Rework diff generation and usage

Currently we omit new zones in the diff. The goal is to include them and also to be able to differentiate between a change in an existing zone and a new zone.

This is a first step towards zone bootstrap automation.

Given a change in config1 and config2 in zone1 the diff currently looks like this:

{
  "zone1": [ "config1", "config2" ]
}

The proposal is to change it to something like:

{
  "<ZONE_NAME>": {
    "<CONFIG_NAME>": "<CHANGE_TYPE>"
  }
}

CHANGE_TYPE is create | update | delete

When a config is removed another is updated:

{
  "zone1": {
    "config1": "update"
    "config2": "delete"
  }
}

When a new zone appears it could look like this:

{
  "zone2": {
    "configa": "create"
    "configb": "create"
    "configc": "create"
  }
}

Now after the diff generation we should be able to determine if a new zone needs to be bootstrapped:

If all configs change type is "create", then we are bootstrapping a zone:

create a branch (but no MR)
generate the bootstrap pipeline using this branch
the different jobs in the pipeline will be running trackbone apply -z zone -c config -t bootstrap=true

If no new zone:

create MR like it is done currently

Since the diff can now contain configurations to create, update or delete we should have a dedicated command to apply a diff.

Maybe we can rename the diff command to make-diff or generate-diff
Add a plan-diff <DIFF_PATH> command
Add a apply-diff <DIFF_PATH> command
the --diff-file option of plan, apply and destroy commands is removed

The plan-diff command set action to plan for configurations with create or update change type, for delete we can just comment the fact that the configuration will be deleted.

The apply-diff command will properly set action to realize on the different configs. create, update -> apply, delete -> destroy

Edited Jan 15, 2021 by Ghost User