Returning custom HTTP Status codes for WCF SOAP Exceptions

When WCF encounters an unhandled exception, the thrown exception is wrapped up in a FaultException and returned to the client. You can of course throw your own FaultException to have better control over the contents of the error, but one thing you cannot control by default is the HTTP response code. It is always 500.

I had a situation where I needed to return a 503 error under certain circumstances in a WCF SOAP service.

The solution is not as simple as I might like, but it isn’t terrible. There are examples out there for changing the http status code for all exceptions (https://msdn.microsoft.com/es-es/library/ee844556(v=vs.95).aspx), but this was not at all what we want.

The solution uses the same basic framework as the mentioned article however. An Endpoint Behavior Extension registers a Dispatch Message Inspector that watches for faults. In the case of a fault. Inside the BeforeSendReply method, you have access to the reply message. For performance reasons it is best to not unwrap the XML, so we use the SOAP Action header to trigger the HTTP Status code update.

I decided to create a simple custom exception class that sets the SOAP Action to a predefined value

    /// <summary>
    /// Custom Fault Exception that when used with the CustomFaultStatusBehavior Endpoint Behavior
    /// allows returning custom HTTP status codes to the client
    /// </summary>
    public class StatusFaultException : FaultException
    {
        /// <summary>
        /// Create new exception
        /// </summary>
        /// <param name="statusCode">HTTP Status code to be returned to the client</param>
        /// <param name="faultReason">SOAP Fault Reason</param>
        /// <param name="faultCode">SOAP Fault Code</param>
        public StatusFaultException(System.Net.HttpStatusCode statusCode, string faultReason, FaultCode faultCode)
            :base(faultReason, faultCode, "CustomFaultStatus" + ((int)statusCode).ToString())
        {
            //StatusCode is placed in the response Action. Action would be "CustomFaultStatus503" to return a 503 error code
        }
    }

Here is the Custom Behavior that consumes the SOAP Action

    /// <summary>
    /// Endpoint Behavior that allows returning custom HTTP response codes for SOAP Faults 
    /// </summary>
    public class CustomFaultStatusBehavior : BehaviorExtensionElement, IEndpointBehavior
    {
        //based on https://msdn.microsoft.com/es-es/library/ee844556(v=vs.95).aspx
        public override Type BehaviorType
        {
            get
            {
                return typeof(CustomFaultStatusBehavior);
            }
        }

        public void AddBindingParameters(ServiceEndpoint endpoint, BindingParameterCollection bindingParameters)
        {
        }

        public void ApplyClientBehavior(ServiceEndpoint endpoint, ClientRuntime clientRuntime)
        {
        }

        public void ApplyDispatchBehavior(ServiceEndpoint endpoint, EndpointDispatcher endpointDispatcher)
        {
            CustomFaultStatusMessageInspector inspector = new CustomFaultStatusMessageInspector();
            endpointDispatcher.DispatchRuntime.MessageInspectors.Add(inspector);
        }

        public void Validate(ServiceEndpoint endpoint)
        {
        }

        protected override object CreateBehavior()
        {
            return new CustomFaultStatusBehavior();
        }
    }

    /// <summary>
    /// Message Inspector that updates the HTTP response code for faulted messages with a CustomFaultStatus action
    /// </summary>
    public class CustomFaultStatusMessageInspector : IDispatchMessageInspector
    {
        public object AfterReceiveRequest(ref Message request, IClientChannel channel, InstanceContext instanceContext)
        {
            return null;
        }

        public void BeforeSendReply(ref Message reply, object correlationState)
        {
            if (!reply.IsFault) return;
            if (!reply.Headers.Action.StartsWith("CustomFaultStatus", StringComparison.Ordinal)) return;
            //get the string value for desired response code
            string statusCodeString = reply.Headers.Action.Substring(17);
            //convert to int
            int statusCodeInt;
            if (!int.TryParse(statusCodeString, out statusCodeInt)) return;

            //cast to HttpStatusCode
            System.Net.HttpStatusCode statusCode = System.Net.HttpStatusCode.InternalServerError;
            try
            {
                statusCode = (System.Net.HttpStatusCode)statusCodeInt;
            }
            catch (Exception ex)
            {
                return;
            }

            // Here the response code is changed
            reply.Properties[HttpResponseMessageProperty.Name] = new HttpResponseMessageProperty() { StatusCode = statusCode };
        }
    }

The CustomFaultStatusBehavior must be registered in your web.config as a behavior extension, then it must be referenced in an endpoint behavior. Finally this behavior should be applied to the endpoint using the behaviorConfiguration attribute.

Dirt simple WCF Routing Service configuration tutorial with authentication

As I’ve said in previous posts, I’m a big fan of WCF in general – recently I have had the opportunity to work with a less popular, but extremely powerful feature: the WCF Routing Service (https://msdn.microsoft.com/en-us/library/ee517423(v=vs.110).aspx)

The MSDN documentation is confusing. The existing blog articles are confusing or broken. Examples of hosting it in IIS using simplified configuration are pretty much missing…

The routing service allows you to create proxy service endpoints. In my case, I needed to expose some internal web services to the internet through a VPN connection in Azure. Also, I needed to implement security on these publicly exposed endpoints.

