Fleet mode

A Ready Up server runs in one of two modes:

• Standalone. Auto Tournament drives the server over RCON and webhooks: ru match load, ru_webhook_url, ru_heartbeat_url. Admins and skins come from local JSON files. This is how the rest of these pages describe it.
• Fleet mode. The server keeps one outbound WebSocket to the platform. The platform sends matches, commands, the admin list and skins loadouts over it. The server streams match state, events and round backups back. No RCON, no webhooks, and no settings to set by hand on each server after it is enrolled.

Fleet mode is the fleet plugin, readyup/plugins/fleet.so. It is in the Essentials and Full bundles, and stays idle until you give it a platform URL. Without it, or without a platform URL, the server is standalone.

Enroll a server

A server enrolls once. It then keeps its own token and reconnects on its own. There are two ways to enroll.

[fleet]
url=https://tournament.example.com
enroll_code=RUE-7F3K-9QX2-LM4D-P8TW

[fleet]
url=https://tournament.example.com
enroll_key=rfk_...

ru fleet enroll https://tournament.example.com RUE-7F3K-9QX2-LM4D-P8TW

ru fleet status

After enrollment, the server stores its identity in game/csgo/readyup/plugins/fleet/:

• install_id, a random ID written once. A server that enrolls again with the same key gets its old record back instead of a new one.
• credentials.json, with the server ID, the token and the platform URL. Only the file's owner can read it. The token is never written to the log.

A code is spent after one use, so you can remove it from the config. A key can stay. Keep the plugins/fleet/ folder when you update or move the server.

Tokens

• The platform rotates the token every 90 days on its own. The server saves the new one and keeps working.
• If the platform revokes the server, the link closes and the server keeps running standalone. With a key in the config, it enrolls again. With a code, it waits for a new ru fleet enroll. A refused code is never retried.

Connection security

• Only wss:// and https://. The certificate is always checked.
• For a private certificate authority, set ca_file. To pin the platform's public key, set pin_sha256.
• insecure_dev=1 allows ws:// and http://, but only to a loopback or private (RFC 1918) address. For development only.
• The token only goes in the Authorization header, never in a URL.

All [fleet] keys are listed in Configuration.

What the platform controls

Matches

The platform assigns a match to a server. The assignment carries the teams and rosters, the maps and sides, the rules (the ruleset and overrides, round limit, overtime, knife side pick time, tiebreak, technical pauses, force ready, forfeit when a team leaves, the .gg vote), match admins, engine cvars, and a password for this match.

When a match is assigned, the server:

1. ends a running scrim without reporting it, tells players This server was assigned a tournament match. Thanks for playing!, and kicks everyone who is not on the roster, a spectator or an admin after 5 seconds,
2. sets sv_password to the match password,
3. changes to the first map and starts the match warmup.

From there the match runs as described in Scrims and matches. The whitelist and team lock still apply.

The platform can change a running match: add or remove a player, rename a team, change the password or the rules. When the match is over or cancelled, the platform unassigns it. The server clears the password, kicks players and goes back to idle.

A match loaded by hand with ru match load is never reported to the platform. While one is loaded, the server answers new assignments with busy. Do not load local matches on a fleet server.

Admin commands

Platform admins can send these commands to the server. Each one gets exactly one answer.

Admins

The platform keeps one admin list for all servers. The server saves a copy in plugins/match/fleet-admins.json, so it knows its admins even when it starts while the platform is down.

The local admins.json still counts on top of it, so the server owner stays an admin even when the platform sends no list, or a list without them. .ru admins list, add and remove work on the local file. The platform's list is edited on the platform. Match admins from the assignment count too.

Skins

If the server has the skins plugin and skins are turned on for the platform, the platform sends each player's loadout. StatTrak counters live on the platform. See Loadouts from the platform.

Not yet

These are designed but not built yet:

• Server-wide settings from the platform (chat prefix, demo settings, kick delays). Until then, set them on the server.
• Draining a server (finish the current match, take no new one).
• Uploading demos to the platform. Demos stay on disk, or go to ru_demo_upload_url if it is set.
• Failover to another server.

