Huddle

Huddle is an auto-deployment file and application management tool designed to work well in auto-scaling environments in which the end number of servers or clients is unknown. This includes auto-scaling web servers and IoT devices, amongst others.

Applications are controlled through huddle using configuration files which are written in .json ir .ini format. Each file or application to be controlled will have its own configuration file, which may be as simple or as complex as the user wishes.

An example configuration in JSON:

{
  "repository": {
    "remote": "origin",
    "remote path": "https://github.com/slightlynybbled/dummy.git",
    "local path": "C:/_code/_git_example",
    "branch": "master",
    "executable": "C:/Program Files/Git/bin/git.exe"
  },

  "timing": {
    "minimum": 10,
    "maximum": 20
  },

  "application": {
    "start": "python -m dummy_app.py"
  }
}

The same configuration as an INI file:

[repository]
remote = origin
remote path: https://github.com/slightlynybbled/dummy.git
local path: C:/_code/_git_example
branch: master
executable: C:/Program Files/Git/bin/git.exe

[timing]
minimum = 10
maximum = 20

[application]
start = python -m dummy_app.py

Motivation

It would appear that most auto-deployment tools are focused on using git hooks and use a 'push' model to deploy from Git to various nodes. In the simplest case, this involves knowing the IP address or similar information for each server. In the most complex cases, this involves a coordinator and workers testing, deploying, and pushing to the servers.

In an environment of unknown scale - such as auto-scaling web servers and IoT devices, the devices themselves need to be intelligent enough to self-deploy. This is where huddle comes in. Each device takes charge of its self and does a pull as the git repository is updated. This has the advantage of not requiring any sort of global registry or count of devices and scales very well.

Capabilities

Huddle will:

• periodically sync the local application branch with the remote repository
• stop/start the local application as required in order to update
• perform health checks using http or via raw sockets, restarting your application if necessary
• perform all of the above on any number of applications

In addition, huddle will pull from any number of git repositories and initiate any number of applications. Huddle only requires one configuration script per application.

Server Requirements

In auto-scaling environments, typically an initial image is created for the server or IoT device which contains:

• python environment
• huddle installation
• appropriate huddle configuration script(s)
• huddle loaded at startup (probably @restart cron job)

Each server then manages its own application suite!