Diagnose a server that's down (not responding)
Check status, logs, CPU/RAM, and infrastructure status when the server isn't responding.
Diagnose a server that’s down (not responding)
When your server stops responding (no one can connect, the panel shows it offline), follow a quick diagnostic method to isolate the cause before contacting support.
Cause / The problem
An outage can come from the server itself (process crash, RAM saturated, corrupted world), from the network (blocked port, Anti-DDoS active), or more rarely from the onesubnet infrastructure. Without a method, you waste time blaming the wrong layer.
Solution
- Check the status in the panel: Dashboard → your server. The indicator should read Running (green). If Stopped or Crashed, jump to step 3.
- Check the global status: consult the onesubnet status page (status.onesubnet.com) to see whether maintenance or an incident is in progress on the Paris node. If so, that’s the most likely cause.
- Start the server from the panel (Start button). If it crashes immediately, the problem is in the server files (faulty plugin, corrupted world, RAM too low).
- Read the logs: Console tab or the
/logs/latest.logfile. Look at the last lines before the crash:OutOfMemoryError→ not enough RAM, upgrade the plan or lower-Xmx.Caused by: ...from a plugin → remove that plugin.Failed to bind port→ port already taken, change it in Variables.Corrupted chunk→ damaged world, restore a backup.
- Check RAM and CPU in the panel’s Resources tab. If usage is at 100%, the server can no longer respond. A clean restart frees up memory.
- Test the network connection from your PC:
ping server_IPand trytelnet IP port(or Test-NetConnection in PowerShell). If ping works but the port doesn’t, the firewall or Anti-DDoS is filtering. - Check the Gcore Anti-DDoS: if it has detected an attack, traffic may be temporarily mitigated. Wait it out or contact support.
- Restore a backup if the crash persists after removing suspect plugins/worlds.
If nothing works after these steps, open a ticket and attach the last 50 log lines plus the exact time of the problem — that’s the most useful information for support.