Deploying a microservice with its own database: how to guarantee replication and reliability?

Question

I am building an application to be deployed in a cloud environment. I am using Kube.netes with Docker.

My image contains two containers: one for the application, and another with Postgres DBMS (including its own database that is stored through a persistent volume claim).

I did it like this because this way I could scale my application linearly, as my main requirement is to be able to scale it like that.

So when I scale my deployment, it will look more or less like the draw bellow (having N pods of my application):

I have done performance tests into this architecture, and it works very well for my needs, and it scales as I needed too.

However, I am worried about the data replication and its reliability.

So here are my questions:

1. How to do replication with this type of architecture? Is there a way in which Kube.netes takes care of it through a specific storage class? What is the best approach?

2. In general, what is a DBA expectation when looking into this architecture? Is it acceptable as I may end up with too many databases?

Answer 1

Postgres scaling cannot be performed as you suggested and persistent volumes cannot be mounted in read/write mode in multiple pods ( with a few exceptions )

To scale Postgres you can consider CrunchyData

More over, decoupling the application layer from the DB, has many benefits, including the granular control of the scaling policies upon each layer.

PS: Nice diagram. It makes the question very intelligible.