Setting up SS14.Watchdog

SS14.Watchdog (codename Ian) is our server-hosting wrapper thing, similar to TGS for BYOND (but much simpler for the time being). It handles auto updates, monitoring, automatic restarts, and administration. We recommend you use this for proper deployments.

Setup Process

1. Check Prerequisites

You need to have:

• .NET 8 SDK
• ASP .NET Core 8 Runtime

Both of these can be found at the .NET 8 download page.

On Linux use your favourite package manager (apt, dnf, pacman, brew etc) according to Microsoft’s installation instructions.

2. Build

The following set of commands should build the Watchdog on a Linux system. You’ll have to adjust it according to your actual system, of course.

# Download the SS14.Watchdog repository and any submodules/etc.
git clone --recursive https://github.com/space-wizards/SS14.Watchdog

# Switch into the SS14.Watchdog directory.
cd SS14.Watchdog

# Build the Watchdog.
# The result is placed into: SS14.Watchdog/bin/Release/net8.0/linux-x64/publish
dotnet publish -c Release -r linux-x64 --no-self-contained

The contents of SS14.Watchdog/bin/Release/net8.0/linux-x64/publish can then be copied to some other place. You will continue your work here.

3. Run

Assuming you’ve followed the structure laid out above, you simply need to have a terminal in the folder you copied above, and run the SS14.Watchdog executable.

Watchdog Configuration

Watchdog’s config file is appsettings.yml

The watchdog configuration is split into two major sections:

• Global elements, shared across all instances (servers).
• Per-instance elements.

Serilog, AllowedHosts

These shouldn’t usually need to be changed, and are too complex to describe here.

BaseUrl

This represents the external URL of the Watchdog. This is passed to the instances automatically so that they can check in with the Watchdog. It’s also used in update modes which require clients to connect to the Watchdog for assets.

# Usually what you want, unless being used for client ZIPs.
BaseUrl: "http://localhost:5000/"

Urls

This controls on which interfaces the Watchdog hosts, which may be important in some cases. In particular, this can be used to expose the Watchdog outside of localhost without a reverse proxy, as so:

Urls: "http://*:5000"

See the relevant documentation for more details: docs.microsoft.com

Be sure to adjust BaseUrl accordingly!

Notifications

You can now set a notification discord webhook to receive notifications whenever a server crashes, the integration is as simple as adding the following to your config.

Notification:
  DiscordWebhook: "https://discord.com/api/webhooks/..."

Instances

Each instance is a separate game server, so the terms “instance” and “server” can be used semi-interchangably.

Servers:
  Instances:
    example:
      # This is intended to be the "human" name of the instance.
      # In practice, this is occasionally used in logging.
      # It doesn't affect game.hostname, for example.
      Name: "Example"

      # This is the API token for external access.
      # (As opposed to the API token used internally between the Watchdog and game server.
      #  That is randomly generated.)
      ApiToken: "you should choose a better token"

      # Somewhat misleadingly, this is the port of the game server on localhost.
      # This will NOT be automatically synchronized with the real port in the config.
      ApiPort: 1212

      # The update type and further options control where the server software is acquired from.
      # This example is for the official server builds.
      UpdateType: "Manifest"
      Updates:
        ManifestUrl: "https://wizards.cdn.spacestation14.com/fork/wizards/manifest"

      # The server is expected to ping the Watchdog occastionally.
      # (The aforementioned BaseUrl is passed to the server to facilitate this.)
      # This confirms the server has not, say, crashed.
      # If it has crashed, the server is forcibly restarted.
      # However, startup can a bit of a long process on some systems.
      TimeoutSeconds: 60
      # If enabled, if a freeze occurs, data on the state of the server is saved for analysis.
      # DumpOnTimeout: true
      # TimeoutDumpType controls how this is set, but I'm not sure on the details.

      # The program used to run the server can be set from here.
      # Note that realistically, this shouldn't need to be changed unless you are:
      # A. Trying to perform more advanced diagnostics (i.e. attaching a debugger)
      # B. Doing something very different from running a Space Station 14 server
      # RunCommand: "./wrapper.sh"

      # Environment variables can be set from here.
      # See for example Performance Tweaks in the Server Operator's Handbook.
      # EnvironmentVariables:
      #   ROBUST_NUMERICS_AVX: "true"

Server Instance Folder

The watchdog will automatically create a folder structure for each server instance. This is at instances/<instanceId>, e.g. instances/wizards_den / instances/wizards_den_two, relative to the current working directory when executing the watchdog. In the example config above this would be instances/example

Each instance folder has the following files and folders:

• binaries/: Is used to store client binaries when using the “Local” update type, see below.
• bin/: Contains the actual extracted server binaries.
• data/: Stores server data like player preferences.
• config.toml: Is the config file the server will load (the watchdog overrides the default location, server_config.toml next to the .exe, to avoid it getting deleted when the server resets). You may have to create this file manually the first time.
• data.json: Contains watchdog information. If you changed the update type and are getting errors, delete this.

