Skip to content
CubeMind_beta
Sign up

Health, latency and dead dedicated IPs

How a dedicated IP is checked, and what happens when one dies.

Dedicated IPs are checked on their own, not only when a client tries to use one. Each check opens a real connection through the endpoint to the game server of a client that uses it. An endpoint no client uses yet can only be checked against a generic host, and says so: attach it to a client and check again.

What you see for each one

  • Status: HEALTHY, DEGRADED after a failed check, DEAD after several in a row.
  • Latency, in milliseconds. Also shown when you pick one for a client, so you can choose a fast one without leaving the form.
  • When it was last checked, and how many checks have failed.
  • How many clients use it, and how many of those are running.
  • not used by any client, for spares.

Test checks it again now; the status updates a few seconds later.

Latency matters more than it looks: it is added to every packet between the client and the game server. An endpoint on the far side of the world will make a client feel laggy and can cause timeouts on strict servers. Where you can, pick one near the server.

When one dies

A dedicated IP that stops answering is marked dead, you get a notification, and every client assigned to it is flagged on the Clients page. A dead one is still checked, less often, and comes back by itself when it answers again.

  • Swap dedicated IP: move that client to a healthy one.
  • Fix all dedicated IP faults: narrows the Clients page to every affected client, so you can work through them. This is the one to use when a whole provider goes down.

You can also clear it entirely, which puts the client on the platform's shared IPs.

A dead dedicated IP does not silently keep a client down without saying so. That is the point of the attention list.

Clients stop, they do not limp

A client whose dedicated IP fails mid-session behaves like any other disconnect: it reconnects if auto-reconnect is on, and it will keep failing while the endpoint is dead. Swap it and the client recovers. A client started while its dedicated IP is marked dead connects without it.

Health, latency and dead dedicated IPs · CubeMind docs