Get emailed when your managed server starts struggling — and when it recovers.
Server Health Alerts
The status page tells you how your world is doing while you're looking at it. Health alerts cover the rest of the time: the platform checks your managed server on a schedule and emails you when something is wrong, then emails you again when it's fixed.
This is automatic for every managed server. There is nothing to switch on.
What Triggers an Alert#
| Alert | Raised when | What it usually means |
|---|---|---|
| Server unreachable | Your server stops answering status checks | The machine is down, restarting, or out of memory |
| Health check failing | The health endpoint returns an error, or the world database can't be read | Players are being refused; usually a database problem |
| Boot errors | The server started with errors | Part of your world is missing — a zone that never loaded, content that never synced |
| Tick time over budget | Tick p95 goes above its budget (50 ms by default) | Players will feel lag |
| Tick rate shortfall | The world achieves fewer ticks per second than it's configured for | Something is repeatedly blocking the world |
| Memory pressure | The game process uses more than 85% of the machine's memory | The server is close to being killed by the operating system |
| Bandwidth per player | Traffic per player goes over its limit | A broadcast storm, or a message that's far bigger than it should be |
A single bad moment never alerts you. Everything except boot errors has to stay wrong for three checks in a row before you hear about it, and has to be good for three checks in a row before it's considered recovered. Boot errors are the exception and are reported on the first check, because a half-loaded world won't fix itself.
Tick and tick-rate checks are skipped for the first couple of minutes after a restart, so a server that's still warming up doesn't alert.
What the Email Says#
You get two emails per incident, never more:
- When it starts — subject begins with
[ALERT], names your project and the specific problem, gives the figure that crossed the line (for example "Tick p95 is 63 ms against a 50 ms budget"), and links to your status page. - When it recovers — subject begins with
[RESOLVED], and tells you how long the problem lasted.
Nothing repeats in between. A server that's been degraded for three hours has sent you exactly one email.
Alerts go to the project owner and to the platform's operations address, so someone is always looking even if you're away.
The Health Panel on Your Dashboard#
Studio → your project → the server card shows the live picture while you have the tab open. It refreshes every few seconds and pauses in a background tab.
| Shown | Meaning |
|---|---|
| Status dot and verdict | The one-line answer: healthy, under pressure, or needs attention |
| Players | How many are online right now |
| Tick p95 | Tick time against its budget, coloured, with a small graph of the last few minutes |
| Memory | How much the game process is using |
| Bandwidth | Network traffic per player, when your server reports it |
| Uptime | How long the server has been running since its last restart |
| Open alerts | Anything currently alerting, and since when |
| Notes | Things an operator should know — for example, a server provisioned without the shared cache a sharded realm needs |
If your browser can't reach your server directly, the panel falls back to the platform's most recent check and tells you the reading is stale rather than showing nothing.
Status page on the panel opens the full status readout for the project.
Alerts tell you that something is wrong. The status page tells you what — open it first when an alert arrives.