NodeFlare/docs

NodeFlare

NodeFlare is a lightweight, self-hosted server monitoring panel: view server status in a web UI while a small agent collects and pushes telemetry. It supports live metrics, TCP/ICMP probing, Telegram alerts, remote command execution, TOTP two-factor authentication, theme customization, and SQLite / PostgreSQL backup, restore, and online migration.

Chinese: README.md

• Screenshots
• Features
• Supported platforms
• Install the server
• Install the agent
• Metrics and sampling
• Configuration
• Default paths and logs
• Reverse proxy
• Database and backups
• Development
• Troubleshooting

Screenshots

Features

Monitoring

• CPU, memory, and network speed are sampled every second and uploaded in compressed batches every 3 seconds by default; slower metrics such as disks, GPU, and connection counts are cached separately
• Latency probing: TCP and ICMP tasks assigned per node, with per-carrier (China Telecom / Mobile / Unicom) display
• Public dashboard in Simplified Chinese and English; site name, announcement, logo, background image, and visible sections are configurable
• Theme store: built-in themes, one-click install from the repository, local ZIP upload, and per-theme settings

Alerts

• Resource threshold rules for CPU / memory / disk / inbound / outbound, using averages or continuous breach with a configurable duration
• Offline, expiry, and traffic alerts delivered through Telegram with a customizable message template

Management

• Node grouping, tags, region, billing cycle, price, expiry date, traffic quota, and reset day
• Remote command execution: requires TOTP; a command runs for up to 10 minutes and is not cancelled when you leave the page
• Daily exchange-rate snapshot for multi-currency pricing

Security

• TOTP two-factor authentication; Cloudflare Turnstile (can protect admin login and the public dashboard independently)
• Login rate limiting, session management (list login devices and revoke them), configurable session lifetime
• Listens on 127.0.0.1 by default; the agent connects outbound, so no inbound port is required; the agent token can be passed as a file or an environment variable

Data

• SQLite and PostgreSQL, with online migration between them
• One-click ZIP backup / restore, space reclamation, and automatic cleanup of history by retention days

Supported platforms

Role	Platform	Architecture
Server	Linux	x64 / ARM64
Server	Windows	x64
Server	macOS	ARM64
Server	FreeBSD 13+	x64 / ARM64
Agent	Linux / Windows / macOS / FreeBSD	Same as above (macOS is Apple Silicon only)

The installer detects the service manager (systemd / OpenRC / launchd / FreeBSD rc / Windows scheduled task) and registers the panel to start on boot.

Install the server

The installer downloads the latest release and registers the system service.

Linux / macOS:

curl -fsSL https://raw.githubusercontent.com/elysia62/NodeFlare/main/install.sh | sudo sh

FreeBSD:

fetch -qo - https://raw.githubusercontent.com/elysia62/NodeFlare/main/install.sh | sudo sh

Windows PowerShell (administrator):

Invoke-WebRequest -UseBasicParsing https://raw.githubusercontent.com/elysia62/NodeFlare/main/install.ps1 -OutFile "$env:TEMP\nodeflare-install.ps1"
Unblock-File "$env:TEMP\nodeflare-install.ps1"
& "$env:TEMP\nodeflare-install.ps1"

The first install asks for the admin username, password, listening port (default 2206), and database URL (SQLite by default). The panel listens on 127.0.0.1:2206, so open http://127.0.0.1:2206/admin/login locally; public access requires an HTTPS reverse proxy, see Reverse proxy.

The admin password is only needed to initialize the database for the first time; it is cleared from the configuration afterwards.

Installer options

Command	Description
sudo sh install.sh	Interactive menu
sudo sh install.sh --install	Install or update
sudo sh install.sh --status	Show service status
sudo sh install.sh --restart	Restart the service
sudo sh install.sh --uninstall	Uninstall, keeping configuration and data
sudo sh install.sh --uninstall --purge	Uninstall and delete configuration and data

On Windows the equivalent options are -Install / -Status / -Restart / -Uninstall [-Purge].

To update, run the installer again; configuration and data are preserved. Installs and updates verify the release digest and roll back to the previous version on failure.

Install the agent

Create a node on the "Servers" page in the admin panel and run the command shown in the dialog. The agent connects outbound to the panel, so no inbound port is required.

Linux:

curl -fsSL https://raw.githubusercontent.com/elysia62/NodeFlare/main/agent/agent.sh \
  | sudo sh -s -- -e 'https://nodeflare.example.com' -t 'Agent Token'

Per-platform scripts:

Platform	Script	Service
Linux	agent/agent.sh	systemd / OpenRC
macOS (Apple Silicon)	agent/install-macos.sh	launchd
FreeBSD	agent/install-freebsd.sh	rc.d
Windows	agent/install.ps1	Scheduled task

Agent options

Option	Description
-e	NodeFlare endpoint (required)
-t	Agent token (required)
-i	Initial history interval in seconds, 15–3600 (default 60)
-m	Optional GitHub download mirror prefix, e.g. https://ghproxy.net
--update	Update the agent, reusing the saved endpoint and token, verifying the digest and rolling back on failure
--status	Show agent status
--uninstall	Uninstall the agent

On Windows the equivalents are -Endpoint / -Token / -Interval / -Mirror and -Update / -Status / -Uninstall.

On Linux with systemd, the installer writes the token directly into the service unit as Environment=NODEFLARE_AGENT_TOKEN=.... The --update command reads the endpoint, token, and history interval from that service configuration.

