Monitoring Overview

Swarm Hosts includes host-local monitoring for swarm hosts and deployments. When monitoring is enabled on a swarm host, the browser can show resource charts, collector health, and game health signals without sending raw metric history through the control plane.

What You Can Monitor

  • swarm host CPU, memory, filesystem, disk I/O, and network usage
  • deployment CPU, memory, disk I/O, and network usage
  • game health metrics, such as player count or tick rate, when the selected game exposes those metrics
  • monitoring stack health and current alert state

The deployment monitor focuses on the selected game server container. The swarm host monitor focuses on the machine running the agent and Docker.

How Monitoring Data Is Handled

Raw time-series history stays on the swarm host. The control plane handles login, permissions, monitor session setup, and low-rate health summaries, but it does not proxy chart payloads.

The browser uses a short-lived monitor session to reach the host-local monitor gateway. When direct access is not available, the browser and swarm host can use WebRTC signaling through the control plane; successful metric payloads still move between the browser and the swarm host.

Requirements

Monitoring needs:

  • a swarm host agent version with monitoring support
  • Docker on the swarm host
  • enough local disk for the retention window
  • a working browser-to-swarm-host monitor path, either direct or WebRTC-backed

Older agents and hosts without the monitoring stack still show the normal deployment status, logs, and heartbeat data. Monitoring failures should not stop game containers.