1. Check That the Subscription Updated and Loaded Successfully

First, confirm which configuration the core is running. Nodes may still appear in the list because the client kept the last successfully loaded configuration. Updating the subscription, applying the configuration, and starting the core are separate steps.

  1. In the client's configuration page, find the currently active profile and update it manually. Note the update time and any errors. If the profile was never enabled, a successful update won't change the active connection.
  2. For an HTTP 401 or 403 response, check the subscription URL and account status, and confirm whether the server restricts access. For a 404, check whether the URL has changed. These responses describe the subscription request; they don't prove that a node is offline.
  3. If the download completes but loading fails, check the first parsing error in the log. If the response is a login or error page, fix the subscription URL. If the error points to a protocol type or field, first verify which core and version the client is actually using, then compare it with the configuration requirements.
  4. After applying the configuration, check that the expected nodes and proxy groups appear on the proxy page. If the configuration uses proxy-providers to fetch nodes separately, also check the provider's update time and logs. A successful update of the main configuration doesn't mean the provider fetched a new list.

2. Check Whether the Local Proxy Port Is Listening

First, verify that the application can connect to the core on your device. mixed-port is the local proxy endpoint shared by HTTP and SOCKS; it isn't the remote node's port. The example below uses 7890. For testing, check the port actually in effect in the client settings, since the client may override the port in the subscription.

mixed-port: 7890
allow-lan: false

Check the listener first, then send an explicit proxy request

In Windows PowerShell, check which process is listening on the port. Match OwningProcess to the PID in Task Manager's Details tab to confirm the listener is the current client's core, not another proxy process.

Get-NetTCPConnection -State Listen -LocalPort 7890 |
  Select-Object LocalAddress, LocalPort, OwningProcess

Next, use curl to explicitly specify the local HTTP proxy. On Windows, use curl.exe; on macOS and Linux, you can use curl instead. Set the connection timeout to 5 seconds and the total request timeout to 15 seconds. These are test settings, not node speed-test results.

curl.exe -v --noproxy "" --proxy http://127.0.0.1:7890 --connect-timeout 5 --max-time 15 https://example.com/
  • Connection to the local port is refused: First check whether the core is running, whether the port is correct, and whether the listening address includes the local IPv4 loopback address. If the startup log says the port is already in use, identify the process using it before stopping the old instance or choosing another port. If you change the port, update the system proxy and test command too.
  • The local connection succeeds, but the request then fails: The local proxy endpoint is reachable. Check the core logs for the same time to distinguish local proxy authentication, DNS resolution, rule rejection, and outbound connection timeouts.
  • The target returns a response: To verify a specific node, confirm in the logs that the request actually used that node rather than DIRECT. A successful request only verifies the path used for that request.

3. Check System Proxy, App Proxy, and TUN Separately

If an explicit proxy request works but the browser doesn't, first check whether the browser is sending requests to the core. Visit the same target and check the connections page or logs for an entry at the same time. If there isn't one, verify the proxy endpoint before switching nodes.

Entry pointWhere to checkWhat to check and do
Windows 11 system proxy Settings → Network & internet → Proxy If the client uses the manual system proxy, check that the address and port match the listener, and make sure the target isn't covered by a proxy exception.
App-specific proxy The app's network or proxy settings Check whether the app follows the system settings or uses a separate address. If its proxy is configured as SOCKS, use a SOCKS-compatible port; don't reuse an HTTP-only endpoint.
TUN mode The client's TUN toggle and core startup log If the issue only occurs with TUN enabled, check whether the virtual interface, routes, and DNS interception are working. Also check whether another VPN is changing the default route at the same time.

The system proxy affects only apps that use that setting. TUN handles traffic routed through its virtual interface; it's a separate mechanism, not another system proxy toggle. During troubleshooting, change one entry point at a time and retest the same target. Node speed tests built into the client's proxy page are usually initiated by the core, so they may work even with the system proxy off. A successful speed test doesn't prove the browser is using the proxy.

