miton18 / cerebro

Infrastructure automation made fast

cerebro

Infrastructure automation made fast

How it works

cerebro use a client-server model, an agent is installed on every host and a main server is started somewhere. When an agent start, it generate a secret key and try to contact his master, use mTLS

Cerebro server can easily perform a first SSH connection in order to install and put his IP on the target host for Cerebro agent configuration. Agent and Server us gRPC.

environment data are store on the server under environment/ directory, every environment has 2 informations kind /environment/prod/host relative to host scoped informations and /environment/prod/roles which holds informations about a group of server (webs ones, databases or proxy ones) (thx Ansible).

Every thing you need to do to configure your hosts is called actions an action must be idempotent, because the server can send several time the same action to the host.

Actions are groupped into roles, roles match a group of hosts and apply a collection of action

Actions

Every action must fill the following spec, see the first examples

Name	Version	Help	Request payload	Result payload
file	v1	Copy a templatized file onto the target host	path/name/content/mode	-

An action is a simple as writing a Golang struct

type Action interface{
    Name() string
    Version() semver.Version
    Help() string
    Execute(Request) Response // TODO
    Schema() string // OpenAPI Schema
}

type Request interface{
    DryRun() bool
    onChanges() []Action
    Payload() map[string]interface{}
}

type Response interface{
    HasChanged() bool
    Payload() map[string]interface{}
}

Reporters

You can also link roles or actions to reporters, reporters are like monitoring probes, it allow you, for example, to start an NGINX server and be sure it is always up.

type Reporter interface {
  Name() string
  Version() semver.Version
  Help() string
  Execute(Request) Response
  Schema() string // OpenAPI Schema
}