What should I consider for my pool size in database.yml?

Question

I've been having some thoughts on setting the appropriate pool size (in database.yml).

I saw a similar question at what is the use of the pool option in database.yml but needed to go further to understand this more.

Here is my setup. I have instances running Unicorn with 5 worker process'. I also have Sidekiq instances with concurrency set to 5 (I am asumming thats 5 concurrent threads).

What should I consider for my pool size? What are other factors should I take into consideration?

This is my understanding so far (as quoted from an engineer I was speaking to earlier):

Suppose you leave it at the default 5. It just means that, inside "a single process", the pool size is 5. The pool is per process. It is not system-wide so having a value of 5 does not mean that you are limited to 5 processes it just means that each process has its own pool, of size 5.

In summary, each instance has a db pool size of 5. This also means the db pool size setting is not app wide, but per process/instance. Which also means, since Sidekiq and Unicorn will run in its own instance. It will have its own db pool size of 5. Is this assumption correct?

At this point, I can assume that the default pool size of 5 is generally safe. Your thoughts?

PS. If you can share your pool size for your AWS RDS instances. That would be appreciated.

Answer 1

Your pool size in database.yml should be not less then max number of sidekiq/unicorn processes (not sum). Example, if unicorn running with 5 processes and sidekiq starting with 15 concurrency, you should set 15 pool size. Or, if unicorn with 7 and sidekiq with 5, you need 7 connections in database.yml.