Installation
Requirements
Supported systems, the ports to open, and what the master and the agents need.
Operating system
| Supported | |
|---|---|
| Systems | Any Linux with systemd: Ubuntu, Debian, AlmaLinux, Rocky Linux, RHEL, CentOS Stream and others |
| CPU | x86_64 (amd64), arm64, armv7 or 386 |
| Access | root, or a user with sudo, to install |
The same requirements apply to the master and to the agents. The release package contains the program for all four architectures, and the installer picks the right one.
Master
- A TCP port that browsers and agents can reach: 8899 by default (change it with
--port). If you prefer to open the panel through a domain such ashttps://monitor.example.com, you also need Nginx or Apache on the master (see HTTPS). - Nothing else to install: the master has its own web server and an embedded SQLite database. It doesn't use PHP, MySQL or a web server, so it keeps working, and sends its alerts, when those services fail on the same machine.
- It runs as its own unprivileged user,
3rmonitor. - Outbound internet access for alert emails (to your SMTP server), the vulnerability databases, the monthly update of the IP-to-country database, update checks on 3rsoft.com.pa, the addresses of your monitors and, if you use AI analysis, your model provider.
Where to put the master
If the master's own server goes down, nothing is left to send the alert. When you can, install the master on a server that doesn't share its fate with your most important services, for example a small server at another provider or in another data center.
Agents
- Outbound HTTP(S) access to the master, at the panel address. The agent opens no ports and doesn't need to be reachable from anywhere.
curlorwgetfor the installation.- The agent runs as the unprivileged user
3rmonitor-agent. The few things that need root (the processes behind listening ports, certificates, software and web server checks, Docker, the blocklist firewall, updates and AI operations) go through a small root helper,3rmonitor-helper, on the same server. The installer sets both up. - Little bandwidth: about 58 MB a day with the default 3-second interval. See Bandwidth usage.
Optional, depending on the features you use:
| Feature | What the server needs |
|---|---|
| Traffic statistics | Apache or Nginx access logs. The usual locations (Debian, RHEL, aaPanel, cPanel, Plesk, XAMPP) are found automatically. |
| Blocklist / allowlist | nftables, or ipset + iptables |
| MySQL / MariaDB status | A credentials file in mysql_defaults_file. On Debian, /etc/mysql/debian.cnf is used automatically. |
| Vulnerability checks of system packages | Debian, Ubuntu, AlmaLinux, Rocky Linux or Alpine. On other systems the main services and the project dependencies are still checked. |
| Sandboxed AI diagnostics | systemd 236 or newer and D-Bus. Without them, the AI on-call only runs its fixed checks there. |
Ports
| Port | Direction | Used for |
|---|---|---|
8899/tcp (or your --port) | Into the master | The panel in the browser, and agent reports. The installer opens it in ufw or firewalld (unless you use --no-firewall). |
| 80 and 443/tcp | Into the master | Only with --domain / --https through Nginx or Apache. Port 80 must be reachable from the internet while Let's Encrypt issues the certificate. |
| 465 or 587/tcp | Out of the master | SMTP, for alert emails |
| 443/tcp | Out of the master | Vulnerability databases, the IP-to-country database, update checks and, with AI analysis, your model provider |
| None | Into the agents | Agents only connect out, to the master |
Many cloud providers have their own firewall (security groups, cloud firewall) in front of the server. Open port 8899 there too, or the panel won't load even though the installer opened it on the server.
Optional
- An SMTP account for alert emails. Gmail works with an "app password".
- A domain and Nginx or Apache on the master, for a free Let's Encrypt certificate with
install.sh --https. HTTPS is also needed for passkeys, which sign the AI repairs you approve. - An API key from a model provider (DeepSeek, OpenAI, Anthropic, Google, Qwen…), or a local Ollama, for AI analysis.
Next: install the master.