EverTask.Abstractions 1.5.0

Overview

EverTask is a .NET library for executing background tasks in .NET applications. It is designed to be simple and focuses on task persistence, ensuring that pending tasks resume upon application restart.

More detailed documentation will be provided in the future.

Features

• Background execution:<br> Easily run background tasks with parameters in .NET
• Persistence<br>Resumes pending tasks after application restarts.
• Managed Parallelism<br>Efficiently handles concurrent task execution with configurable parallelism.
• Scheduled, Delayed and recurring tasks<br>Schedule tasks for future execution or delay them using a TimeSpan. You can also create recurring tasks, for now only with cron expressions, but a fluent scheduler is on the way!
• Resilient execution<br>A powerful resilience feature to ensure your tasks are robust against transient failures. Fully customizable even with your custom retry policies.
• Optional specialized CPU-bound execution<br>Optimized handling for CPU-intensive tasks with an option to execute in a separate thread, ensuring efficient processing without impacting I/O-bound operations. Use judiciously for tasks requiring significant computational resources.
• Timeout management<br>Configure the maximum execution time for your tasks.
• Error Handling<br>Method overrides for error observation and task completion/cancellation.
• Monitoring (local and remote)<br>Monitor your task with the included in-memory monitoring or remotely with SignalR!<br>
• SQL Storage<br>Includes support for SQL Server storage, enabling persistent task management.<br>
• In-Memory Storage<br>Provides an in-memory storage solution for testing and lightweight applications.
• Serilog Integration<br>Supports integration with Serilog for detailed and customizable logging.<br>
• Extensible Storage & Logging<br>Designed to allow easy plug-in of additional database solutions or logging systems.
• Async All The Way<br>Fully asynchronous architecture, enhancing performance and scalability in modern environments.
• Simplicity by Design<br>Created for simplicity, using the latest .NET technologies.
• Inspiration from MediaTr<br>Implementation based on creating requests and handlers.

Efficient Task Processing

EverTask employs a non-polling approach for task management, utilizing the .NET's System.Threading.Channels to create a BoundedQueue. This queue efficiently manages task execution without the need for constant database polling. Upon application restart after a stop, any unprocessed tasks are retrieved from the database in bulk and re-queued in the channel's queue for execution by the background service. This design ensures a seamless and efficient task processing cycle, even across application restarts.

Basic Configuration

builder.Services.AddEverTask(opt =>
{
    opt.RegisterTasksFromAssembly(typeof(Program).Assembly);
})
.AddMemoryStorage();

Advanced Configuration

builder.Services.AddEverTask(opt =>
{
    opt.SetChannelOptions(500)
       .SetThrowIfUnableToPersist(true)
       .RegisterTasksFromAssembly(typeof(AppSettings).Assembly);
})
.AddSqlServerStorage(configuration.GetConnectionString("QueueWorkerSqlStorage")!,
    opt =>
    {
        opt.SchemaName          = "EverTask";
        opt.AutoApplyMigrations = true;
    })
.AddSerilog(opt => 
    opt.ReadFrom.Configuration(configuration, new ConfigurationReaderOptions { SectionName = "EverTaskSerilog" }));

For setting up all configurations, refer to the Fluent Service Configuration section.

Creating Requests and Handlers

This example demonstrates how to create a request and its corresponding handler in EverTask. The SampleTaskRequest and SampleTaskRequestHandler illustrate the basic structure. Additionally, the handler includes optional overrides that allow you to control and monitor the lifecycle of a background task, providing hooks for when a task starts, completes, is disposed, or encounters an error.

public record SampleTaskRequest(string TestProperty) : IEverTask;

public class SampleTaskRequestHanlder : EverTaskHandler<SampleTaskRequest>
{
    private readonly ILogger<SampleTaskRequestHanlder> _logger;

    public SampleTaskRequestHanlder(ILogger<SampleTaskRequestHanlder> logger)
    {
        _logger = logger;
    }

    public override Task Handle(SampleTaskRequest backgroundTask, CancellationToken cancellationToken)
    {
        _logger.LogInformation($"Property value: {backgroundTask.TestProperty}");
        return Task.CompletedTask;
    }
}

