Question

We create one Linq2Sql DataContext for every request using the following class:

public static class DataContextManager
{

    private const string HTTPCONTEXT_KEY = "DataContextManagerKey";

    private static CompanyDataContext _staticContext; //when there's no HttpContext (in test/debug situations). 

    public static CompanyDataContext Context
    {
        get
        {
            if (_Context == null)
            {
                _Context = NewContext();
            }
            return _Context;
        }
    }

    private static CompanyDataContext _Context
    {
        get
        {
            return (CompanyDataContext)(HttpContext.Current != null ? HttpContext.Current.Items[HTTPCONTEXT_KEY] : _staticContext);
        }
        set
        {
            if (HttpContext.Current != null)
            {
                HttpContext.Current.Items[HTTPCONTEXT_KEY] = value;
            }
            else
            {
                DataContextManager._staticContext = value;
            }    
        }
    }

    public static void Dispose()
    {
        CompanyDataContext context = _Context;
        if (context != null)
        {
            if (Config.Instance.TestMode) context.Log.WriteLine("--- DISPOSING DATACONTEXT ---");
            context.Dispose();
            _Context = null;
        }
    }

    public static CompanyDataContext NewContext()
    {  
        CompanyDataContext db = new CompanyDataContext();
        db.CommandTimeout = Config.SQL_COMMAND_TIMEOUT;
        if (Config.Instance.TestMode)
        {
            db.Log = new ConsoleTextWriter();
            db.Log.WriteLine("--- CREATING NEW DATACONTEXT ---");
        }
        return db;
    }

}

And in Global.asax:

protected void Application_EndRequest(Object sender, EventArgs e)
{
    DataContextManager.Dispose();
}

The reason why I’m asking is that we’re suddenly getting random “SqlException: Server failed to resume the transaction” exceptions once or twice every day with code that used to work perfectly. After the exception we get a lot of other exceptions until we restart the web application. Anyone seen this behaviour before?

We’re running ASP .Net 2.0 with SQL server 2005 on IIS 6.

UPDATE:

Just so no one else does the same horrible mistake as we did:

It turned out that some worker threads also used the DataContext but with no HttpContext they of course got the _staticContext (a feature in DataContextManager only to be used when testing). We rewrote the code in the worker threads to make sure one DataContext per thread and disposing it when done. And so far everything has worked for 2 weeks 🙂

Answer 1

This is a bad pattern. First you should never have a static data context it implements IDisposable so one thread could try to use the context while another is disposing of it plus many other potential problems. One data context per http request is no good either, data contexts are designed to be used for a single transaction then disposed of. You get problems if you retrieve update/insert/delete and retrieve using the same context, the second retrieve does not reflect the changes of the update/insert/delete. Remove the static context and just have the Context property return a new context every time. You could still dispose of the all at the end of a request by sticking them all in a List property and iterating through it.