Feature: Clustered Database Connections

[hodgesds]

Background

I work in an environment where we used a clustered database setup where there are replicas in use. The replicas are used in different ways such as bulk ETL operations for business analysts, reindexing for search engines, and other cron jobs that don't need the most up to date data. However, some of instances, such as cron jobs that need to do writes, do need to use transactions on the master database. I had opened #1673 but I think this is the appropriate place to track this.

Things that would be nice:

• A method to detect if a query was mutable
• Health checks on database nodes and potentially marking nodes in the pool as unhealthy

There's probably some other nice things to have but I can't think of them right now. As I'm just getting familiar with the diesel codebase I don't think I can provide great insight as to the best way to implement this feature, but would be more than willing to help out once everything is well defined.

Thanks!