The internal web services are hosted by a non-microsoft service bus, use HTTP with no authentication.
The external endpoints need to run over HTTPS (SSL) and require authentication.

Yes, you can do this with the Routing Service. No, it does not require a bunch of code.

The routing service is capable of using xpath matching on the message header or message itself to determine the endpoint routing. You could use the SOAP Action for example. Or… a much simpler approach is to use URL-pattern based routing.

In fact, you can do 95% of it in the web.config only. The only place you will need code is to apply authorization (what users can access the proxy), but it is very simple.

Suppose you have three SOAP endpoints you would like to proxy.
http://mailserver/sendservice
http://customermanagement/customerretrieval
http://customermanagement/customerupdate

Suppose you want to serve these up through you service as
https://proxyservice.example.org/sendmail
https://proxyservice.example.org/getcustomer
https://proxyservice.example.org/updatecustomer

You want Basic Authentication over SSL, using Windows for credentials. Finally, you want a means of limiting which authenticated users can access the proxy.

So, step by step.

1. Create a WCF Project

1.1 Setup IIS to host your project
Create a site in IIS pointing to the folder containing Service1.svc. Disable anonymous auth, enable basic auth, set it up for SSL.
(I recommend using a real ssl cert with a hosts entry pointing back at localhost. Easier than getting self signed certs working…)

2. Delete the IService1.cs; expand the Service1.svc and delete the code behind (Service1.svc.cs).

3. Rt click the Service1.svc and edit markup. Replace it with the following:

<%@ ServiceHost Language="C#" Debug="true" Service="System.ServiceModel.Routing.RoutingService,System.ServiceModel.Routing, version=4.0.0.0, Culture=neutral,PublicKeyToken=31bf3856ad364e35" %>

This will cause your Service1.svc to invoke the routing service. No code required (yet) – you do the rest in the web.config.

4. Update your web.config with the following.

<?xml version="1.0"?>
<configuration>
  <configSections>
  </configSections>
  <system.web>
    <compilation debug="true" targetFramework="4.5"/>
  </system.web>
  <system.serviceModel>
    <client>
      <!--Define named internal (private-side) endpoints here. Defined like any client endpoint except the contract doesn't need to be specified. Define binding as needed -->
      <endpoint name="ep_sendmail" address="http://mailserver/sendservice" binding="basicHttpBinding" contract="*"/>
      <endpoint name="ep_getcustomer" address="http://customermanagement/customerretrieval" binding="basicHttpBinding" contract="*"/>
      <endpoint name="ep_updatecustomer" address="http://customermanagement/customerupdate" binding="basicHttpBinding" contract="*"/>
    </client>
    <routing>
      <filters>
        <!--Define named filters that will be applied to requests coming into the (public side of) the router. (hostname is ignored - anything can be used - protocol and path/query strings must match identically)-->
        <filter name="f_sendmail" filterType="EndpointAddress" filterData="https://host/Service1.svc/SendMail"/>
        <filter name="f_getcustomer" filterType="EndpointAddress" filterData="https://host/Service1.svc/GetCustomer"/>
        <filter name="f_updatecustomer" filterType="EndpointAddress" filterData="https://host/Service1.svc/UpdateCustomer"/>
      </filters>
      <filterTables>
        <filterTable name="filterTable1">
          <!--Define the mapping between the filter match and endpoint. I'm using a 1:1:1 mapping-->
          <add filterName="f_sendmail" endpointName="ep_sendmail"/>
          <add filterName="f_getcustomer" endpointName="ep_getcustomer"/>
          <add filterName="f_updatecustomer" endpointName="ep_updatecustomer"/>
        </filterTable>
      </filterTables>
    </routing>
    <services>
      <service behaviorConfiguration="RoutingBehavior" name="System.ServiceModel.Routing.RoutingService">
        <!--Server endpoint must be defined with the appropriate contract (Most likely IRequestReplyRouter) and associated with its own binding-->
        <endpoint address="" binding="basicHttpBinding" bindingConfiguration="externalBinding" name="RouterEndpoint1" contract="System.ServiceModel.Routing.IRequestReplyRouter"/>
      </service>
    </services>
    <bindings>
      <basicHttpBinding>
        <!--For this example, the default or internal binding are just using the defaults-->
        <binding name="internalBinding"/>
        <!--Binding configuration for the router's "endpoint" - configures it to expect Basic Authentication over SSL.-->
        <binding name="externalBinding">
          <security mode="Transport">
            <transport clientCredentialType="Basic"/>
          </security>
        </binding>
      </basicHttpBinding>
    </bindings>
    <behaviors>
      <serviceBehaviors>
        <behavior>
          <!--These behaviors are likely unused-->
          <serviceMetadata httpGetEnabled="true"/>
          <serviceDebug includeExceptionDetailInFaults="true"/>
        </behavior>
        <behavior name="RoutingBehavior">
          <!--The behaviors applied specifically to the routing service - must specify the filtertable name -->
          <routing routeOnHeadersOnly="true" filterTableName="filterTable1"/>
          <serviceDebug includeExceptionDetailInFaults="true"/>
          <!--The router's provided metadata is pretty much useless - it is the IRequestReply contract-->
          <serviceMetadata httpGetEnabled="false"/>
          <!--This tells the router to use Windows for establishing identities, and to use our custom class for determining permission-->
          <serviceAuthorization principalPermissionMode="UseWindowsGroups" serviceAuthorizationManagerType="ExampleNamespace.ExampleAuthorizationManager, ExampleAssemblyName"/>
        </behavior>
      </serviceBehaviors>
    </behaviors>
    <!--Very important - there is a bug of sorts in the router when using basicHttpBinding with asp compatibility. Just disable it.-->
    <serviceHostingEnvironment aspNetCompatibilityEnabled="false" multipleSiteBindingsEnabled="true"/>
  </system.serviceModel>
</configuration>

