Nodes
A node is a Linux machine that runs servers for your network. Each node runs the Atlas worker, which takes its orders from the panel.
Add a node with a command
In the panel, open Settings, Nodes and choose Add node, then Command. The panel makes a join command like this one:
$ curl -fsSL https://panel.example.net/join/<token> | sudo sh Run it as root on the new machine. It needs Linux on x86-64 or ARM64, and it adds Docker when the machine doesn't have it. The token in the command joins one machine and expires after an hour. When it runs out, make a new one.
The node shows up in the list once its worker says hello. If it doesn't, the worker's log says why:
$ journalctl -u atlas-worker -f Add a node over SSH
With the Network plan, the panel can log in and install the worker itself. Choose SSH and enter the host, the port and a user that can use sudo without a password. Atlas makes a new key for you to add to that user's authorized_keys, or takes a private key of your own without a passphrase.
Before it installs anything, the panel shows the machine's fingerprint. Check it against the machine, then choose Trust and install. The same key repairs the worker later, from the node's page.
Players come in through the gate
Every node runs a gate that listens on port 25565 and hands each player to the proxy group their address leads to. Servers and proxies never open ports of their own, so 25565 is the one port players need.
If something else already holds 25565, the node's page says so, and the gate starts once the port is free.
Nodes talk over a private network
Servers on different nodes reach each other over a private network that Atlas sets up. It uses UDP port 51871, which has to be open between your nodes. You can change the port and the address range under Settings, Network.
Keep workers up to date
A panel accepts workers of its own minor version or the one before. A node with an older worker shows Needs update. Update it on that machine:
$ sudo atlas update Take a node out
Run this on the node:
$ sudo atlas uninstall It stops the worker and removes every container, network and image Atlas made on that machine. Docker stays. Add --keep-data to keep the servers' files.