Rails Engines可伸缩性问题

Question

I need to find a way for scaling my rails monolith application with the help of rails engines.

Goal: I have database connection timeout issue and monolith has more than 200+ models. what we want to do is divide our models into the tree-like structure of engines. and we will be able to use a separate database for each engine.

UseCase: let's say we have engine A as the base engine and which is included in engine B and C respectively. Both B and C engines live on the same level of the tree.

So i have models seggrated in different engines.

Engine A: has all data related to user.

Class User 
end

Engine B: has all data related to products

class Product
end

Engine C: has all data related to reports.

class Report
end

now the main issue comes while defining the associations . earlier we were having associations and several other methods which access associations. For eg.

class User
  has_many products

  def get_title_product
    products.pluck(:title)
  end
end

now I cannot define it in engine A as products table doesn't live there.

Option: what I know is I have to open that User model inside Engine B and define all the association and get_title_product logic related to this Domain in engine B itself. I can't even include Engine B in Engine A because it will result in circular dependency.

I don't want to follow above approach because it will get messy and my application is significantly large, additionally I don't think it is good as per rails best practices.

Thanks in advance.

Answer 1

Your post contains many questions. Your asking about database sharding, architecture with rails engines and performances / scalability with the timeout issue.

Performances / Timeout

First your timeout issue is not related to the number of models as you suggest. The number of models as no impact at all on performances. To know where is the performance problem or the bottleneck you should use a monitoring tool. This one is my favorite https://www.rorvswild.com (DISCLAIMER: I'm the author ^_^). There is other competitors, use the one your like more.

For the timeouts maybe you should check around your database config. We have not enough informations here to go deeper.

Database sharding

That is not trivial at all since you cannot JOIN and reference foreign keys when they are not in the same db. That is why you have to carefully chose where to shard your db. Ideally it's where you have the least joins. It's a long work which has an important impact on your code base. A stack overflow post is not enough to talk about sharding. Fortunately there is a lot of articles and gems to help you for that.

Just be sure you understand that you can split the load across many databases, but it comes at an extra price on your code base.

For the relationships across databases you cannot use the Rails builtin has_many and so on. You have to define the relationships yourself or use a gem which will help for that. To give you an idea:

class User
  def products
    Product.where(user_id: id)
  end
end

Rails engines

They are great to built reusable piece of features across applications ( http://guides.rubyonrails.org/engines.html ). Looks like reusing is not your goal. So I'm afraid your going in the wrong direction.

If you don't want to reopen the class you can use a module:

module HasProducts
  def self.included(model)
    model.has_many(:products)
  end

  def get_title_product
    products.pluck(:title)
  end
end