WebApi's {“message”:“an error has occurred”} on IIS7, not in IIS Express

181
votes

I'm working with ASP.NET MVC 4 WebApi and am having a lot of fun with it running it on my local computer on IIS Express. I've configured IIS Express to serve remote machines too, and so other's in my company are using my computer as our webserver.

After deciding this was a less-than-optimal solution, we decided to put the WebApi on a remote server after installing .NET 4.5. When I use fiddler and sent a POST to a controller on my local machine it returns the correct response, yet when I change the domain to the webserver running IIS7 the same POST returns a cryptic

{"message":"an error has occurred"}

message. Anyone have any idea what could be going on?

9
iis-7asp.net-mvc-4asp.net-web-api
What's the HTTP status code on the error response? If it's 500 it's very likely the web site/application config is invalid for the remote machine with IIS 7. Create a simple HTML file on the remote machine, browse it on the remote machine if possible to make sure it can be viewed and then try to hit it from your machine to see if it is successful or not. – Sixto Saez
It is a 500-error. Thanks for the suggestion, but the default index.html page that the WebApi provides works. Also should add that some of the API webservices work and others don't, whereas all of them work on my local machine. – nsg
You'll need to enable IIS request tracing to get more specifics when you see a 500 error. A 500 error usually occurs before the Web API routing kicks but I guess its possible to trigger it by something your code is doing. Look at the IIS trace logging and see if that offers any clues. – Sixto Saez
You might be able to get the server to give you more verbose error information in its response by initiating the request from a browser on the server machine itself (e.g. using a Remote Desktop session). – Jon Schneider

9 Answers

286
votes

The problem was a missing dependency that wasn't on the server but was on my local machine. In our case, it was a Devart.Data.Linq dll.

To get to that answer, I turned on IIS tracing for 500 errors. That gave a little bit of information, but the really helpful thing was in the web.config setting the <system.web><customErrors mode="Off"/></system.web> This pointed to a missing dynamically-loaded dependency. After adding this dependency and telling it to be copied locally, the server started working.

103
votes

Basically:

Use IncludeErrorDetailPolicy instead if CustomErrors doesn't solve it for you (e.g. if you're ASP.NET stack is >2012):

GlobalConfiguration.Configuration.IncludeErrorDetailPolicy 
= IncludeErrorDetailPolicy.Always;

Note: Be careful returning detailed error info can reveal sensitive information to 'hackers'. See Simon's comment on this answer below.

TL;DR version

For me CustomErrors didn't really help. It was already set to Off, but I still only got a measly an error has occurred message. I guess the accepted answer is from 3 years ago which is a long time in the web word nowadays. I'm using Web API 2 and ASP.NET 5 (MVC 5) and Microsoft has moved away from an IIS-only strategy, while CustomErrors is old skool IIS ;).

Anyway, I had an issue on production that I didn't have locally. And then found I couldn't see the errors in Chrome's Network tab like I could on my dev machine. In the end I managed to solve it by installing Chrome on my production server and then browsing to the app there on the server itself (e.g. on 'localhost'). Then more detailed errors appeared with stack traces and all.

Only afterwards I found this article from Jimmy Bogard (Note: Jimmy is mr. AutoMapper!). The funny thing is that his article is also from 2012, but in it he already explains that CustomErrors doesn't help for this anymore, but that you CAN change the 'Error detail' by setting a different IncludeErrorDetailPolicy in the global WebApi configuration (e.g. WebApiConfig.cs):

GlobalConfiguration.Configuration.IncludeErrorDetailPolicy 
= IncludeErrorDetailPolicy.Always;

Luckily he also explains how to set it up that webapi (2) DOES listen to your CustomErrors settings. That's a pretty sensible approach, and this allows you to go back to 2012 :P.

Note: The default value is 'LocalOnly', which explains why I was able to solve the problem the way I described, before finding this post. But I understand that not everybody can just remote to production and startup a browser (I know I mostly couldn't until I decided to go freelance AND DevOps).

39
votes

None of the other answers worked for me.

This did: (in Startup.cs)

public class Startup
{
    public void Configuration(IAppBuilder app)
    {
        var config = new HttpConfiguration();

        WebApiConfig.Register(config);

        // Here: 
        config.IncludeErrorDetailPolicy = IncludeErrorDetailPolicy.Always;
    }
}

(or you can put it in WebApiConfig.cs):

public static class WebApiConfig
{
    public static void Register(HttpConfiguration config)
    {
        // Web API routes
        config.MapHttpAttributeRoutes();

        config.Routes.MapHttpRoute(
            name: "DefaultApi",
            routeTemplate: "api/{controller}/{action}/{id}",
            defaults: new { id = RouteParameter.Optional }
        );

        // Here: 
        config.IncludeErrorDetailPolicy = IncludeErrorDetailPolicy.Always;
    }
}
15
votes

I always come to this question when I hit an error in the test environment and remember, "I've done this before, but I can do it straight in the web.config without having to modify code and re-deploy to the test environment, but it takes 2 changes... what was it again?"

For future reference

<system.web>
   <customErrors mode="Off"></customErrors>
</system.web>

AND

<system.webServer>
  <httpErrors errorMode="Detailed" existingResponse="PassThrough"></httpErrors>
</system.webServer>
12
votes

I had a similar problem when posting to the WebAPI endpoint. By turning the CustomErrors=Off, i was able to see the actual error which is one of the dlls was missing.

11
votes

In case this helps anyone:

I had a similar issue, and following Nates instructions I added:

<system.web>
     <customErrors mode="Off"/>
 </system.web>

This showed me more information about the error:

"ExceptionMessage": "Unable to load the specified metadata resource.", "ExceptionType": "System.Data.Entity.Core.MetadataException", "StackTrace": " at System.Data.Entity.Core.Metadata.Edm.MetadataArtifactLoaderCompositeResource.LoadResources(...

This is when I remembered that I had moved the edmx file to a different location and had forgotten to change the connectionstrings node in the config (connectionsstrings node was placed in a seperate file using "configSource", but that's another story).

0
votes

My swagger XML file was not deployed into \bin:

GlobalConfiguration.Configuration
  .EnableSwagger(c =>
  {
    c.SingleApiVersion("v1", "SwaggerDemoApi");
    c.IncludeXmlComments(string.Format(@"{0}\bin\SwaggerDemoApi.XML", 
                         System.AppDomain.CurrentDomain.BaseDirectory));
    c.DescribeAllEnumsAsStrings();
  })

http://wmpratt.com/swagger-and-asp-net-web-api-part-1/

enter image description here

It had to be set in the Release Configuration as well as in the Debug Configuration.

0
votes

If you have <deployment retail="true"/> in your .NET Framework's machine.config, you won't see detailed error messages. Make sure that setting is false, or not present.

0
votes

So i tried all the suggested solutions to no avail. All i did was to set run the app from the server and it displayed the error in full, this should have worked when i set customErrors mode to false but it didn't. The moment i browsed the API form the server i was able to see the problem.