redcraft_server_management

README

RedCraft Server Management (rcsm)

This Golang project was made to manage Minecraft servers on Linux, with efficiency and scalability in mind.

Requirements to run

rcsm requires tmux to be installed and in the PATH to work.

How to install

• Download the latest release of rcsm from the GitHub releases as rcsm
• Set it as runnable with chmod +x rcsm
• Download .env.example as .env and edit the settings
• Edit .env to make sure it's set as you want
• Start rcsm with ./rcsm

Development

Installation

Run go install, it should install all the dependencies from go.mod.

Start the project

rcsm requires tmux, golang 1.14 or later to develop for, and that's pretty much it!

You can start the dev server by running scripts/test.sh and compile the project using scripts/build.sh.

Features

rcsm was designed to automatically handle server updates from S3, alerts on Discord, and a way to communicate between services with little to no configuration using Redis.

Once set up, you shouldn't have anything to change.

Config

rcsm uses env files that are .ini compliant. By default, rcsm will look for a .env file, it's recommended to copy the .env.example to .env and then edit it to suit your needs.

Environment variables

You can also override some parameters using env variables. For example, you can share one config file with all your servers and override INSTANCE_NAME and MINECRAFT_SERVERS_TO_CREATE by running rcsm like this: INSTANCE_NAME=survival MINECRAFT_SERVERS_TO_CREATE=survival rcsm Environment variables can also be used with systemd scripts.

Custom env file location

If you install rcsm in your path such as in /usr/bin/rcsm, you'll want to store your configuration somewhere else like in /etc/rcsm.

You can in fact do this using the ENV_FILE variable, like ENV_FILE=/etc/rcsm /usr/bin/rcsm (or using systemd files of course)

Handling of servers

rcsm was primarily built to handle Minecraft servers, and it has multiple features and configuration for this.

Servers are started using tmux to be able to re-attach the console and manually take direct control of the server without rcsm.

tmux session prefixes can be changed with MINECRAFT_TMUX_SESSION_PREFIX (default is rcsm_)

S3 templates

rcsm was built primarily because other solutions didn't have any -good- solution for plugin updates.

At RedCraft.org, we use server templates that generate a usable .tar archives that can be restored on a server.

This is very useful if you use cron tasks with a tool to update a plugin repository such as our plugin updater.

If S3_ENABLED is set to true, then when a server starts (including automated restart), S3 will be used to check for templates in the specified S3 bucket.

Please notice that you can also use a 3rd party S3 compatible provider, such as Scaleway Object Storage (in fact that's what we use) or even host it yourself by changing S3_ENDPOINT.

⚠ ⚠ ⚠ If you store data in your plugin folders, S3 templates might delete or overwrite them! Please use plugins that use external databases to avoid this issue.

Server config

When rcsm starts, it will do a discovery of the folder specified by MINECRAFT_SERVERS_DIRECTORY. For every server, it will try to read a rcsm_config.json file that contains the following configuration:

• start_command to specify Java flags such as memory usage. By default, it's set to use 6 GB of memory and uses these flags. ⚠ By default the command is made to run server.jar
• stop_command which is the command to gracefully stop the server, by default it's stop but for BungeeCord you'll have to set it to end for example.

Auto start/stop and "health checks"

rcsm was also built for high availability in mind.

Therefore, a few settings are dedicated to auto restart/stop/etc.

Auto start/stop

By default, rcsm will start every server specified in MINECRAFT_SERVERS_DIRECTORY when it starts, if not already started. This behavior can be disabled by setting AUTO_START_ON_BOOT to false.

That's right, rcsm isn't technically a wrapper because servers will continue to run even if rcsm is closed by default, but you can specify AUTO_STOP_ON_CLOSE to true to stop them on SIGINT (regular kill or ctrl + c)

Health checks

By default, rcsm checks if servers stops and restart them if they stop on their own or even if someone uses /stop from the console or in-game. If you want to stop a server, you'll have to send a redis pub command (cf down bellow for Redis help).

This behavior can be disabled by setting AUTO_RESTART_CRASH_ENABLED to false.

Also, if a server fails to reboot 3 times in under 2 minutes, the server will be marked as crashed and rcsm won't attempt to restart it automatically. These values can be changed with AUTO_RESTART_CRASH_MAX_TRIES and AUTO_RESTART_CRASH_TIMEOUT_SEC

Webhooks

rcsm has support for webhooks, more specifically for Discord webhooks.

You can enable this feature by setting WEBHOOKS_ENABLED to true.

Make sure to add a webhook integration on the desired Discord channel and set the endpoint on rcsm using WEBHOOKS_ENDPOINT.

Here's how the formatting looks:

Redis

Redis is a cache database, but a very interesting feature added years ago is the pub/sub feature that rcsm supports.

You can enable this feature by setting REDIS_ENABLED to true.

Basically, on any log, rcsm will publish a message on a channel (set with REDIS_PUB_SUB_CHANNEL) and other parts of your infrastructure can subscribe to this channel to get messages.

rcsm will also listen for commands on the channel.

Command format for rcsm

rcsm will listen on the pub/sub channel for JSON formats using the following fields:

• target (can be a server name or * for all servers)
• action (can be start/stop/restart or command)
• content (used only for command for now, it's the command to run in the console)

Please notice that:

• rcsm works well with UTF-8 characters, you can even send unicode characters in commands
• rcsm won't send you back the response for a command, but will acknowledge via an event

Examples

Restarting the test1 server:

{
    "target": "test1",
    "action": "restart"
}

Running /op lululombard on the test2 server:

{
    "target": "test2",
    "action": "command",
    "content": "op lululombard"
}

Format of the logs sent by rcsm

rcsm will publish logs with the following format:

• level (can be debug/info/warn/severe/fatal)
• instance (it's the instance name, by default server and can be changed with INSTANCE_NAME)
• service (it's the server name or any of the components like redis, healthcheck, setup, updater or rcsm)
• message (it's the log message)

Example:

{
    "level": "SEVERE",
    "instance": "server",
    "service": "test1",
    "message": "Could not start: Server crashed on start, check server logs"
}

Auto update of rcsm

By default, rcsm will check for updates and auto update itself.

You can disable auto updates when setting AUTO_UPDATE_ENABLED to false.

Here's how the process works:

• rcsm checks for update on the GitHub repo at boot and every hour
• if an update is available, it will download it and replace itself with the new binary
• if EXIT_ON_AUTO_UPDATE is set, rcsm will quit, otherwise the update will be applied at the next restart of rcsm

⚠ EXIT_ON_AUTO_UPDATE will NOT restart rcsm, it will only exit it. That's useful if you're using it as a system daemon and you automatically restart it on quit.