Stephen A. Fuqua (SAF) is a Bahá'í, software developer, and conservation and interfaith advocate in the DFW area of Texas.

Notes on WCF in Business Applications, Part 1: Server Side

February 2, 2014

Self-Hosting

WCF is the .Net component for "web services;" naturally the default mechanism of hosting a service is through the web server. While this can be easy to install, it does require that you can install and configure IIS on the server machine. If you have no other needs for a web server, or want the advantages of bindings other than HTTP, then self-hosting is a better option.

In this model, the developer typically builds a Windows service that provides the host. Client applications connect to a URL just as they do for a "hosted" (IIS) solution. The Windows service must create an instance of the ServiceContract class, and inject that into a ServiceHost. The following code assumes that the service's endpoint (URL), binding, and behavior are setup in the application's config file (note).

If you used the Windows Service template when creating the project in Visual Studio, then you'll need to manually add the endpoint, binding, and behavior to the application's config file, whereas they will be present by default if you used the WCF template. Alternately, you can right click the app.config file in Solution Explorer, and choose Edit WCF Configuration. This will load the config file into a helpful editor.

public partial class MyService : ServiceBase
{
   // be sure to close this in the Dispose() method, which will be in the MyService.Designer.cs file   
   protected ServiceHost ServiceHost { get; set; }

   public MyService()
   {
      InitializeComponent();
   }

   protected override void OnStart(string[] args)
   {
      ServiceHost = new ServiceHost(typeof(MyService));
      ServiceHost.Open();
   }

   protected override void OnStop()
   {
      closeService();
   }

   private void closeService()
   {
      if (ServiceHost != null)
      {
         if (ServiceHost.State == (CommunicationState.Opened ^ CommunicationState.Opening))
         {
            ServiceHost.Close();
         }
      }
   }
}

Threaded Startup

When Windows starts up a service, it only gives 30 seconds to the service's OnStart command. This is a synchronous method, and if no response is received in that window, then the service times out. There is a method on the ServiceBase, RequestAdditionalTime, that can be used to extend beyond that timeout – but the timeout is there for a reason. Assuming that your service starts when the server starts, it is best to let the OS get on with the business of starting up without having to sit around waiting for your service to finish starting.

Therefore, start the service via a thread. Don't Wait() for the thread to complete; just let it do its own thing. I like to use the Task Parallel Library instead of the Thread class.

protected object threadLock = new object();

protected override void OnStart(string[] args)
{
   Task.Run(() =>
      {
         lock (threadLock)
         {
            ServiceHost = new ServiceHost(typeof(MyService));
            ServiceHost.Open();
         }
      });
}

Development Debugging

Windows services are difficult to debug; you can't start the Service from Visual Studio using the "Start Debugging" command. The simplest way to debug a WCF service is to run it in a console instead of as a service. When using the default Visual Studio template for a Windows Service Application, you get a Main() function like this:

static void Main()
{
   ServiceBase[] ServicesToRun;
   ServicesToRun = new ServiceBase[]
   { 
      new MyService() 
   };
   ServiceBase.Run(ServicesToRun);
}

Use compiler directives to change this so that the program runs as a stand-alone application from Visual Studio when in Debug mode, but remains a Windows service when compiled in Release mode. Refactor the OnStart method so that it calls a public Start method, which itself can be called from the static Main() method. I will also clean up that default code a little bit…

        var myService = new MyService();
#if DEBUG
        myService.Start();
        Console.WriteLine("Press any key to close down the service");
        Console.Read();
#else
        ServiceBase[] ServicesToRun = new ServiceBase[] { myService };
        ServiceBase.Run(ServicesToRun);
#endif

Installing the Service

There are two ways to install the service:

1. Add an Installer class per the instructions in Walkthrough: Creating a Windows Service Application in the Component Designer, or
2. Build an MSI using the WiX Toolset (StackOverflow post on Installing a service using WiX)

The first option is simple and therefore inappropriate in many cases. Building an MSI allows you to

• Bundle dependent DLL's together to make sure nothing is left out;
• Avoid asking someone to pull up a command-prompt and get the right .Net Framework path for InstallUtil.exe;
• Prompt the installer to enter a service account name and password, instead of making them open the Service Manager and manually enter those values as an additional step (note);
• Perhaps provide options, such as installing with a Test-environment config file instead of the default config file geared toward the Production environment.

By default, the service would otherwise install using the "LOCAL SERVICE" account, which is built into Windows. But this does not provide a satisfactory level of control for a service that needs to access resources, such as files or a database. In that case, it is best to create a dedicated domain account. Grant that account permission to access necessary files and to execute stored procedures or perform CRUD operations directly on tables.

None of these advantages are strictly necessary; you could do them all manually. But manual steps are prone to error and frustration. So building an MSI package is a manifestation of the Be Kind admonition from my unfinished TACKLE theory of software development.

Run-time Debugging

As with any other application, presumably the service also has some logging mechanism built in. But sometimes the basic logging is not enough — you'll need to know if applications are actually managing to call the service, the order of operation calls, and which calls generated errors / faults. It is trivial to enable logging a substantial amount of event log data; in fact, it is very easy to accidentally log an overwhelming amount of data.

This logging is configured via the system.diagnostics node in the application's config file. The default configuration provided by Microsoft includes two types of events: Information and ActivityTracing. I find that "Information" provides far too much data, unless perhaps I am tuning for performance. As a commenter in the link above notes, be sure that the account running your service has write access to the directory that you configure for the log file. For error debugging, turn on the Warning, Error, and Critical switches. Again, this is configurable via the Microsoft Service Configuration Editor; however, it only allows you to configure one switch for the trace source, so I had to manually add the ,Error,Critical. If that is still too much information, turn off Warnings and ActivityTracing. The generated file can be opened by the developer using the Service Trace Viewer Tool.

<system.diagnostics>
      <sources>
            <source propagateActivity="true" name="System.ServiceModel" switchValue="Warning,Error,Critical,ActivityTracing">
                  <listeners>
                        <add type="System.Diagnostics.DefaultTraceListener" name="Default">
                            <filter type="" />
                        </add>
                  </listeners>
            </source>
      </sources>
      <sharedListeners>
            <add initializeData="C:\Logs\MyService.svclog" type="System.Diagnostics.XmlWriterTraceListener, System, Version=4.0.0.0, Culture=neutral, PublicKeyToken=b77a5c561934e089"
                  name="NewListener" traceOutputOptions="LogicalOperationStack, DateTime, Timestamp, ProcessId, ThreadId, Callstack">
                <filter type="" />
            </add>
      </sharedListeners>
</system.diagnostics>

Logging takes a bit of a toll on performance, so I recommend commenting out this whole section of the config file unless and until you need it. Since this is a Windows service, don't forget to stop and re-start the service after making a change to the installed config file.

Preview of the Next Parts

In two follow-ups to this note, I plan to discuss:

1. Shared Concerns
   • Security
   • Binding
   • Faults and Exceptions
   • Dependency Injection
   • Shared Library
2. Client Side
   • Custom Service Client / Proxy
   • Channel Caching
   • Using Statement
   • General References