多线程实体框架:连接未关闭。连接的当前状态是连接

时间:2012-09-20 22:15:55

标签: multithreading entity-framework dbcontext

所以我有一个执行工作流程的Windows服务流程。后端在Entity Framework之上使用Repository和UnitofWork Pattern and Unity,以及从edmx生成的实体类。我不会详细介绍它,因为它不是必需的,但基本上工作流程有5个步骤。特定过程可能在任何时间点的任何阶段(按当然顺序)。第一步只生成第二步的数据,它通过长时间运行的进程验证数据到另一台服务器。然后步骤生成带有该数据的pdf。对于每个阶段,我们生成一个计时器,但是它可以配置为允许为每个阶段生成多个计时器。其中就是问题所在。当我将处理器添加到特定阶段时,它会随机出现以下错误:

连接未关闭。连接的当前状态是连接。

读到这一点似乎很明显,这种情况正在发生,因为上下文试图从两个线程访问同一个实体。但这就是让我陷入困境的地方。我可以在此找到的所有信息表明我们应该在每个线程中使用实例上下文。据我所知,我正在做什么(见下面的代码)。我没有使用单例模式或静态或任何东西所以我不确定为什么会发生这种情况或如何避免它。我已在下面发布了我的代码的相关部分供您审核。

基础资料库:

 public class BaseRepository
{
    /// <summary>
    /// Initializes a repository and registers with a <see cref="IUnitOfWork"/>
    /// </summary>
    /// <param name="unitOfWork"></param>
    public BaseRepository(IUnitOfWork unitOfWork)
    {
        if (unitOfWork == null) throw new ArgumentException("unitofWork");
        UnitOfWork = unitOfWork;
    }


    /// <summary>
    /// Returns a <see cref="DbSet"/> of entities.
    /// </summary>
    /// <typeparam name="TEntity">Entity type the dbset needs to return.</typeparam>
    /// <returns></returns>
    protected virtual DbSet<TEntity> GetDbSet<TEntity>() where TEntity : class
    {

        return Context.Set<TEntity>();
    }

    /// <summary>
    /// Sets the state of an entity.
    /// </summary>
    /// <param name="entity">object to set state.</param>
    /// <param name="entityState"><see cref="EntityState"/></param>
    protected virtual void SetEntityState(object entity, EntityState entityState)
    {
        Context.Entry(entity).State = entityState;
    }

    /// <summary>
    /// Unit of work controlling this repository.       
    /// </summary>
    protected IUnitOfWork UnitOfWork { get; set; }

    /// <summary>
    /// 
    /// </summary>
    /// <param name="entity"></param>
    protected virtual void Attach(object entity)
    {
        if (Context.Entry(entity).State == EntityState.Detached)
            Context.Entry(entity).State = EntityState.Modified;
    }

    protected virtual void Detach(object entity)
    {
        Context.Entry(entity).State = EntityState.Detached;
    }

    /// <summary>
    /// Provides access to the ef context we are working with
    /// </summary>
    internal StatementAutoEntities Context
    {
        get
        {                
            return (StatementAutoEntities)UnitOfWork;
        }
    }
}

StatementAutoEntities是自动生成的EF类。

存储库实现:

public class ProcessingQueueRepository : BaseRepository, IProcessingQueueRepository
{

    /// <summary>
    /// Creates a new repository and associated with a <see cref="IUnitOfWork"/>
    /// </summary>
    /// <param name="unitOfWork"></param>
    public ProcessingQueueRepository(IUnitOfWork unitOfWork) : base(unitOfWork)
    {
    }

    /// <summary>
    /// Create a new <see cref="ProcessingQueue"/> entry in database
    /// </summary>
    /// <param name="Queue">
    ///     <see cref="ProcessingQueue"/>
    /// </param>
    public void Create(ProcessingQueue Queue)
    {
        GetDbSet<ProcessingQueue>().Add(Queue);
        UnitOfWork.SaveChanges();
    }

    /// <summary>
    /// Updates a <see cref="ProcessingQueue"/> entry in database
    /// </summary>
    /// <param name="queue">
    ///     <see cref="ProcessingQueue"/>
    /// </param>
    public void Update(ProcessingQueue queue)
    {
        //Attach(queue);
        UnitOfWork.SaveChanges();
    }

    /// <summary>
    /// Delete a <see cref="ProcessingQueue"/> entry in database
    /// </summary>
    /// <param name="Queue">
    ///     <see cref="ProcessingQueue"/>
    /// </param>
    public void Delete(ProcessingQueue Queue)
    {
        GetDbSet<ProcessingQueue>().Remove(Queue);  
        UnitOfWork.SaveChanges();
    }

    /// <summary>
    /// Gets a <see cref="ProcessingQueue"/> by its unique Id
    /// </summary>
    /// <param name="id"></param>
    /// <returns></returns>
    public ProcessingQueue GetById(int id)
    {
        return (from e in Context.ProcessingQueue_SelectById(id) select e).FirstOrDefault();
    }

    /// <summary>
    /// Gets a list of <see cref="ProcessingQueue"/> entries by status
    /// </summary>
    /// <param name="status"></param>
    /// <returns></returns>
    public IList<ProcessingQueue> GetByStatus(int status)
    {
        return (from e in Context.ProcessingQueue_SelectByStatus(status) select e).ToList();
    }

