Ansible deep merge hash array in vars - ansible

Question

A want declare hash array in vars/main.yml or default/main.yml of some role, e.g.:
mysql:
instances:
new:
port: 3306
dir: /mydir
config:
innodb_log_file_size: '128M'
tmp_table_size: '128M'
innodb_buffer_pool_size: '10G'
...
And I want change only some values of keys in YAML inventory or group_vars. Other values must be taken from vars/main.yml of some role:
mysql:
instances:
new:
config:
innodb_buffer_pool_size: '2G'
I want result for used in jinja2 template:
mysql:
instances:
new:
port: 3306
dir: /mydir
config:
innodb_log_file_size: '128M'
tmp_table_size: '128M'
innodb_buffer_pool_size: '2G'
...

Answer 1

Question: "Want to change only some values of keys in YAML inventory or group_vars. Other values must be taken from vars/main.yml of some role".
1) Service role (nginx, mysql, pgsql ...). In this role, I describe the default settings, ...
2) Then I create a project role in which I can include the service role and I will use most of the default settings described in the service role. Only a small part of the service settings can be changed in the project role.
Answer:
In the "service" roles create in defaults special variables for parameters that might be changed later. For example
mysql_port: "3306"
mysql_dir: "mydir"
mysql_innodb_log_file_size: "128M"
mysql:
instances:
new:
port: "{{ mysql_port }}"
dir: "{{ mysql_dir }}"
config:
innodb_log_file_size: "{{ mysql_innodb_log_file_size }}"
...
In the "project" role any variable with higher precedence will override the role's defaults.