Replace many-to-many relations between Nodes↔︎Agents and Services↔︎Agents to one-to-many. Change database structure, simplify code.
Update Confluence documentation.
Why we do it and how it will affect the other parts of code?
To make it easier to understand for users and us we decided to not use many-to-many relations.
Because we don't have cases where we need it.
For rds_exporter we are going to store each `credentials - instance` pairs as separate rds_exporter record in database.
But we will run only one instance of rds_exporter per pmm-agent with combined config file.
Config files will be generated based on all rds_exporters related to current pmm-agent.
Generating config file and everything related to rds_exporter is out of scope of this task.