EverTaskHandler Optional overrides for lifecycle control

public override ValueTask OnStarted(Guid taskId)
{
    _logger.LogInformation($"====== TASK WITH ID {taskId} STARTED IN BACKGROUND ======");
    return ValueTask.CompletedTask;
}

public override ValueTask OnCompleted(Guid taskIdtaskId)
{
    _logger.LogInformation($"====== TASK WITH ID {taskId} COMPLETED IN BACKGROUND ======");
    return ValueTask.CompletedTask;
}

public override ValueTask OnError(Guid taskId, Exception? exception, string? message)
{
    _logger.LogError(exception, $"Error in task with ID {taskId}: {message}");
    return ValueTask.CompletedTask;
}

protected override ValueTask DisposeAsyncCore()
{
    _logger.LogInformation("====== TASK DISPOSED IN BACKGROUND ======");
    return base.DisposeAsyncCore();
}

Task Dispatch

To dispatch a task, obtain an instance of ITaskDispatcher. This can be done using Dependency Injection:

// Retrieving ITaskDispatcher via method injection
var _dispatcher = serviceProvider.GetService<ITaskDispatcher>();

// Alternatively, ITaskDispatcher can be injected directly into the constructor of your class
_dispatcher.Dispatch(new SampleTaskRequest("Hello World"));

Dispatching Tasks with Delay

You can also schedule tasks to be executed after a certain delay. This can be achieved using either TimeSpan or DateTimeOffset.

Using TimeSpan for Relative Delay

To delay task execution by a relative time period, use TimeSpan. This is useful when you want to postpone a task by a specific duration, such as 30 minutes or 2 hours from now.

// Delaying task execution by 30 minutes
var delay = TimeSpan.FromMinutes(30);
_dispatcher.Dispatch(new SampleTaskRequest("Delayed Task"), delay);

Using DateTimeOffset for Absolute Delay

Alternatively, use DateTimeOffset for scheduling a task at a specific future point in time. This is particularly useful for tasks that need to be executed at a specific date and time, regardless of the current moment.

// Scheduling a task for a specific time in the future
var scheduledTime = DateTimeOffset.Now.AddHours(2); // 2 hours from now
_dispatcher.Dispatch(new SampleTaskRequest("Scheduled Task"), scheduledTime);

Recurring Tasks

In addition to delayed and scheduled tasks, you can also configure recurring tasks. These tasks repeat at specified intervals, providing a powerful way to automate ongoing processes.

Currently, EverTask supports only cron expressions for task scheduling. However, a fluent builder that will further simplify this process is nearly ready.

Implementing Recurring Tasks

You can schedule recurring tasks using various approaches. Below are some examples with explanations:

Immediate Execution with Cron Schedule:

// Execute task immediately, then repeat according to a Cron schedule
await _dispatcher.Dispatch(task, builder => builder.RunNow().Then().UseCron("*/2 * * * *").MaxRuns(3));

Delayed Start with Cron Schedule:

// Execute task after a short delay, then repeat according to a Cron schedule
await _dispatcher.Dispatch(task, builder => builder.RunDelayed(TimeSpan.FromSeconds(0.5)).Then().UseCron("*/2 * * * *"));

Scheduled Start with Cron Schedule

// Schedule task to start at a specific time, then repeat according to a Cron schedule
await _dispatcher.Dispatch(task, builder => builder.RunAt(dateTimeOffset)).Then().UseCron("*/2 * * * *").MaxRuns(3));

Scheduling Only with Cron

// Schedule a recurring task based solely on a Cron schedule
await _dispatcher.Dispatch(task, builder => builder.Schedule().UseCron("*/2 * * * *").MaxRuns(3));

 

💡 Remember: Delayed, scheduled and recurring tasks are also persistent. If your app restarts, you won't lose these tasks – they'll be executed at the right time!

Task Cancellation

When you dispatch, you can capture the returned GUID to keep track of the task. If you need to cancel this task before it starts, use the Cancel method of ITaskDispatcher with this ID.