If you comment out the serviceAuthorization tag, your project should build and run – so long as you have ssl and basic auth working…

You should update the “ExampleNamespace.ExampleAuthorizationManager, ExampleAssemblyName” with the Namespace, class name and assembly name that you actually use
(often the namespace and assembly name are the same)

If you simply want a router – with no auth and no ssl remove the enternalBinding’s security node AND the serviceAuthorization node

5. Implement the ExampleAuthorizationManager

namespace ExampleNamespace
{
    public class ExampleAuthorizationManager : ServiceAuthorizationManager
    {
        protected override bool CheckAccessCore(OperationContext operationContext)
        {
            //check that the user is allowed and return true to allow, false to deny
            return true;
            //return base.CheckAccessCore(operationContext);
        }
    }
}

6. Update clients

Since there is no WSDL, you will need build your clients against the internal endpoints’ WSDLs, then update them to use the external endpoint. Additionally you will need to update the binding configuration to use Transport security with Basic Authentication (unless you have turned this off).

You now have a secure proxy for WCF services without having to reimplement the services in WCF – adding new services requires 3 lines of configuration. I’m not sure if Microsoft could have made this any easier… except perhaps by documenting it better ­čśŤ

Generating a single flattened WSDL from an existing WCF Service

UPDATE 4/12/2013: Tool included
Download WSDL Flattener Tool
I’m not including source because it’s messy (I prefer to cleanup my code before sharing…) but feel free to decompile with ILSpy if you don’t trust me.

UPDATE 8/11/2015: Command line tool included
WSDL Flattener Command Line Utility
Somehow I forgot to save my update – thanks Luke for pointing it out!

WCF is great – its flexible, standards compliant, super easy to work with, and just altogether nice. That is, until you try to use it with, say, Oracle products. WCF generates so-called modular wsdl files, with separate xsd files for types, and even sometimes multiple wsdl files. Not technically in violation of any standards, but kind-of pushing the limits.

There are plenty of articles out there about modifying a WCF service to generate a flat WSDL. Also if you upgrade to .NET 4.5, you can use the URL parameter ?singleWsdl to get a flattened WSDL.

But what if you don’t have control of the web service? What if you cannot upgrade or modify the code?

For some odd reason, Microsoft added the internal ability to generate a single WSDL but only exposed it through the WCF Service implementation. They didn’t add it to svcutil.exe, and they didn’t expose it in the Framework… publicly that is *evil grin*

Turns out, in the System.ServiceModel.Description namespace, there is an internal static class called “WsdlHelper”, and this class has a public static Method called “GetSingleWsdl” that takes a MetadataSet for a parameter, and returns a┬áSystem.Web.Services.Description.ServiceDescription object representing the Wsdl as a single entity. All you have to do is call .Write(filename) and you have your file.

“But I don’t see any WsdlHelper class!!!!!!”

It’s internal – you have to use reflection to invoke it.
First, you have to load the metadata from the existing WCF service into a MetadataSet using a MetadataExchangeClient. (Hint: you will probably need to use a custom binding). Then you run that set into something like this:

            var asy = System.Reflection.Assembly.GetAssembly(typeof(WsdlExporter));
            Type t = asy.GetType("System.ServiceModel.Description.WsdlHelper", true);
            var method = t.GetMethod("GetSingleWsdl", System.Reflection.BindingFlags.Public | System.Reflection.BindingFlags.Static);

            object retVal = method.Invoke(null, new object[] { metadataSet });
            //cast retVal into System.Web.Services.Description.ServiceDescription, and call the .Write() method. All done

Using this, I was able to create a tool that takes a WCF url, and allows you to save a single .wsdl file. You want the tool? Later perhaps. For now put your thinking cap and Google gloves on, and write your own.