Reputation: 58422
I have create an mvc web api 2 webhook for shopify:
public class ShopifyController : ApiController
{
// PUT: api/Afilliate/SaveOrder
[ResponseType(typeof(string))]
public IHttpActionResult WebHook(ShopifyOrder order)
{
// need to return 202 response otherwise webhook is deleted
return Ok(ProcessOrder(order));
}
}
Where ProcessOrder
loops through the order and saves the details to our internal database.
However if the process takes too long then the webhook calls the api again as it thinks it has failed. Is there any way to return the ok
response first but then do the processing after?
Kind of like when you return a redirect in an mvc controller and have the option of continuing with processing the rest of the action after the redirect.
Please note that I will always need to return the ok response as Shopify in all it's wisdom has decided to delete the webhook if it fails 19 times (and processing too long is counted as a failure)
Upvotes: 15
Views: 29970
Reputation: 1753
I used Response.CompleteAsync();
like below. I also added a neat middleware and attribute to indicate no post-request processing.
[SkipMiddlewareAfterwards]
[HttpPost]
[Route("/test")]
public async Task Test()
{
/*
let them know you've 202 (Accepted) the request
instead of 200 (Ok), because you don't know that yet.
*/
HttpContext.Response.StatusCode = 202;
await HttpContext.Response.CompleteAsync();
await SomeExpensiveMethod();
//Don't return, because default middleware will kick in. (e.g. error page middleware)
}
public class SkipMiddlewareAfterwards : ActionFilterAttribute
{
//ILB
}
public class SomeMiddleware
{
private readonly RequestDelegate next;
public SomeMiddleware(RequestDelegate next)
{
this.next = next;
}
public async Task Invoke(HttpContext context)
{
await next(context);
if (context.Features.Get<IEndpointFeature>().Endpoint.Metadata
.Any(m => m is SkipMiddlewareAfterwards)) return;
//post-request actions here
}
}
Upvotes: 2
Reputation: 169
Task.Run(() => ImportantThing()
is not an appropriate solution, as it exposes you to a number of potential problems, some of which have already been explained above. Imo, the most nefarious of these are probably unhandled exceptions on the worker process that can actually straight up kill your worker process with no trace of the error outside of event logs or something at captured at the OS, if that's even available. Not good.
There are many more appropriate ways to handle this scenarion, like a handoff a service bus or implementing a HostedService
.
Upvotes: 1
Reputation: 30208
There are a few options to accomplish this:
Hangfire
or Quartz
run the actual processing, where your web request just kicks off the task.RabbitMQ
, to run the actual process, and the web request just adds a message to the queue... be careful this one is probably the best but can require some significant know-how to setup.Javascript AJAX
kick off the process in the background and maybe you can turn retry off on that request... still that keeps the request going in the background so maybe not exactly your cup of tea.Upvotes: 6
Reputation: 58422
I have managed to solve my problem by running the processing asynchronously by using Task
:
// PUT: api/Afilliate/SaveOrder
public IHttpActionResult WebHook(ShopifyOrder order)
{
// this should process the order asynchronously
var tasks = new[]
{
Task.Run(() => ProcessOrder(order))
};
// without the await here, this should be hit before the order processing is complete
return Ok("ok");
}
Upvotes: 17