// Dispatching a task and getting its unique ID
Guid taskId = _dispatcher.Dispatch(new SampleTaskRequest("Cancelable Task"));

// Cancelling the task (if not started yet)
_dispatcher.Cancel(taskId);

💡 Note: The Cancel method triggers the CancellationToken in the task's Handle method. Tasks should check this token regularly to enable cooperative cancellation. Remember, this only affects tasks in progress; tasks cancelled before execution won't run.

Task Execution Timeout

EverTask provides a flexible approach to managing task execution times with its timeout functionality.

Global Default Timeout

A global default timeout for all tasks can be set using the SetDefaultTimeout option in the EverTask configuration. This timeout defines a uniform maximum duration for task execution, after which the CancellationToken will be marked as cancelled.

// Example of setting a global default timeout
builder.Services.AddEverTask(opt =>
{
    opt.SetDefaultTimeout(TimeSpan.FromMinutes(5)); // Sets a global timeout of 5 minutes
});

Customizing Timeout per Task Handler

In addition to the global timeout, individual task handlers can specify their own timeout periods. This is achieved by setting the Timeout property in the task handler, allowing for task-specific timeout durations.

public class MyCustomTimeoutTaskHandler : EverTaskHandler<MyTask>
{
    public MyCustomTimeoutTaskHandler()
    {
        // Setting a custom timeout for this specific handler
        Timeout = TimeSpan.FromMinutes(2);
    }
    
    public override Task Handle(CancellationToken cancellationToken)
    {
        // Task handling logic
    }
}

CPU-bound Task Handling

In addition to its default async-await behavior, which is ideal for I/O-bound operations (like database reads, email sending, file generation, etc.), EverTask provides a specialized handling mechanism for CPU-bound tasks. This is crucial for tasks that involve intensive computational work, such as data processing or complex calculations.

This is controlled by the CpuBoundOperation property in the EverTaskHandler.

public class MyCustomCPUIntensiveTaskHandler : EverTaskHandler<MyCPUIntensiveTask>
{
    public MyCustomCPUIntensiveTaskHandler()
    {
    CpuBoundOperation = true;
    }

    public override Task Handle(CancellationToken cancellationToken)
    {
        // CPU-intensive task logic
    }
}

💡 Note: When using CpuBoundOperation to execute tasks in a separate thread, be mindful of potential overhead and concurrency issues. This feature is best used for genuinely CPU-intensive tasks. Excessive use can lead to increased resource utilization and complexity.

Resilience and Retry Policies

EverTask now includes a powerful resilience feature to ensure your tasks are robust against transient failures. This is achieved through customizable retry policies, which are applied to task executions.

Default Linear Retry Policy

Tasks by default use the LinearRetryPolicy, set in the global configuration (SetDefaultRetryPolicy). This default policy attempts three executions with a 500-millisecond delay between them, addressing temporary issues that might hinder task completion.

// Example of setting a global default RetryPolicy
builder.Services.AddEverTask(opt =>
{
    opt.SetDefaultRetryPolicy(new LinearRetryPolicy(3, TimeSpan.FromMilliseconds(500)));
});

// Handler automatically inherits the default LinearRetryPolicy
public class MyTaskHandler : EverTaskHandler<MyTask>
{
    public override Task Handle(CancellationToken cancellationToken)
    {
        // Task handling logic
    }
}

Customizing Retry Policies

For each task handler, you can define or override the default retry policy settings. This includes changing the number of execution attempts, fixed execution times, or providing an array of TimeSpan values for the retry delays.

Retries with a fixed delay:

public class MyCustomRetryTaskHandler : EverTaskHandler<MyCustomRetryTask>
{
    public MyCustomRetryTaskHandler()
    {
        // Setting a custom LinearRetryPolicy with 2 retries and 300ms delay
        RetryPolicy = new LinearRetryPolicy(2, TimeSpan.FromMilliseconds(300));
    }
    
    public override Task Handle(CancellationToken cancellationToken)
    {
        // Task handling logic
    }
}

With an array of timespan:

public class MyCustomRetryTaskHandler : EverTaskHandler<MyCustomRetryTask>
{
    public MyCustomRetryTaskHandler()
    {
        // Define a LinearRetryPolicy with custom delays
        RetryPolicy = new LinearRetryPolicy(new TimeSpan[] 
        {
            TimeSpan.FromMilliseconds(200),  // First delay
            TimeSpan.FromMilliseconds(300),  // Second delay
            TimeSpan.FromMilliseconds(600)   // Third delay
        });
    }
    
    public override Task Handle(CancellationToken cancellationToken)
    {
        // Task handling logic
    }
}