What the server sends

• The match state, on every change. Each change has a number, so the platform can tell if it missed one. A full copy goes out on each reconnect and every 60 seconds.
• Events: ready, pauses, rounds, the knife result, map results, the series end, demos.
• Round backups. CS2 writes a backup file at the start of each round. The server sends it to the platform about 1.5 seconds later, inside the WebSocket.

Scrims and local matches are never reported.

When the platform is down

The server keeps playing on what it already has: the match config, the admin list and loadouts.

• Everything it would have sent is kept on disk, up to 50,000 messages or 64 MB (spool_max_msgs, spool_max_mb). It is sent when the link is back.
• It reconnects on its own, with a growing wait between tries. ru fleet reconnect tries now.
• In warmup and during the knife round, nothing changes. Admins in chat still work.
• When idle, scrims run as usual.

Offline auto-pause

A live match pauses on its own when the platform has been gone for offline_pause_minutes (3 by default). The pause starts at the next freeze time. Players see Paused: server lost contact with the tournament platform. An admin in game can unpause, or the platform can after it is back. /status shows a countdown while the link is down.

Set offline_pause_minutes=0 to turn this off.

Failover

If a server dies during a match, the plan is that the platform notices, proposes a spare server, and an admin confirms. The admin picks which round backup to go back to. Nothing moves on its own.

Today the server side can restore a round backup on command (restore_round), also from a backup the platform sends. Resuming a match on a different server is coming.

Status endpoint

The core serves a small read-only HTTP endpoint. It works in both modes, and with no plugins at all. Monitoring tools, uptime checks and CS2 Server Manager use it. It never talks to the game thread directly, so a slow client cannot slow the server down.

By default it listens on 127.0.0.1, on the game port + 7 (27022 for a server on 27015).

The core runs a selftest on its own 15 seconds after the first map, so /health has a result.

curl -s http://127.0.0.1:27022/status

The summary in /status has flat fields for a table row: mode, phase, map, round, score, series_score, players, match_id, teams, paused and more. Without match.so, it says "match_plugin": "none".

update_safe is true when nothing is lost by restarting or updating the server now: idle, a scrim, practice, or a finished series with no demo upload left. It is false from the moment a real match loads until the series is over.

Token

Requests from the same machine need no token. Anything else must send Authorization: Bearer <token> (or X-ReadyUp-Token, or ?token=). /health never needs one.

If you set status_http_bind to a non-loopback address, you must also set status_http_token (16 characters or more), or the endpoint does not start. The endpoint is plain HTTP. For remote access, put a reverse proxy with TLS in front. The status token is separate from the fleet token.

status.json

At start, the core writes game/csgo/readyup/status.json:

{ "port": 27022, "bind": "127.0.0.1", "token": "rst_...", "pid": 12345, "game_port": 27015, "started_at": 1790340000 }

This is how local tools find the endpoint. If you do not set a token, the generated one is kept in this file and used again after a restart.

Settings are in Configuration.

With CS2 Server Manager (beta)

CS2 Server Manager (CSM) is being extended to link a whole machine to the platform with one command. After that, you create servers from the platform, and each new server enrolls itself with a fleet enrollment key that CSM writes for it. You do not copy codes.

This is in development. There is no final command to show yet, and the machines page in the platform is not released. Until then, enroll each server yourself with ru fleet enroll <url> <code>, or with url, enroll_code or enroll_key in cfg/ReadyUp/fleet.cfg, as above.

CS2 Server Manager

CS2 Server Manager can read each server's status.json and /stream to show a live table of all Ready Up servers, and wait for update_safe before it restarts or updates a server during a match. This is being added to CS2 Server Manager now and is not released yet.

Later, CS2 Server Manager will also enroll with the platform as a host agent. The platform can then start, stop, restart and update servers through it, and write the fleet key into new servers so they enroll on their own.