In init systems that don't support automatic restart this might be handy when we fail because we boot up too fast.
This is an attempt to fix https://github.com/c3os-io/c3os/issues/47
Move out cmd pieces and split into its own package. Also make sure we reload
plugins before bootstrapping nodes.
Also commons out the agent configuration in a specific YAML file.
Now there is a `install` section in the config that has the fields that previously where in `c3os` but
were actually only used during install phase.
Also the k3s and c3os config were moved to the provider instead that in the global config.
This changeset also adds a `config_url` and `options` keyword in the c3os config.
Along with that the config logic is changed so the configuration is taken also from boot commands and merged in the final installed config file.
* 🎨 Split off cli into separate binaries
This commit splits off the cli into 3 binaries:
- agent
- cli
- provider
The provider now is a separate component that can be tested by itself
and have its own lifecycle. This paves the way to a ligher c3os variant,
HA support and other features that can be provided on runtime.
This is working, but still there are low hanging fruit to care about.
Fixes#14
* 🤖 Add provider bin to releases
* ⚙️ Handle signals
* ⚙️ Reduce buildsize footprint
* 🎨 Scan for providers also in /system/providers
* 🤖 Run goreleaser
* 🎨 Refactoring
