How do you manage database connections using Akka actors?

Question

I've been studying Akka for some time and I am building an app that utilizes the actor model and requires to maintain a connection to the database, but I have a dilemma: Where do I put this connection and how can I manage it?

This is my reasoning so far:

The connection should be initialized only once.
Since the connection represents a state, it should reside inside an actor. Let's call this actor DatabaseConnection (just to be creative).
Since I don't want to share the actor's state, all querying should happen inside this actor.
The database driver I am using is reactive-mongo so each query returns a Future that can be piped to the sender.

Even though the querying is done through Futures, I can't help thinking that this model cannot scale. Only one thread managing all database access messages? It sounds unreasonable even utilizing Futures. I have thought of making child workers to manage database querying but I would have to share the connection with the children. I like this last idea because if the DatabaseConnection actor dies, all its children die as well in theory. But I don't know if there is another better way to solve this problem without sharing state. Is there?

Possible duplicate of [Akka and ReactiveMongo](https://stackoverflow.com/questions/18638570/akka-and-reactivemongo) — cvargascr, Sep 14 '17 at 06:42
Possible duplicate of [ReactiveMongo: single connection pool instance Vs many connection pool instances Vs multiple connection pools](https://stackoverflow.com/questions/45859204/reactivemongo-single-connection-pool-instance-vs-many-connection-pool-instances) — cchantep, Sep 15 '17 at 22:00

score 0 · Accepted Answer · answered Sep 14 '17 at 10:54

Based on the example code for reactive mongo it doesn't look like Actor encapsulation is necessary. Examining the code sample:

val collection = db[BSONCollection]("articles")

val query = BSONDocument()

val cursor = collection.find(query).cursor[BSONDocument]

val futureList = cursor.toList

The collection object already acts like an entity which encapsulates the state of the "connection" and responds to "queries" with Future results. That is almost the exact definition of the Actor you were thinking of creating.

Wrapping the above code inside of an Actor body doesn't seem to buy you much...

Jeffrey Chung · Answer 2 · 2017-09-14T13:29:29.657

ReactiveMongo, as of the current 0.12.6 version, is itself built on Akka. The main processing component (housed here) is an actor:

trait MongoDBSystem extends Actor {
   ...
}

Basically, ReactiveMongo has already implemented what you're trying to achieve, so if you're using ReactiveMongo, there's no point in attempting to deal with the low-level connection details in your own actor.

If you want to understand one way to use actors to handle connections and serve requests, study ReactiveMongo's implementation. In addition to the MongoDBSystem trait linked above, check out the MongoDriver and MongoConnection classes, as well as the tests in DriverSpec.

How do you manage database connections using Akka actors?

2 Answers2