Scope: this page starts from an existing configuration with the interface showing a connected state. If you have not installed or imported anything yet, the setup walkthrough comes first; subscription fetch or format errors belong to the subscription pages. Evidence boundary: assembled from the official project, Android VPN documentation, the project's own interface wording, and one clearly labelled community report. No Android device reproduction, node reachability test or fix-rate data is claimed here — what follows is a diagnostic order, not a measured repair result.
Pick the entry point by symptom
| Symptom | Check first | Then |
|---|---|---|
| Ordinary browsing also fails after disconnecting | Android Always-on VPN, block-connections-without-VPN, local network | Safe exit at the end of this page, then rebuild a network baseline |
| Network is fine, every app fails while connected | Selected server, start-up log, Android VPN policy | Configuration and core branch first, policy second |
| Some apps work, the browser does not | Per-app proxy and bypass mode | Whether the affected app is included or excluded |
| Log mentions name resolution | DNS settings and recent changes | DNS branch only, no blanket rules |
| Only the automatic test says success | Actual app behaviour and its log | Do not treat the test result as a connectivity verdict |
Change one setting per round and keep the original value. That is what lets you tell which step caused the change — and what lets you get back.
Step 1: build an ordinary network baseline
Record the selected profile, the per-app mode and the Android VPN settings, then stop the v2rayNG connection and open an HTTPS page you normally reach.
If it still fails, check whether the system has Always-on VPN enabled and whether connections outside the VPN are blocked. Android supports both restrictions, and while such a policy is active, disconnecting the VPN does not necessarily restore ordinary networking. On a device you manage, record the original settings and lift the restriction temporarily; on a managed device you cannot change, keep the restriction in the record and do not attribute the symptom to a node.
Only compare before and after starting v2rayNG once ordinary networking works.
Step 2: confirm the profile and the core actually started
Return to v2rayNG and confirm the selected item is the configuration you intend to use. Record its name or group only — no full configuration, subscription address or credential in the notes.
- Note the time you started the connection.
- Look around that timestamp for profile-load, core-start or connection-handling errors.
- If the log names a cause, keep the original line and a little context.
- Line the log time up with the time the page or app failed.
The official project supports the Xray and v2fly cores. When the failure looks configuration- or core-related, record the app version, the active core and any core version information you can obtain, and hand it to the profile provider or project support. With insufficient evidence, do not infer from "connected" that the protocol, core and remote service are all healthy.
If you have a second profile, compare it without changing anything else, but the candidate still needs verification — the project's own interface text warns that the automatic test result can be inaccurate.
A specific case, not a rule: v2rayNG issue #6141, checked 2026-11-12, reported the 2.3.5 F-Droid build with a TLS node showing latency −1, with Failed to build TLS config and allowInsecure has been removed in the log. That is one environment. A −1 does not by itself mean the same cause; if your log shows the removed-field error, work through the allowInsecure page instead of continuing here.
Step 3: audit the Android VPN policy
Android documents that Always-on VPN can start the VPN service at boot, that the system can block connections that bypass the VPN, and that a badly configured per-app allow or exclude list can leave some apps without network access. So answer two concrete questions: does the system require every connection to go through the VPN, and is the failing app outside the VPN scope?
Record the current state before changing anything, and do not modify the system policy and several in-app options in the same round.
Step 4: if only some apps fail, check per-app routing
The current project interface includes per-app proxy and bypass mode. As generally described in this site's tutorials: in normal mode, ticked apps enter the proxy scope; in bypass mode, ticked apps are excluded and go direct. Confirm against the interface text of the version you run before acting.
Compare one app that works with one that does not, and check whether per-app proxy is enabled, whether bypass mode is enabled, whether both apps are ticked, and whether that combination matches the path you want. Reconnect and retest the same failing app. If the result changes, record which option changed it; if not, restore the original settings and keep going.
Some apps working does not prove the node is healthy, and it does not prove DNS is at fault. It is a clue about application scope, and nothing more.
Step 5: enter the DNS branch only with a resolution clue
Look at DNS when the log mentions name resolution, or when the problem began right after a DNS change. Record the current local DNS, FakeDNS and related routing settings; if you changed them recently, restore them one at a time and watch the same app and the same target.
| Setting | What the interface wording supports | How to use it while diagnosing |
|---|---|---|
| Local DNS | Described as DNS handled by the core | Do not infer from the name that a particular direct path is used |
| FakeDNS | Described as faster, incompatible with some apps | If enabled and a specific app is suspect, compare with it off |
| Mux | Warned as possibly causing unstable connections | If enabled, use it as one separate comparison — not as the proven cause |
These notes are not a recommended configuration.
The UDP 53 community case: in a project discussion dated 2026-10-06, an Android 16 user reported that some apps worked while the browser or system apps reported no network under the VPN; the reporter and another participant said a UDP 53 direct rule helped their respective cases. That is a community report. It does not show your device shares the cause, it is not an official general fix, and this page has no reproduction, scope or privacy assessment for it. A "direct" rule is a routing decision with privacy consequences — do not apply it in bulk before you understand your own DNS and routing setup.
Step 6: verify with real access
- Does the page that failed now open?
- Does the app that failed now reach the network?
- Do the apps that worked still work?
- After disconnecting, does ordinary networking return as expected?
An IP check can help compare the egress before and after, but it is not a substitute for testing other apps, DNS or node stability. When the automatic test disagrees with real use, keep both results instead of picking the convenient one.
Safe exit: get ordinary networking back
- Stop the v2rayNG connection and check the Android VPN state.
- Check Always-on VPN and block-connections-without-VPN.
- If the goal is restored direct networking and you manage the device, lift the restriction as needed.
- Restore the per-app, DNS, FakeDNS or Mux settings you changed in this round.
- Retest the pages and apps that used to work.
Do not treat clearing the subscription as the first step to restoring networking. If it still does not recover, keep the policy state, app version, profile source and logs, and keep narrowing.
What to keep when asking for help
Android version, v2rayNG version, current core information, whether all or only some apps are affected, how ordinary networking behaves with the VPN stopped, and the log from the failure. Remove subscription tokens, passwords and full configurations before sharing. There is no single reproducible error string for this symptom, and "connected but no internet" does not yet identify one cause.
Common questions
The app says connected. Can I conclude the node works?
No. Connected is a local state: the client believes a core is running with a profile. Verify with a real HTTPS page and an egress IP check.
Should I turn off Always-on VPN to test?
Only on a device you manage, and record the original settings first. On a managed device keep the policy and ask the administrator.
Is the UDP 53 direct rule a general fix?
It is a community report, not an official general fix. It changes the DNS connection path and has privacy implications. Do not apply it in bulk before you understand your own DNS and routing configuration.
Related
allowInsecure has been removed · TUN enabled but no internet · Error index