Implementing Custom Retry Policies

EverTask's design allows for the implementation of custom retry policies. Create your own policy by implementing the IRetryPolicy interface. This enables you to craft unique retry mechanisms tailored to specific requirements of your tasks.

public class MyCustomRetryPolicy : IRetryPolicy
{
    public async Task Execute(Func<CancellationToken, Task> action, CancellationToken token = default)
    {
        // Implementation of your custom retry logic
    }
}

The on your handler:

public class MyCustomRetryTaskHandler : EverTaskHandler<MyCustomRetryTask>
{
    public MyCustomRetryTaskHandler()
    {
        // Setting your custom RetryPolicy
        RetryPolicy = new MyCustomRetryPolicy();
    }
    
    public override Task Handle(CancellationToken cancellationToken)
    {
        // Task handling logic
    }
}

Handling WorkerSerivce stops with CancellationToken in Task Handlers

In EverTask, every task handler's Handle method is provided with a CancellationToken. This token is crucial for effectively managing task interruptions when the EverTask WorkerService is stopped.

If the WorkerService is halted, the CancellationToken is set to a cancelled state. Tasks interrupted in this manner are marked as ServiceStopped and are re-queued upon the application's next restart.

This ensures no tasks are lost due to service stops, and the presence of CancellationToken in the Handle method allows for custom logic to track and manage partial execution of tasks.

Task Continuations and Rescheduling: Advanced Workflow Management

EverTask supports not only task continuations but also the rescheduling of tasks, providing a comprehensive approach to advanced workflow management.

Chaining Tasks for Sequential Execution

You can chain tasks for sequential execution by utilizing the lifecycle methods in task handlers. Overriding methods like OnComplete or OnError allows you to dispatch new tasks based on the outcome of the current task.

public class MyTaskHandler : EverTaskHandler<MyTaskRequest>
{
    private readonly IEverTaskDispatcher _dispatcher;

    public MyTaskHandler(IEverTaskDispatcher dispatcher)
    {
        _dispatcher = dispatcher;
    }

    public override ValueTask OnCompleted(Guid taskId)
    {
        // Dispatch another task upon completion
        _dispatcher.Dispatch(new AnotherTaskRequest());
        return base.OnCompleted(taskId);
    }
}

Task Rescheduling via Lifecycle Methods

In addition to task chaining, EverTask also enables task rescheduling within the same lifecycle methods. This feature is particularly useful for tasks that need to be executed repeatedly or at different intervals, depending on certain conditions or outcomes.

public class MyReschedulingTaskHandler : EverTaskHandler<MyTaskRequest>
{
    private readonly IEverTaskDispatcher _dispatcher;

    public MyReschedulingTaskHandler(IEverTaskDispatcher dispatcher)
    {
        _dispatcher = dispatcher;
    }

    public override ValueTask OnCompleted(Guid taskId)
    {
        // Reschedule the same task based on certain conditions or logic
        _dispatcher.Dispatch(new MyTaskRequest(), TimeSpan.FromMinutes(30)); // Rescheduling after 30 minutes
        return base.OnCompleted(taskId);
    }
}

With these capabilities, EverTask offers unparalleled flexibility in creating sophisticated, custom-tailored workflows for background task processing. Whether you need to chain tasks sequentially or reschedule them based on specific criteria, EverTask provides the necessary tools for robust and efficient task management.