Controlling The Watchdog

There are two key situations when the watchdog needs to be controlled.

Firstly, the watchdog will only update when it is either explicitly notified to check for an update, or when it is restarted.

Secondly, you may want to simply force a server to be restarted.

Lastly, you may want to shutdown the server when the round ends. Example for maintenance.

These tasks can be achieved with the following commands:

curl -v -X POST -u myInstance:ApiToken http://localhost:5000/instances/myInstance/restart

curl -v -X POST -u myInstance:ApiToken http://localhost:5000/instances/myInstance/update

curl -v -X POST -u myInstance:ApiToken http://localhost:5000/instances/myInstance/stop

Update Types

Manifest Update

Manifest type is the method all Wizard’s Den servers use and we recommend using. Manifest is required to be able to record replays

Servers:
  Instances:
    example:
      # (This is an example, do NOT blindly copy paste this. Or you may end up with an unfinished configuration)
      UpdateType: "Manifest"
      Updates:
        ManifestUrl: "https://wizards.cdn.spacestation14.com/fork/wizards/manifest"

Git-based Updates

SS14.Watchdog can compile and update the server when commits are pushed to a branch of the Git repository that contains the source to your fork.

Servers:
  Instances:
    example:
      # (This is an example, do NOT blindly copy paste this. Or you may end up with an unfinished configuration)
      UpdateType: "Git"
      Updates:
        # BaseUrl: The URL of the Git repository to watch.
        # This is distinct from the Watchdog-wide BaseUrl.
        BaseUrl: "https://github.com/moonheart08/outer-rim-14/"
        # Branch: The branch to watch.
        Branch: "master"
        # Hybrid ACZ: When enabled, the game server hosts the client zip rather than the watchdog.
        # As of the introduction of delta updating this is now the better way to handle this.
        HybridACZ: true

Jenkins Updates

This is an ancient method, but it should still work.

Servers:
  Instances:
    example:
      # (This is an example, do NOT blindly copy paste this. Or you may end up with an unfinished configuration)
      UpdateType: "Jenkins"
      Updates:
        BaseUrl: "http://localhost:9938"
        JobName: "Star"

“Dummy” Update Provider

The “Dummy” update provider will fake an update whenever it is queried, and otherwise simply assume that a server has already been extracted to bin/.

As the Watchdog does not automatically periodically check for updates, the fake updates shouldn’t get in the way.

To configure this, use the following update configuration in your appsettings.yml, in the entry for your server instance:

Servers:
  Instances:
    example:
      # (This is an example, do NOT blindly copy paste this. Or you may end up with an unfinished configuration)
      UpdateType: "Dummy"

Custom Auto Updates

Not supported, but it should be relatively trivial to edit the code for SS14.Watchdog to add support for whatever update mechanism you need. See UpdateProvider.cs.

Update Types (DIY Edition)

“Dummy” Update Provider, DIY Updates Edition

Before trying this, ensure you’re familiar with how to use the “Dummy” Update Provider in general.

It is relatively simple to configure things in such a way as to allow for updates. The following instructions are for Unix-likes, but the idea should be comprehensible in any case.

Start with configuring as so:

Servers:
  Instances:
    example:
      # (This is an example, do NOT blindly copy paste this. Or you may end up with an unfinished configuration)
      UpdateType: "Dummy"
      RunCommand: "./currentServer"

The basic premise of this mechanism is to use symbolic links to switch over which directory is actually used.

As such, three scripts, switchTo, switchTo1 and switchTo2, are needed:

switchTo:

#!/bin/sh
rm currentServer switch inactiveBin ; mkdir -p $1
ln -s $1/Robust.Server currentServer ; ln -s $2 switch ; ln -s $3 inactiveBin
echo Switching to $1
curl -v -X POST -u myInstance:ApiToken http://localhost:5000/instances/myInstance/update

switchTo1:

#!/bin/sh
./switchTo bin1 switchTo2 bin2

switchTo2:

#!/bin/sh
./switchTo bin2 switchTo1 bin1

Once these are made executable (chmod +x switchTo*), and one of them is run, running ./switch after that will toggle between the two directories.

As such, the workflow is to remove everything in inactiveBin, then extract a new server there, then run ./switch to confirm it.

Before clearing and extracting a new server build to inactiveBin, be sure to make sure the server has actually restarted from any previous update and is actually no longer using that directory.

DIY Manifest Server

This is a quick script useful in setting up a DIY server for the Manifest update type described in the manifest section.