    /// <summary>
    /// Gets a list of all <see cref="ProcessingQueue"/> entries
    /// </summary>
    /// <returns></returns>
    public IList<ProcessingQueue> GetAll()
    {
        return (from e in Context.ProcessingQueue_Select() select e).ToList();
    }

    /// <summary>
    /// Gets the next pending item id in the queue for a specific work        
    /// </summary>
    /// <param name="serverId">Unique id of the server that will process the item in the queue</param>
    /// <param name="workTypeId">type of <see cref="WorkType"/> we are looking for</param>
    /// <param name="operationId">if defined only operations of the type indicated are considered.</param>
    /// <returns>Next pending item in the queue for the work type or null if no pending work is found</returns>
    public int GetNextPendingItemId(int serverId, int workTypeId, int? operationId)
    {
        var id = Context.ProcessingQueue_GetNextPending(serverId, workTypeId,  operationId).SingleOrDefault();
        return id.HasValue ? id.Value : -1;
    }

    /// <summary>
    /// Returns a list of <see cref="ProcessingQueueStatus_dto"/>s objects with all
    /// active entries in the queue
    /// </summary>
    /// <returns></returns>
    public IList<ProcessingQueueStatus_dto> GetActiveStatusEntries()
    {
        return (from e in Context.ProcessingQueueStatus_Select() select e).ToList();
    }
    /// <summary>
    /// Bumps an entry to the front of the queue 
    /// </summary>
    /// <param name="processingQueueId"></param>
    public void Bump(int processingQueueId)
    {
        Context.ProcessingQueue_Bump(processingQueueId);
    }
}

我们使用Unity进行依赖注入,一些调用代码例如:

#region Members
    private readonly IProcessingQueueRepository _queueRepository;       
    #endregion

    #region Constructors
    /// <summary>Initializes ProcessingQueue services with repositories</summary>
    /// <param name="queueRepository"><see cref="IProcessingQueueRepository"/></param>        
    public ProcessingQueueService(IProcessingQueueRepository queueRepository)
    {
        Check.Require(queueRepository != null, "processingQueueRepository is required");
        _queueRepository = queueRepository;

    }
    #endregion

启动计时器的Windows服务中的代码如下:

            _staWorkTypeConfigLock.EnterReadLock();
        foreach (var timer in from operation in (from o in _staWorkTypeConfig.WorkOperations where o.UseQueueForExecution && o.AssignedProcessors > 0 select o) 
                              let interval = operation.SpawnInternval < 30 ? 30 : operation.SpawnInternval 
                              select new StaTimer
                            {
                                Interval = _runImmediate ? 5000 : interval*1000,
                                Operation = (ProcessingQueue.RequestedOperation) operation.OperationId
                            })
        {
            timer.Elapsed += ApxQueueProcessingOnElapsedInterval;
            timer.Enabled = true;
            Logger.DebugFormat("Queue processing for operations of type {0} will execute every {1} seconds", timer.Operation, timer.Interval/1000);                
        }
        _staWorkTypeConfigLock.ExitReadLock();

StaTimer只是定时器添加操作类型的包装器。 ApxQueueProcessingOnElapsedInterval然后基本上只是根据操作将工作分配给流程。

我还将添加一些ApxQueueProcessingOnElapsedInterval代码,用于生成任务。

            _staTasksLock.EnterWriteLock();
        for (var x = 0; x < tasksNeeded; x++)
        {
            var t = new Task(obj => ProcessStaQueue((QueueProcessConfig) obj),
                             CreateQueueProcessConfig(true, operation), _cancellationToken);


            _staTasks.Add(new Tuple<ProcessingQueue.RequestedOperation, DateTime, Task>(operation, DateTime.Now,t));

            t.Start();
            Thread.Sleep(300); //so there are less conflicts fighting for jobs in the queue table
        }
        _staTasksLock.ExitWriteLock();

3 个答案:

答案 0 :(得分:27)

看起来您的服务,存储库和上下文应该在您的应用程序的整个生命周期中存在,但这是不正确的。您可以同时触发多个计时器。这意味着多个线程将并行使用您的服务,他们将在他们的线程中执行您的服务代码=在多个线程之间共享上下文=&gt;异常,因为上下文不是线程安全的。

唯一的选择是为要执行的每个操作使用新的上下文实例。例如,您可以更改类以接受上下文工厂而不是上下文,并为每个操作获取新的上下文。

答案 1 :(得分:7)

如果这有助于任何人:

在我的情况下,我确实确保非线程安全的TransientLifetimeActionFilters(使用Ninject),但它仍然导致并发问题!事实证明,在我的一些自定义DbContext中,我使用依赖注入来访问构造函数中的ActionFilters,但是OnActionExecuting具有使其在多个请求中实例化的生命周期,因此上下文没有重新创建。

我通过手动解析{{1}}方法中的依赖项而不是构造函数来修复它,以便每次都是一个新的实例。

答案 2 :(得分:-1)

在我的情况下,我遇到了这个问题,因为在我的一个DAL函数调用之前忘记了sudo lsusb关键字。将await解决了。