Task Monitoring in EverTask

EverTask provides basic task monitoring through the TaskEventOccurredAsync event, accessible via dependency injection from IEverTaskWorkerExecutor. This event aggregates all the events generated by EverTask's WorkerService. Here's an example of how it can be utilized:

In your service class, you can subscribe to the TaskEventOccurredAsync event. For instance:

public class MyService
{
    public MyService(IEverTaskWorkerExecutor executor, ILogger<EverTaskTestController> logger)
    {
        executor.TaskEventOccurredAsync += data =>
        {
            logger.LogInformation("Message received from EverTask Worker Server: {@eventData}", data);
            return Task.CompletedTask;
        };
    }
}

The above code will produce this output:

Message received from EverTask Worker Server: EverTaskEventData { TaskId = dc49351d-476d-49f0-a1e8-3e2a39182d22, EventDateUtc = 19/11/2023 16:10:20 +00:00, Severity = Information, TaskType = EverTask.Example.AspnetCore.SampleTaskRequest, TaskHandlerType = EverTask.Example.AspnetCore.SampleTaskRequestHanlder, TaskParameters = {"TestProperty":"Hello World"}, Message = Task with id dc49351d-476d-49f0-a1e8-3e2a39182d22 was completed., Exception =  }

Here, data is of type EverTaskEventData which includes detailed information about the event:

public record EverTaskEventData(
    Guid TaskId,
    DateTimeOffset EventDateUtc,
    string Severity,
    string TaskType,
    string TaskHandlerType,
    string TaskParameters,
    string Message,
    string? Exception = null);

//Possible enum values
public enum SeverityLevel
{
    Information,
    Warning,
    Error
}

This allows users to trigger custom code based on the events produced by EverTask.

Real-Time Monitoring with SignalR

Additionally, EverTask offers real-time monitoring through a SignalR hub, available with the EverTask.Monitoring.AspnetCore.SignalR package for ASP.NET Core applications.

To use it, first register the service in your startup configuration:

builder.Services.AddEverTask(opt =>
       {
           opt
              .RegisterTasksFromAssembly(typeof(Program).Assembly);
       })
       .AddMemoryStorage()
       .AddSignalRMonitoring();

Then, register the middleware:

app.MapEverTaskMonitorHub();

This creates a SignalR hub at the URL /evertask/monitoring. You can customize this URL, for example:

app.MapEverTaskMonitorHub("/task-monitoring");

The hub sends all events in real-time to all connected clients. The event type is always EverTaskEventData.

Future updates will include new monitoring systems such as Email Alerts, ApplicationInsights, OpenTelemetry, etc.

💡 Note: Depending on EverTask's configuration, all events are also written to the logs as per the chosen configuration. Additionally, if SQL persistence is used, the task status and its audit trail can always be checked in the database.

 

---

<a name="fluent-service-configuration"></a>

Fluent Service Configuration

EverTaskService can be configured using a series of fluent methods, allowing a clear and user-friendly way to set up the service. These methods enable precise control over task processing, persistence, and parallel execution. Below are the available configuration methods, along with their default values and types:

SetChannelOptions (Overloaded Methods)

• Type: Action<BoundedChannelOptions> or int
• Default: Capacity set to 500, FullMode set to BoundedChannelFullMode.Wait
• Functionality: Configures the behavior of the task queue. You can directly specify the queue capacity or provide a BoundedChannelOptions object. This defines the maximum number of tasks that can be queued and the behavior when the queue is full.

SetThrowIfUnableToPersist

• Type: bool
• Default: true
• Functionality: Determines whether the service should throw an exception if it is unable to persist a task. When enabled, it ensures that task persistence failures are explicitly managed, aiding in data integrity.

SetMaxDegreeOfParallelism

• Type: int
• Default: 1
• Functionality: Sets the maximum number of tasks that can be executed concurrently. The default sequential execution can be adjusted to enable parallel processing, optimizing task throughput in multi-core systems.

SetDefaultRetryPolicy

• Type: IRetryPolicy
• Default: LinearRetryPolicy (with 3 tries every 500 milliseconds)
• Functionality: Defines a global default retry policy for tasks, using LinearRetryPolicy (3 attempts, 500 ms delay) unless overridden in task handlers. Supports custom policies via IRetryPolicy interface implementation.

SetDefaultTimeout

• Type: TimeSpan?
• Default: null
• Functionality: Specifies a global default timeout for tasks. If set, the CancellationToken provided to task handlers will be cancelled after the timeout duration. Users must handle this cancellation in their task logic. Setting to null means tasks have no default timeout and will run until completion or external cancellation.

RegisterTasksFromAssembly

• Functionality: Facilitates the registration of task handlers from a single assembly. This is particularly beneficial for applications structured in a modular fashion, enabling easy integration of task handlers.

