Entity Framework Core service default lifetime

Yes, the default life time for DbContext is scoped. This is intended this way.

Instantiating DbContext is pretty cheap and it makes sure that the your do not use to many resources. If you'd have a DbContext with a singleton lifetime, then all records that you read once will be tracked by the DbContext, unless you specifically disable tracking. This will require much more memory usage and it will keep growing.

And the more the DbContext tracks, the lower the performance will be. That's why you often see DbContext only used within a using(var context = new AppDbContext()) block.

In web applications however, using the using block is bad, because the lifetime is managed by the framework and if you dispose it to early the calls after that will fail with an exception.

If you use transient lifetime on the other side, you will lose the "transaction" functionality. With scoped, the DbContext has a transaction scope that's as long as the request.

If you need more fine-grained control, you have to use the Unit of Work pattern (which DbContext already kind of utilize).

For your second question:

If you create a service, it must have a lifetime that's equal to the one of the scope or shorter (read: Scoped or transient).

If you explicitly need a longer life-time for a service, you should inject a DbContext factory service or factory method into your service.

You can accomplish this with something like

services.AddTransient<Func<AppDbContext>>( (provider) => new Func<MyDbContext>( () => new AppDbContext()));
services.AddSingleton<IMySingletonService, MySingletonService>();

And your service may look like this:

public class MySingletonService : IMySingletonService, IDisposable
{
    private readonly AppDbContext context;

    public MySingletonService(Func<AppDbContext> contextFactory)
    {
        if(contextFactory == null)
            throw new ArgumentNullException(nameof(contextFactory));

        // it creates an transient factory, make sure to dispose it in `Dispose()` method.
        // Since it's member of the MySingletonService, it's lifetime
        // is effectively bound to it. 
        context = contextFactory();
    }
}

---

Note: In EF Core 2 there is now a new method AddDbContextPool which creates a pool of contexts that can be reused. The scoping is still the same, but the instance will be 'reset' and returned to the pool. I would have thought the overhead of 'resetting' would be the same as just creating a new one but I guess that isn't the case.

If this method is used, at the time a DbContext instance is requested by a controller we will first check if there is an instance available in the pool. Once the request processing finalizes, any state on the instance is reset and the instance is itself returned to the pool.+ This is conceptually similar to how connection pooling operates in ADO.NET providers and has the advantage of saving some of the cost of initialization of DbContext instance.

https://docs.microsoft.com/en-us/ef/core/what-is-new/ef-core-2.0#high-performance

---

Admittedly anecdotal, this is what I’ve learned. And it’s from a lazy KISS devotion. I avoided other complications and solved my issue of premature EF Core DbContext disposes w/o concern of pools, scopes etc. I was merely throwing together a POC w/o regard to async return values, chained up from the Controller to a Repository and so on, which all essentially returned “void”, i.e. literally the singular, “Task”. That caused iterations amongst my DbContext members in lower routines to inexplicably dispose of the underlying DbContext. All I had to do was make every async method return a Task<whatever return> value and everything worked. EF Core don’t like async void return values.

---