Troubleshoot a noobtunnel connection
A public tunnel has several stages. Testing them in order gives you a useful failure report and avoids changing unrelated routes or firewall rules. Work only on devices and services you are allowed to inspect.
Start at the application
Request the backend from the agent's machine or container using its configured host, port and protocol. If that fails, check whether the process is running, which address it listens on and whether a local firewall blocks that port. A backend bound to 127.0.0.1 cannot be reached as localhost from a different container or host.
Check the backend's own logs at the test time. A successful request from your browser on the application machine is not enough when the agent uses a different network namespace.
Agent can't reach the resource
An agent can be online while its resource target is unreachable. Start with the Address and Port saved in Resources, then test that exact backend from the agent's network environment. A request that works on the host may still fail inside an isolated Docker container.
Check Docker networking
Replace YOUR-AGENT-CONTAINER with your agent container's name and inspect its network mode:
docker inspect --format '{{.HostConfig.NetworkMode}}' YOUR-AGENT-CONTAINERThe noobtunnel Docker installer sets network_mode: host. On Linux, this shares the host's network, so 127.0.0.1 reaches services on that host. Docker Desktop requires host networking support to be enabled; check its networking settings if you run the agent there.
With bridge networking or another isolated container network, 127.0.0.1 refers to the agent container itself. For a backend in another container, attach both containers to the same user-defined Docker network and use the backend's container name and listening port as the resource target. For a backend on the host, use a host address reachable from the container and make sure the application listens on that address.
Verify the backend address and port
- Check that the application is running and listening on the configured address and port. A listener bound only to its own container's loopback address cannot accept connections from the agent container.
- For separate containers on the same Docker network, use the application's container port. For access through a port published on the host, use the published host port.
- Test the backend from inside the agent container using an available HTTP client or network diagnostic tool. If it fails there, check container network membership, name resolution, routes and firewall rules before changing public DNS.
- After the backend responds from the agent's network environment, save the correct target in Resources and retest the public URL.
For the local test website and publishing steps, return to Host a resource behind tunnel.
Check the agent and mesh separately
Look for the agent's online status and recent handshake or connection information. If it is offline, inspect its service log and outbound route to the configured control endpoint. Verify the clock and certificate settings instead of disabling certificate validation.
For a private mesh problem, test the destination's mesh address and application port from another enrolled device. If the control panel works but mesh traffic fails, inspect the mesh route, destination firewall and application listener. The control connection and the WireGuard data path are separate.
Verify the public exit and DNS
Read the exit assigned to the resource and compare it with the public hostname's DNS records. An A record pointing at the control panel instead of the chosen publishing exit can send the request to the wrong machine. Check AAAA records too; an outdated IPv6 record can cause failures for some visitors while IPv4 tests succeed.
DNS caches can retain older answers until their TTL expires. Compare the authoritative records with what your client resolves. Test the actual public hostname from an external connection and note which address it reached.
Separate TLS and proxy errors
A certificate name mismatch means the connection reached an endpoint that did not present a certificate valid for your hostname. Confirm the domain verification, DNS and certificate status. A visitor-facing valid certificate does not prove that the backend HTTPS connection is configured correctly.
If the proxy returns a gateway error, compare its request log with the backend log. No backend request suggests a failure before the application: address, port, agent route or protocol. A backend request with an application error points to the application itself. Allowance exhaustion can also change behavior; check the resource's monthly usage and the published limits.
Prepare a useful report
Record the public hostname, affected resource, UTC time, resolved IP, response code and whether local backend and mesh tests succeeded. Include a short sanitized log excerpt. Send private service reports through Contact; remove cookies, passwords, tokens and private keys.