It assumes you have some arbitrary static HTTP server, and you just need a script to output the JSON with an updated date (so you can just transfer two files to said static HTTP server and trigger an update).

import json, datetime
nowish = datetime.datetime.now().isoformat()
print(json.dumps({"builds":{nowish: {"time": nowish, "client": {"url": "", "sha256": ""}, "server": {"linux-x64": {"url": "http://localhost:9283/SS14.Server_linux-x64.zip", "sha256": ""}}}}}))

Systemd service

To allow watchdog to run in the background and automatically start up with the server, you can make a service file. It will look something like this.

Of course, configure it to the actual directory of your watchdog.

If your distro does not use systemd as it’s init then you will have to convert this to your relevent init.

/etc/systemd/system/SS14.Watchdog.service

[Unit]
Description=SS14 Watchdog
After=network.target

[Service]
ExecStart=/path/to/SS14.Watchdog
WorkingDirectory=/path/to
Restart=always
# This is used for git method to not fail instantly.
Environment="DOTNET_CLI_HOME=/tmp"

[Install]
WantedBy=default.target

Now reload your systemd daemon and enable the service as you normally would.

# Reload the systemd daemon (required when making a new service file)
systemctl daemon-reload

# Start Watchdog service in the background
systemctl start SS14.Watchdog

# Enable the Watchdog service so it will start on system startup.
systemctl enable SS14.Watchdog

If you are not already aware of how to use systemctl now would be a good time.

To view logs you can use journalctl from now on.

General Troubleshooting

Server keeps restarting every 30 seconds

This means the server isn’t communicating with the watchdog correctly and the watchdog is forced to assume that the server is locked up or similar. This happens if BaseUrl in the watchdog configuration is set incorrectly or otherwise inaccessible by the game server.

System.IO.FileNotFoundException: Could not load file or assembly 'Mono.Posix.NETStandard, Version=1.0.0.0, Culture=neutral (…)

Current working theory is that this is caused by improper dotnet publish options. The below set of test results should help explain.

dotnet publish -c Release -r linux-x64 --no-self-contained SS14.Watchdog -o test
 RESULT: Mono.Posix.NETStandard.dll included, System.dll not included (as expected)

dotnet publish -c Release -r linux-x64 SS14.Watchdog -o test
 RESULT: Mono.Posix.NETStandard.dll included, System.dll included

dotnet publish -c Release SS14.Watchdog -o test
 RESULT: Mono.Posix.NETStandard.dll not included, System.dll not included

Since Watchdog uses Mono.Posix.NETStandard.dll to mark executables as executable on Linux and Mac OS X, it’s important to have it around on those OSes.

Old Example Config

(This section has been extracted from an older version of this guide with slight modification in case it is still of use. It may be outdated.)

Example from our official servers (obviously tokens redacted):

Serilog:
  Using: [ "Serilog.Sinks.Console", "Serilog.Sinks.Loki" ]
  MinimumLevel:
    Default: Information
    Override:
      SS14: Information
      Microsoft: "Warning"
      Microsoft.Hosting.Lifetime: "Information"
      Microsoft.AspNetCore: Warning

  WriteTo:
    - Name: Console
      Args:
        OutputTemplate: "[{Timestamp:HH:mm:ss} {Level:u3} {SourceContext}] {Message:lj}{NewLine}{Exception}"

  Enrich: [ "FromLogContext" ]

  # Uncomment to have watchdog log to Loki
  #Loki:
  #  Address: "{{ loki_addr }}"
  #  Name: "{{ server_id }}"
  #  Username: "{{ loki_user }}"
  #  password: "{{ loki_pass }}"

AllowedHosts: "*"

Notification:
  DiscordWebhook: "https://discord.com/api/webhooks/..."

# API URL that your watchdog is accessible from.
# This HAS to be set so the game servers can communicate with the watchdog.
# If you don't want the watchdog to be publically accessible, do `http://localhost:5000/` here.
BaseUrl: https://builds.spacestation14.io/watchdog/

Servers:
  Instances:
    # ID of your server.
    wizards_den:
      # Name of the server
      Name: "Wizard's Den"
      ApiToken: "foobar" # API token to control this instance remotely like run updates, restart server.
      ApiPort: 1212 # API port OF THE GAME SERVER. This has to match the 1212 HTTP status API (described below). Otherwise the watchdog can't contact the game server for stuff.
      
      # Auto update configuration. This can be left out if you do not need auto updates. Example is for our officially hosted builds.
      # See above for alternatives.
      UpdateType: "Manifest"
      Updates:
        ManifestUrl: "https://wizards.cdn.spacestation14.com/fork/wizards/manifest"
        
      # Any environment variables you may want to specify.
      EnvironmentVariables:
        Foo: bar