Servers and performance
The Servers page — status, rename, pause, revoke and the proxy mapping — and the Performance page with TPS, MSPT, CPU, memory, uptime, restarts and crashes.
Servers
Every agent linked to the workspace, with:
| Column | Shows |
|---|---|
| Name | Editable by members who manage servers, up to 40 characters. |
| Type | proxy, backend or standalone, as the agent reported on its last start. |
| Software | Platform and Minecraft version, and the agent version. |
| Status | See below. |
| Last heartbeat | How long ago the server last reported in. |
| Online | Players on it right now, while it is online. |
Status
| Status | Meaning |
|---|---|
| Waiting | Linked, but no heartbeat has arrived yet. |
| Online | Its heartbeats are arriving. |
| Silent | It missed three heartbeats in a row (three minutes with the default interval). Its open sessions and stays are closed at the last heartbeat, as Timed out. |
| Disabled | Paused from this page. |
| Revoked | Its token no longer works. |
A server that stops cleanly sends a stop before going quiet, so the dashboard can tell a restart from a crash: a start whose previous run never sent that stop is marked as a crash on every chart.
Pause and resume
Pause is a kill switch for one server. The agent learns it with its next batch, stops collecting
and stops sending; /analytics status shows "Disabled from the dashboard". Nothing it sees while
paused is ever sent.
A paused agent sends nothing, so it never hears that it was resumed. After Resume, run
/analytics reload on that server (/analyticsv reload on the proxy) or restart it. Its first start
after a pause is marked on the charts as a crash, since the paused run never sent a stop.
Revoke
Revoke retires a server for good. Its token stops working within 30 seconds; the agent is told on its next send, logs "This server's token was revoked", deletes its credentials and whatever it had not sent, and goes back to unlinked. What it had already sent stays in the workspace. To bring it back, link it again: it becomes a new server.
Rename, pause, resume, revoke and the proxy mapping need Link, rename, disable and revoke servers
(manage_servers).
Proxy mapping
A proxy reports its own server list — each entry's name and address — when it starts and whenever the list changes. A backend does not know what the proxy calls it, so the dashboard asks you to match them once:
- The Proxy mapping table lists every backend with its port, and suggests the proxy entry with the same port, marked Suggested.
- Nothing is saved until you Confirm. The match is stored and never guessed again.
- An entry with no confirmed backend is listed under the table: confirm a match, or install the agent on that server.
The table only appears once the workspace has both a proxy that reported its list and at least one backend.
Performance
Is the hardware keeping up with the players? One row per server, then charts for the server picked in the filter bar.
| Column | Definition |
|---|---|
| Status | As on the Servers page. |
| Uptime | Share of the range covered by heartbeats. |
| Restarts | Starts reported in the range. |
| Crashes | Starts not preceded by a stop. |
| TPS | Average ticks per second. |
| MSPT p95 | The slowest 5% of ticks, in milliseconds. |
| Lag minutes | Minutes with MSPT p95 above 50 ms. |
| CPU | Share of all cores the server's process used. |
| Peak heap | The most Java heap in use. |
| Peak players | The most players online at once. |
| Chart | Shows |
|---|---|
| Tick time | Average and p95 MSPT, with the 50 ms line above which a server falls behind 20 TPS. |
| TPS | Average and lowest TPS. |
| Lag | Minutes per period above 50 ms. |
| Uptime | Share of each period covered by heartbeats, with restarts (↻) and crashes (✕). |
| CPU | This server's process against the whole machine. |
| Memory | Java heap in use against the maximum. |
| Players on this server | Average and peak online. |
| Entities and chunks | What the world asks of each tick. |
| Restarts and crashes | Every start in the range, labelled restart, crash or agent update. |
What each platform can report
| Measure | Paper | Spigot | Folia | Velocity |
|---|---|---|---|---|
| TPS | ✓ | ✓ | Global region only | — |
| MSPT | ✓ | — | Global region only | — |
| CPU and heap | ✓ | ✓ | ✓ | ✓ |
| Entities and chunks | ✓ | ✓ | — | — |
| Players | ✓ | ✓ | ✓ | ✓ |
TPS is measured by the agent itself, from how long 20 ticks really take. MSPT comes from Paper's tick-end event, which Spigot does not have. On Folia the agent runs on the global region, which keeps its own pace whatever the world regions do, so its TPS and MSPT read as idle and entity and chunk counts are not reported at all.
Samples are taken every second and summarised into one row every metrics interval (60 seconds by default, set in Remote configuration). Raw metrics are kept for 90 days; older ranges read hourly and daily summaries.
Something missing on this page? Tell us on Discord