带有实体框架的Hangfire-DbContext并发问题

Question

I have a hangfire service responsible for two jobs:

• Creating email objects with their relevant attachments (Time-consuming process)
• Processing and sending out Pending emails.

I am using ASP.NET MVC (5.2.3) , with StructureMap (4.5.1) for creating an IoC container for Dependency Injection, together with Hangfire (1.6.19)

The flow is:

• The email will be triggered to be created from the front-end, where a Fire and Forget hangfire job is created, pushing it into the database with a Pending status.
  • This process uses a few repository objects to create a report of the client in a PDF format (These repositories uses the DbContext to retrieve information from the database)
• Another recurring job, picks up all the Pending emails and composes an email for each of these using System.Net.Mail .

Both of these jobs work fine, except if there are multiple jobs creating emails at the same time.

For instance: If there are 10 jobs fired at the same time from the front-end to generate the emails, the job fails with the error:

A second operation started on this context before a previous asynchronous operation completed. Use 'await' to ensure that any asynchronous operations have completed before calling another method on this context. Any instance members are not guaranteed to be thread safe.

The line that is throwing this error is using await , but I am thinking that it is because there are different threads accessing the context at the same time:

var ids = await context.Objects
    .Where(x => x.id == clientId && !x.IsDeleted)
    .Select(x => x.id)
    .ToListAsync();

I have set up the DbContext trying different scopes, Transient and AlwaysUnique , but the error persists.

Also, this doesn't always happen at the same line, basically anywhere where the context is being used in the involved repositories - if both jobs hit the same line at the same time.

The jobs retry, and the emails are being sent out pretty much one per minute, which might not work if there are hundreds of emails to be processed and sent out (which is the problem, in my scenario)

Answer 1

The issue was resolved when I used the ContainerScoped LifeCycle.

---

StructureMap Docs :

ContainerScoped in this case means that a registration will be built once per Container, such that the root container, any child or profile container, and every single nested container will build its own object instance

This means that there is a new instance of the DataContext for every job fired from hangfire (If I understand it correctly, I might speak under correction)

Specifically for the Datacontext, I had to enable MARS (Multiple Active Result Sets) which enables SQL Server to allow the execution of multiple batches on a single connection.