# Nodes

> Add, monitor, and remove database nodes — status, CPU, memory, disk, connectivity, and offline behavior.

Source: https://fleetdock.dev/docs/nodes


A **node** (shown as a *server* in the dashboard) is a host running the Fleetdock Agent. It
is where managed database instances live and where operations are executed locally.

## What a database node is [#what-a-database-node-is]

A node is any Linux server you connect with the agent. Fleetdock records its status,
operating system, memory, disk, Docker availability, and detected database engine version,
and it keeps a time-series history of resource metrics from heartbeats.

## Adding a node [#adding-a-node]

Open **Servers → Connect server** to generate a single-use registration token and install
command, then run it on the server (see [Agent installation](/docs/agents)). The node
enrolls and appears within about 30 seconds.

## Node status [#node-status]

Each node reports a status derived from its heartbeats:

* **Online** — heartbeating normally.
* **Maintenance** — you can put a node into maintenance while you work on it.
* **Offline** — no heartbeat for the timeout window.

## Resource metrics [#resource-metrics]

Every 30-second heartbeat records a snapshot that is charted on the node's detail page:

<Screenshot name="servers" />

* **CPU** utilization
* **Memory** usage
* **Disk** usage
* **Connections**

Per-heartbeat samples are retained for `FLEETDOCK_METRICS_RETENTION` (default `168h`, seven
days).

## Connectivity [#connectivity]

Agents connect **outbound** to the Manager's `FLEETDOCK_PUBLIC_URL`. The Manager does not
open connections back to the agent for operations — the agent polls. For **live database
administration** and **external instances**, however, the control plane connects directly
to the database port; see [Networking](/docs/networking).

## Capacity [#capacity]

Node capacity is the CPU, memory, and disk headroom reported by heartbeats, plus the count
of instances running on the node. Use it to decide where to provision the next instance.

<Callout type="info">
  Fleetdock reports capacity metrics but does not currently enforce a hard capacity limit
  when provisioning. Review a node's headroom before adding instances.
</Callout>

## Node labels and maintenance mode [#node-labels-and-maintenance-mode]

Use maintenance mode to signal that a node is being worked on. Node labels help you group
and identify servers within a larger fleet.

## Removing a node [#removing-a-node]

Uninstall the agent on the server (see [Agent installation](/docs/agents#uninstall)), then
remove the server from the dashboard. Removing a node does not delete data on the host by
itself — uninstall the agent and clean up any provisioned containers you no longer need.

## When a node goes offline [#when-a-node-goes-offline]

If a node stops heartbeating for `FLEETDOCK_HEARTBEAT_TIMEOUT` (default `2m`), the Manager
flips it to `offline`. Operations targeting a managed instance on that node cannot be
executed by its agent until it returns. You can configure alert rules to notify you when a
server goes offline — see [Backups](/docs/backups) and the Notifications page.