DEV Community

Kevin Griffin @ #Shedquarters
Kevin Griffin @ #Shedquarters

Posted on • Originally published at consultwithgriff.com on

HttpContext Issues with Azure SignalR Service

This is my entry for this years CS Advent. Take a moment to go check out the other great entries!

Last week, the team and I were migrating an application built with ASP.NET Core and SignalR to use the Azure SignalR Service. The migration was pretty straight forward, but we ran into an issue with the HttpContext when using the Azure SignalR Service.

When a normal SignalR integration is running under an ASP.NET host, you will typically have access to the HttpContext for the SignalR request. This is best done by injecting IHttpContextAccessor into your Hub class.

We wrote a bit of code that needed to look at the HttpContext to pull some information from a cookie. In normal operation, this code worked fantastically and it served us well for YEARS.

public class SignalRHub : Hub
{
    private readonly IHttpContextAccessor _httpContextAccessor;
    public SignalRHub(IHttpContextAccessor httpContextAccessor)
    {
        _httpContextAccessor = httpContextAccessor;_
    }

    public async Task SendMessage(string message)
    {
        var httpContext = _httpContextAccessor.HttpContext;

        // look up cookie tzInformation
        var tzInformation = httpContext.Request.Cookies["tzInformation"];

        // do something with tzInformation
    }
}
Enter fullscreen mode Exit fullscreen mode

There is a bit of problem now that we migrated this code to Azure SignalR Service. The IHttpContextAccessor is still injected into the Hub class, but it won't have the same information you'd have in a normal SignalR integration.

If you think about it for a moment, or read this Github issue from 2019, you'll realize that it makes sense. The HttpContext is from a connection between the client and the SignalR. When Azure SignalR Service is used, the hub is still on your server, but the connection is between the client and the Azure SignalR Service. The Azure SignalR Service then forwards the messages to your hub.

What's a better approach?

The primary mistake we made was assuming the state was available when it wasn't. The better approach is to pass the information you need from the client to the hub. In our case, we needed the timezone information from the client. We could have passed that information as a parameter to the SendMessage method.

public class SignalRHub : Hub
{
    public async Task SendMessage(string message, string tzInformation)
    {
        // do something with tzInformation
    }
}
Enter fullscreen mode Exit fullscreen mode
const connection = new signalR.HubConnectionBuilder()
    .withUrl("/signalr")
    .build();

async function sendMessage(message: string) {
    const tzInformation = "America/New_York";
    await connection.invoke("SendMessage", message, tzInformation);
}

async connection.start();
Enter fullscreen mode Exit fullscreen mode

What about Identity?

One of the cool things that Azure SignalR Service will do for you is it will pass along any IPrincipal information that is set on the HttpContext. This is great if you are using Identity and want to know who the user is. You can use the Context.User property to get the user information.

public class SignalRHub : Hub
{
    public async Task SendMessage(string message)
    {
        var user = Context.User;
        // do something with user
    }
}
Enter fullscreen mode Exit fullscreen mode

Conclusion

If you've found yourself here because of a Google search, I hope you have a little bit of resolve now. Maybe if you're looking for more SignalR information, take a look at my course, SignalR Mastery. It's a great way to learn SignalR and it's on sale for the holidays! If you're looking for more information on Azure SignalR Service, check out the Azure SignalR Service documentation.

Top comments (0)