Skip to Content

Go + gRPC + OPA - A Perfect Union - Part 1

Posted on 4 mins read

TL;DR – In a series of blog posts, I will be implementing a simple library application supporting both gRPC and REST interfaces using Go, gRPC, and OPA. My approach might not be the most optimal one, but I am learning these technologies currently. Please give your valuable suggestions and be kind :)


I have been learning the basics of microservices and Golang lately. On the work front, I got a chance to work on Go, gRPC, and Open Policy Agent as my first professional project. In this post, I will be demonstrating what I learned in the last few months. We will be implementing a simple gRPC based library service, which will be able to serve requests using both gRPC and REST calls. It will also incorporate Open Policy Agent (OPA) to provide the authorization to users. Let’s begin with a quick introduction to gRPC and OPA.

gRPC

gRPC is Google’s implementation for Remote Procedure Calls(RPC). RPC is mainly used in building scalable distributed systems. While REST has a limited set of verbs, RPC can define any function calls including synchronous and asynchronous calls.

In gRPC, the client can make procedure calls as if the calls are made to some local function. However, the underlying client stub (auto-generated) will send the call to the server. The server will have a similar server stub which will be able to handle the requests coming from the client. The server will send the response to the client using similar mechanisms over the network. All the communication is serialized to binary format, so it is ideal for distributed systems as binary format tends to be on faster side for large amounts of data.

For more info about gRPC, visit the official website.

Open Policy Agent (OPA)

OPA gives us the ability to define a fine-grained policy control mechanism. However, I think the most important benefit of using OPA is that it gives you the ability to decouple your services and the definition of policies from the enforcement of it.

There are mainly two parts of OPA -

  1. A JSON document store where you can define anything from your users, access roles, permission levels, etc.
  2. A policy is written in a declarative language. This policy gives you new derivative data from the original JSON document store evaluated by the policy. The declarative language is called Rego, and these policies are also documents which generate results according to the defined policy. The users query these results.

This info will be sufficient for our use case. More info about it can be found at the official website.

The skeleton of our Application

We will be building a command line library application. This application will support adding, deleting, searching, issuing and returning of books. There are three types of users – student, faculties, and staff. Not all users have a similar kind of access level. There are some books which are only reserved for students, and some are exclusively reserved for faculties and so on.

I think, emulating the library at Hogwarts will be a good idea here. Students and ordinary people were not allowed inside the restricted section of the library. We will emulate that restricted section using the authorization mechanisms provided by OPA.

The gRPC protocol will handle the communications part in our app, but not everyone in Hogwarts is using gRPC. Wizards are still in love with REST (not a bad thing, though), so we will provide them alike REST interface to interact with our service. We do not want to face the wrath of the wizards, after all.

OPA can either be used as a standalone application or be embedded in Go service as well. Both approaches have their pros and cons. I have decided to use OPA as an independent service, as it will be more comfortable that way for us to push authorization data and policies to it if such a need arises in the future.

One more issue remains - where to store the added books. Shall we persist them? In practical scenarios, persistence is always a good idea, but here I do not want to complicate things too much. We could use any Object storage servers (e.g., Minio) to store the JSON data generated from our gRPC methods, but that will unnecessarily add complexity to our simple scenario. Wizards do not like complicate stuff, you know!

In the coming posts, we will define our gRPC proto file and build the Go service around it. Then we will add a REST interface to our service, for simple CRUD operations in our application. Then we will add OPA authorization to restrict our users from adding and viewing books which they are not supposed to access.

Thank you for reading. Cheers 😊

comments powered by Disqus