Host monitoring

Hosts give you server-level visibility through a lightweight push agent. Install it on a Linux or Windows box and it streams CPU, memory, disk and network metrics — no inbound ports, no scraping.

Read: 4 min
Sections: 4
Limits: Free 1 · Pro 25 · Business ∞

Add a host

Click New host to create a host record and get an enrollment token. The host starts in pending state until the agent checks in. Install the agent on the server with the token:

bash

# Linux — install and enroll with the token from the New host drawer
curl -fsSL https://your-datarunner.com/agent/install.sh | sh -s -- \
  --token dr_host_xxxxxxxxxxxx \
  --url https://your-datarunner.com

On Windows, run the equivalent installer from the drawer. Once the agent reports in, the host flips to online and its hostname, OS, and agent version populate.

Status & freshness

The list ranks hosts that need attention first. Each host is in one of four states:

online — agent reporting on schedule
stale — last sample is overdue; agent may be struggling
offline — no recent samples; the box or agent is down
pending— created but the agent hasn't enrolled yet

What the agent streams

The agent pushes CPU, memory and swap, disk usage per mount, network throughput, load average, uptime and process count on an interval. Samples are then downsampled in two steps: raw is rolled into 1-minute aggregates after 24 hours, and 1-minute into 1-hour after 7 days. Your plan's host-metrics retention applies to those 1-hour aggregates — 7 days on Free, 30 on Pro, 90 on Business — which is what the long-range trend charts on the host detail page read from.

Alerting on hosts

Host conditions feed the same alerting pipeline as monitors. Route host alerts to channels, escalate via on-call, and silence them during planned work — see Alerting and Maintenance windows.

Hosts pair naturally with endpoint monitors: monitors check the service from the outside, hosts watch the box from the inside. Both land in one incident timeline.