Velocity cannot connect the player to the backend (backend unreachable)
- Platforms
- Velocity · whole network
- Verified
- Verified
- Confidence
- Confidence 85%
What it looks like in the log
Any one of these lines is enough. The agent matches them locally, on your server.
logs/latest.log
unable to connect to serverUnable to connect you to
Symptoms
- Players cannot get onto the server: “Unable to connect you to <server>. Please try again later.”
- In the proxy log: “<player>: unable to connect to server <name>” with Connection refused / timeout
Cause
The proxy could not establish a connection to the backend: the backend process is stopped, has crashed or is restarting, has hung (does not respond within read-timeout), or the address in velocity.toml is wrong.
Possible causes
- The backend is off, has crashed or is restarting (Connection refused) — the cloud correlates this with the backend state (NW-01)
- The backend has hung or is overloaded and does not respond within read-timeout (ReadTimeoutException)
- Wrong backend address/port in the [servers] section of velocity.toml, or a firewall between the proxy and the backend (connection timed out / No route to host)
- noteVelocity kick messages are translatable and can be overridden; the primary source is the Velocity agent's API events, and this log signature is supplementary
How to fixLow risk
- Look at the exception under the
unable to connect to server <name>line in the evidence:Connection refused— nothing is listening on the address;connection timed out/No route to host— network or firewall;ReadTimeoutException— the backend accepts the connection but does not respond (hung or lagging). - Check whether backend
<name>is running and not stuck in a restart loop (the hosting console/panel, the end of its latest.log). - Compare the backend address and port under
[servers]invelocity.tomlwithserver-portin that backend'sserver.properties. - If the backend is reachable but responds slowly, look for the cause of the hang on it (TPS/MSPT, GC pauses, watchdog).
- To stop players being dropped from the proxy entirely, keep a working fallback server in
try.
Translated from the Russian original; log lines are quoted verbatim.