Server Metrics

Server metrics bring host-level telemetry — CPU, memory, disk, load — into the same place as your uptime and performance checks. Unlike every other monitor, it's push-based: a small agent on your box sends metrics to UptimeStatus, so it works for private servers with no inbound access.

How it works

You install a lightweight agent on each server. On an interval, the agent samples the host and pushes a metrics payload to your ingest endpoint. Tracked signals:

• CPU utilisation (overall, cpuPercent)
• Memory utilisation (ramPercent) plus used / total in MB (ramUsedMb / ramTotalMb)
• Disk utilisation (diskPercent, optional - send it if your agent collects it)

Because it's push, there's nothing to expose publicly - the agent dials out to UptimeStatus. Every sample is recorded as a check result, so it charts per host and feeds the same history and uptime machinery as any other monitor. A push is a JSON POST to /api/agent/<metrics-token>/metrics (the token is shown when you enable metrics on the monitor):

curl -fsS -X POST https://uptime-status.org/api/agent/<metrics-token>/metrics \
  -H "Content-Type: application/json" \
  -d '{"cpuPercent":37.2,"ramPercent":38.4,"ramUsedMb":6112,"ramTotalMb":16384,"diskPercent":68}'

The agent handles this for you on a schedule; the raw call is shown so you understand the shape. cpuPercent, ramPercent, ramUsedMb, and ramTotalMb are required (percentages 0-100); diskPercent is optional.

What triggers an alert

• A metric crosses its threshold. Each push is evaluated against the monitor's thresholds - defaults are CPU >= 90%, memory >= 90%, and disk >= 85% (disk only when the agent reports it). A breach marks the host down and opens an incident, which fans out to the monitor's notification channels; the next healthy push resolves it. Set any threshold to 0 to disable that metric.
• No metrics received within the expected window (the agent stopped or the host is down) - a missed push works like a heartbeat. The window defaults to 300 seconds and is checked every minute.

Thresholds and the missed-push window live in the monitor's config (cpuThreshold, ramThreshold, diskThreshold, metricsWindowSeconds).

Setting it up

1. Add monitor and choose Server Metrics.
2. Copy the host token / ingest URL for the new host.
3. Install and start the agent on your server with that token.
4. Adjust the thresholds (cpuThreshold, ramThreshold, diskThreshold) and the missed-push window (metricsWindowSeconds) in the monitor's config if the defaults (90 / 90 / 85, 300s) don't suit the host.
5. Attach notifications.

Related

• Cron & Heartbeats · Performance · Port Scan
• Notifications
• Marketing: Server metrics feature