Metrics and sampling

CPU, memory, and network speed are sampled every second and uploaded in compressed batches every 3 seconds by default. Server cards show one real sample per second, giving roughly 2–3 seconds of display latency on a healthy connection; after a reconnect, backlogged samples are skipped. Slower metrics such as disk capacity and GPU are cached separately, and history is aggregated by each node's configured interval. A longer real-time upload interval reduces how often cards update.

Linux memory accounting follows the Komari convention: used memory is MemTotal - MemFree - Cached - SReclaimable - Buffers + Shmem, and used swap subtracts SwapCached. File cache is not counted as used memory and shared memory is; the panel reports whole-machine memory, not NodeFlare's own usage.

Configuration

The configuration file location is listed in Default paths and logs; a full example is in backend/config.example.toml.

Key	Description
database_url	sqlite://nodeflare.db or a PostgreSQL URL such as postgres://user:password@127.0.0.1:5432/nodeflare?sslmode=disable
bind_addr	Listen address, 127.0.0.1:2206 by default
admin_username	Admin username
admin_password	Admin password (8–128 characters), only needed for the first database initialization and cleared afterwards
trusted_proxies	Trusted reverse proxy IP / CIDR list; only X-Forwarded-For from these networks is trusted
turnstile_site_key / turnstile_secret_key	Turnstile keys; leave empty to disable
session_ttl_hours	Session lifetime in hours (1–2160), 168 by default
frontend_dir / admin_frontend_dir	Frontend asset directories; relative paths resolve against the configuration file
theme_dir	Theme extraction directory, themes under the configuration directory by default

A few settings can be overridden on the command line: nodeflare --config <path> --bind <address> --database <url>. When started with --database, the panel's online migration feature is disabled.

Default paths and logs

Platform	Program	Configuration and data
Linux	/opt/nodeflare	/etc/nodeflare
Windows	%ProgramFiles%\NodeFlare	%ProgramData%\NodeFlare\Server
macOS	/usr/local/libexec/nodeflare	/Library/Application Support/NodeFlare/Server
FreeBSD	/usr/local/libexec/nodeflare	/var/db/nodeflare/server

Agent on Linux: program at /opt/nodeflare/agent, configuration and state in /etc/nodeflare/agent. The SQLite file lives in the configuration directory.

Logs:

• Linux (systemd): journalctl -u nodeflare -f; the agent is journalctl -u nodeflare-agent -f
• Linux (OpenRC): rc-service nodeflare status
• macOS: /var/log/nodeflare.log
• Windows: Get-ScheduledTaskInfo -TaskName nodeflare

Reverse proxy

Expose the panel through an HTTPS reverse proxy and add the proxy address to trusted_proxies, otherwise every visitor is treated as the same IP and the session cookie is not marked Secure:

trusted_proxies = ["127.0.0.1/32", "::1/128"]

nginx:

server {
    listen 443 ssl;
    server_name nodeflare.example.com;

    ssl_certificate     /etc/letsencrypt/live/nodeflare.example.com/fullchain.pem;
    ssl_certificate_key /etc/letsencrypt/live/nodeflare.example.com/privkey.pem;

    location / {
        proxy_pass http://127.0.0.1:2206;
        proxy_http_version 1.1;
        proxy_set_header Upgrade $http_upgrade;   # WebSocket live data
        proxy_set_header Connection "upgrade";
        proxy_set_header Host $host;
        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
        proxy_set_header X-Forwarded-Proto $scheme;
    }
}

Caddy:

nodeflare.example.com {
    reverse_proxy 127.0.0.1:2206
}

Database and backups

The "Database" page shows usage, reclaims space, exports and restores ZIP backups, and migrates between SQLite and PostgreSQL. A backup contains settings, nodes, history, notifications, themes, tasks, and security configuration, and requires TOTP or the password. Limits are a 512 MiB ZIP, 4 GiB extracted, 16,384 entries, and 32 MiB per theme file. Migration overwrites the target database, updates the connection string automatically, keeps agent tokens, and takes effect after a restart. History is kept for 30 days by default.

Development

git clone https://github.com/elysia62/NodeFlare.git
cd NodeFlare
bun install --frozen-lockfile
cp backend/config.example.toml backend/config.toml
./dev.sh

dev.sh builds the frontend and starts the backend with cargo run. start.sh builds release binaries and prefers /etc/nodeflare/config.toml, which can be overridden with NODEFLARE_CONFIG.

Repository layout:

Directory	Contents
backend/	Server (Rust / Axum / SQLx): src/routes endpoints, src/db data layer, src/websocket live channels, migrations/ schema
agent/	Agent source (collection / upload / remote execution / self-update) and per-platform install scripts
shared/	Telemetry protocol shared by the agent and the server (serialization + compression)
frontend/	Frontend (React + Vite): src/components, src/styles
scripts/	Build, version resolution, and smoke-test scripts
docs/	English README and systemd service unit

Tests and builds:

bun test --cwd frontend
cargo test --locked --manifest-path backend/Cargo.toml
cargo test --locked --manifest-path agent/Cargo.toml
bun run build

Smoke test (start the panel first):

MONITOR_ADMIN_USERNAME=admin MONITOR_ADMIN_PASSWORD='your-password' bun run test:smoke

Troubleshooting

• The service fails to start: check the logs (see Default paths and logs); usually the port is in use or the database URL is wrong.
• Unknown listening port: check bind_addr in the configuration file, 127.0.0.1:2206 by default.
• The agent shows offline: verify the agent can reach the panel (curl -I https://your-panel), check the token, and check the system clock, since reporting and latency rely on clock calibration.
• Nodes are missing from the public dashboard: make sure the node is not hidden and that "Public dashboard" is enabled in site settings.
• Backup export reports the size limit: shorten the history retention or clear history, then export again.

License

MIT