RegisterTasksFromAssemblies

• Functionality: Allows for the registration of task handlers from multiple assemblies. This approach suits larger applications with distributed task handling logic spread across various modules or libraries.

SQL Server Persistence

• Schema Management: EverTask creates a new schema named EverTask by default. This approach avoids adding clutter to the main data schema.
• Schema Customization: Allows specifying a different schema or using null to default to the main schema.
• Migration Table: If using a custom Schema, event the Entity Framework Core migration table will be placed in that schema.
• Migration Handling: Option to apply database migrations automatically or handle them manually.

Serilog Integration

• Default Logging: Uses .NET configured ILogger by default.
• Serilog Option: Enables adding Serilog as a separate logger for EverTask, with customizable options.
• Example Configuration in appsettings.json:

  "EverTaskSerilog": {
    "MinimumLevel": {
      "Default": "Information",
      "Override": {
        "Default": "Information",
        "Microsoft": "Warning",
        "Microsoft.AspNetCore.SpaProxy": "Information",
        "Microsoft.Hosting.Lifetime": "Information",
        "Microsoft.EntityFrameworkCore.Database.Command": "Information"
      }
    },
    "WriteTo": [
      {
        "Name": "Console"
      },
      {
        "Name": "File",
        "Args": {
          "path": "Logs/evertask-log-.txt",
          "rollingInterval": "Day",
          "fileSizeLimitBytes": 10485760,
          "retainedFileCountLimit": 10,
          "shared": true,
          "flushToDiskInterval": "00:00:01"
        }
      }
    ],
    "Enrich": [
      "FromLogContext",
      "WithMachineName",
      "WithThreadId"
    ],
    "Properties": {
      "Application": "CliClub"
    }
  }
  
  

EverTask and EverTask.Abstractions

EverTask is complemented by the EverTask.Abstractions package, designed for use in Application projects where additional implementations are not required. This allows separation of concerns, keeping your application layer free from infrastructural code.

In your Infrastructure project, where EverTask is added, specify the assembly (or assemblies) containing IEverTask requests. This modular approach ensures that the application layer remains clean and focused, while the infrastructure layer handles task execution and management.

Serialization and deserialization of Requests for Persistence

EverTask uses Newtonsoft.Json for serializing and deserializing task requests, due to its robust support for polymorphism and inheritance, features that are limited in System.Text.Json. It is recommended to use simple objects for task requests, preferably primitives or uncomplicated complex objects, to ensure smooth serialization. In cases where EverTask is unable to serialize a request, it will throw an exception during the Dispatch method. This design choice emphasizes reliability in task persistence, ensuring that only serializable tasks are queued for execution.

Future Developments

Feature	Description
Web Dashboard	Implement a simple web dashboard for monitoring tasks. Also there will be some management capability (start/stop a task, change some execution parameters)
WebApi management	Webapi endpoints to list and manage tasks execution in EverTask remotely
Recurring Tasks	Recurring tasks using cron expressions or fluent builder (cron already implemented in the recurring-cron branch).
Support for new monitoring Options	Email alerts, application insights integration, open telemetry integration, ecc..
Support for new Storage Options	Considering the inclusion of additional storage options like Redis, MySql, Postgres, and various DocumentDBs initially supported by EfCore, with the possibility of expanding to other databases.
Queue customization	Create custom queues (each with his own, custom, degree of parallelism) to split task execution (for example by priority)
Clustering tasks	I'm toying with the idea to allow multiple server running evertask to create a simple cluster for tasks execution, with rules like loading balance, fail-over
Improving documentation	docs needs more love...

 

🌟 Acknowledgements

Special thanks to jbogard for the MediaTr project, providing significant inspiration in the development of key components of this library, especially in the creation of:

TaskDispatcher.cs

TaskHandlerExecutor.cs

TaskHandlerWrapper.cs

HandlerRegistrar.cs

I have included comments within these files to acknowledge and reference the specific parts of the MediaTr project that inspired them.

Their approach and architecture have been instrumental in shaping the functionality and design of these elements.

This project includes code from MediatR, which is licensed under the Apache 2.0 License. The full text of the license can be found in the LICENSE file.