as something to check before blaming configuration compatibility.
| Platform | Check first | Test |
|---|---|---|
| Android | VPN permission, background battery settings, persistent notification | Unlock after locking, then make a new request |
| iOS | VPN configuration ownership, connection status, on-demand settings | After switching networks, check whether the connection recovers |
Android: check whether the app was stopped in the background
If the app works in the foreground but disconnects after the screen has been locked for a while, check whether the system’s VPN status and the client’s persistent notification disappear at the same time. Then check the app’s battery settings in System Settings; the location varies by manufacturer and may be under Battery usage or Background activity. Relax restrictions only for the client you use, then retest with the same lock-screen interval. There’s no need to disable battery saving across the whole device. See Android VPN Permissions and Battery Settings for details.
If the VPN remains connected but some apps have no traffic, check whether the client has an allowlist or exclusion list for apps. Also check whether Always-on VPN or blocking connections without a VPN is enabled in system settings. The latter changes how apps outside the VPN connect; it doesn’t automatically indicate a node problem. Restore your original security settings when testing is done, and don’t leave diagnostic changes in place.
iOS: distinguish a connection that didn’t recover from an old request that didn’t resume
Check which VPN configuration is active in iOS settings, then check the connection status in Clash Plus. If on-demand connection is configured, review its actual conditions instead of assuming it will reconnect immediately on every network. When switching from Wi-Fi to cellular, wait for iOS to confirm the new network is available, then make a new web request. An existing call or download that didn’t resume doesn’t prove the new connection is still unavailable.
If the issue occurs only on cellular, turn off proxying for this test and confirm that mobile data works and the affected app has permission to use it. If it occurs only on guest Wi-Fi, complete the network login before testing the VPN. Don’t keep switching between multiple configurations; stick to one client and one known-working configuration, then record the results on each network.
Recheck supported features when moving to another device
A subscription that imports on a Mac may include desktop configuration features that a phone doesn’t support. Local ports, overrides, LAN policies, and extension fields may need separate settings. Import a test configuration in a format supported by the platform, then verify one public destination and one resource that should connect directly. If only the phone fails after waking or switching networks, first record the traffic-capture status and result of a new request rather than changing the entire ruleset.
After fixing the issue, test again in the foreground, after locking and unlocking, and after switching networks. If it still occurs, include the device OS, network type, steps that trigger the issue, and VPN status at the time of failure in your report. Attach only redacted logs relevant to that attempt; don’t share unrelated app connection lists.
Back to contents ↑ Continue to the setup guide → Browse networking and configuration terms →