4. Distinguish Node Server DNS from Target Website DNS

Check the logs to see which domain failed to resolve. If a node's server is a hostname, a DNS failure can prevent the node from connecting at all. If the failed lookup is for the target, check the routing and proxy protocol to determine whether resolution happens in the app, the local core, or remotely.

Check which DNS resolvers the core is using

This mihomo DNS configuration snippet illustrates the roles of each field: nameserver handles regular DNS queries, while proxy-server-nameserver is specifically for node server hostnames. The example uses IP addresses for upstream resolvers; it is only useful for diagnosis if the current network can reach them. Back up the original configuration before testing one change at a time. Don't replace the DNS configuration used for your current routing rules.

dns:
  enable: true
  listen: 127.0.0.1:1053
  nameserver:
    - 1.1.1.1
  proxy-server-nameserver:
    - 223.5.5.5

If dig is installed, send a query to the listening port shown in the example. This only checks whether the core's DNS listener can answer that query. It doesn't test the upstream reserved for node server hostnames, or change the system DNS settings.

dig @127.0.0.1 -p 1053 example.com A +time=2 +tries=1
  • Query times out: First confirm that the DNS module is listening on 127.0.0.1:1053, then check whether the upstream resolver is reachable. A system lookup without an explicit server may test a different DNS path.
  • The query returns a response, but the logs still show a node hostname resolution failure: Check the node address for typos and verify that the upstream resolver for node server hostnames is reachable on the current network. Resolving the target domain doesn't replace this check.
  • Returns an address such as 198.18.0.x: If fake-IP is enabled and the address falls within the configured fake-IP range, it's usually a virtual address mapped to a domain. Don't use ping on this address to determine whether a remote node is online.

5. Check Rule Matches and Proxy Group Selections

Once the core receives a request, check which rule matched at that time, then check what the selected proxy group actually chose. In rule mode, rules are evaluated in order, and the first match determines the route. A request matching DIRECT didn't use a remote node; one matching REJECT can't be used to test node connectivity either.

For example, if the current configuration already has a proxy group named "Manual Select," temporarily add the specific domain rule below to the existing rules list, ahead of any rules that might match first. Don't add a second top-level rules section or replace the entire ruleset with this example.

- DOMAIN,example.com,Manual Select

Select a node to test in that group, then send the same request again. Record the target domain, matched rule, proxy group, and final node. If the request still goes direct, check that the client is in rule mode, that the group isn't set to DIRECT, and that an earlier rule isn't matching first. If the expected domain doesn't appear in the client logs, check whether the app is connecting directly to an IP address. Remove the temporary rule after testing.

6. Compare Nodes Using the Same Target to Isolate Node or Speed-Test Issues

Compare nodes only after confirming that the request used the node under test. Keep the device, network, and target address the same, and test nodes A and B using the same manual selection group. Establish a new connection each time you switch nodes so the browser doesn't reuse an existing connection. Compare them on the same network first; then, if needed, retest on another network you're authorized to use.

What you seeNext step
A times out; B reaches the same target Check A's outbound errors. For connection timeouts, first check the server address, port, and network path. For TLS or authentication errors, verify the parameters supplied by the subscription, device time, and core compatibility. Don't disable certificate verification as a troubleshooting step.
All speed tests fail, but the site works through a confirmed node Check whether the speed-test URL is reachable through that node and whether its response status meets the configured requirements. Speed tests and regular traffic use different targets, so a failed speed test doesn't prove all nodes are offline.
Several nodes fail at the same stage Check the shared components: node hostname resolution, shared server addresses or ports, and the current network's connection requirements. Don't treat a batch of identical errors as separate failures.

When reporting an issue, include at least the failure time, core version, test target, a redacted node identifier, and the rule match and first outbound error from the same time. Indicate whether the failure occurred while connecting locally, resolving a domain, connecting to the remote server, or during the handshake. This is more useful than simply saying "all nodes timed out."