worker-services

.NET Worker Services

Trigger On

• building long-running background services or scheduled workers
• adding hosted services to an app or extracting them into a worker process
• reviewing graceful shutdown, cancellation, queue processing, or health behavior

Documentation

• Worker Services in .NET
• Background tasks with hosted services in ASP.NET Core
• Create Windows Service using BackgroundService
• App health checks in .NET
• Health checks in ASP.NET Core

References

• patterns.md - BackgroundService patterns, graceful shutdown, and health check implementations
• anti-patterns.md - Common worker service mistakes and how to avoid them

Workflow

1. Use BackgroundService as your base class:

   • Provides standard StartAsync/StopAsync handling
   • Focus on implementing ExecuteAsync only
   • Proper cancellation token management built-in

2. Handle scoped dependencies correctly:

   • Create service scopes for scoped services
   • No scope is created by default in hosted services

3. Implement graceful shutdown:

   • Propagate cancellation tokens throughout
   • Complete work promptly when token fires
   • Avoid ungraceful shutdown at timeout

4. Keep execution loop thin:

   • Move business logic to testable services
   • Handle exceptions to prevent service crashes
   • Use PeriodicTimer for scheduled work

5. Add observability:

   • Use health checks for readiness/liveness
   • Expose metrics and structured logging
   • Consider distributed locks for multi-instance

Current Upstream Notes

• .NET runtime v10.0.11 is servicing. For workers, rerun cgroup-v2 memory-limit detection, cancellation, graceful shutdown, long-running thread-static and GC paths, NativeAOT/WASM startup where relevant, and Windows directory enumeration after upgrading rather than changing architecture by default.
• Use the refreshed Worker Services and hosted-service Learn pages for exact current hosting and health-check APIs when adding new worker entry points.

Basic BackgroundService Pattern

Simple Worker

public class Worker : BackgroundService
{
    private readonly ILogger<Worker> _logger;

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

    protected override async Task ExecuteAsync(CancellationToken stoppingToken)
    {
        _logger.LogInformation("Worker starting");

        while (!stoppingToken.IsCancellationRequested)
        {
            try
            {
                _logger.LogInformation("Worker running at: {Time}", DateTimeOffset.Now);
                await DoWorkAsync(stoppingToken);
                await Task.Delay(TimeSpan.FromSeconds(5), stoppingToken);
            }
            catch (OperationCanceledException) when (stoppingToken.IsCancellationRequested)
            {
                // Graceful shutdown, not an error
                break;
            }
            catch (Exception ex)
            {
                _logger.LogError(ex, "Error in worker iteration");
                // Continue or break based on error severity
            }
        }

        _logger.LogInformation("Worker stopping");
    }

    private async Task DoWorkAsync(CancellationToken cancellationToken)
    {
        // Business logic here
    }
}

Using PeriodicTimer (Recommended)

public class TimedWorker : BackgroundService
{
    private readonly ILogger<TimedWorker> _logger;
    private readonly IServiceScopeFactory _scopeFactory;
    private readonly TimeSpan _period = TimeSpan.FromMinutes(1);

    public TimedWorker(ILogger<TimedWorker> logger, IServiceScopeFactory scopeFactory)
    {
        _logger = logger;
        _scopeFactory = scopeFactory;
    }

    protected override async Task ExecuteAsync(CancellationToken stoppingToken)
    {
        using var timer = new PeriodicTimer(_period);

        while (await timer.WaitForNextTickAsync(stoppingToken))
        {
            try
            {
                await using var scope = _scopeFactory.CreateAsyncScope();
                var processor = scope.ServiceProvider.GetRequiredService<IDataProcessor>();
                await processor.ProcessAsync(stoppingToken);
            }
            catch (Exception ex)
            {
                _logger.LogError(ex, "Error processing scheduled task");
            }
        }
    }
}

Handling Scoped Dependencies

Correct Pattern with Scope Factory

public class ScopedWorker : BackgroundService
{
    private readonly IServiceScopeFactory _scopeFactory;
    private readonly ILogger<ScopedWorker> _logger;

    public ScopedWorker(IServiceScopeFactory scopeFactory, ILogger<ScopedWorker> logger)
    {
        _scopeFactory = scopeFactory;
        _logger = logger;
    }

    protected override async Task ExecuteAsync(CancellationToken stoppingToken)
    {
        while (!stoppingToken.IsCancellationRequested)
        {
            // Create scope for each unit of work
            await using var scope = _scopeFactory.CreateAsyncScope();

            var dbContext = scope.ServiceProvider.GetRequiredService<AppDbContext>();
            var service = scope.ServiceProvider.GetRequiredService<IScopedService>();

            await service.ProcessAsync(dbContext, stoppingToken);
            await Task.Delay(TimeSpan.FromSeconds(10), stoppingToken);
        }
    }
}

Queue Processing Pattern

Message Queue Worker

public class QueueWorker : BackgroundService
{
    private readonly ILogger<QueueWorker> _logger;
    private readonly IServiceScopeFactory _scopeFactory;
    private readonly IBackgroundTaskQueue _taskQueue;

    public QueueWorker(
        ILogger<QueueWorker> logger,
        IServiceScopeFactory scopeFactory,
        IBackgroundTaskQueue taskQueue)
    {
        _logger = logger;
        _scopeFactory = scopeFactory;
        _taskQueue = taskQueue;
    }

    protected override async Task ExecuteAsync(CancellationToken stoppingToken)
    {
        _logger.LogInformation("Queue Worker started");

        while (!stoppingToken.IsCancellationRequested)
        {
            var workItem = await _taskQueue.DequeueAsync(stoppingToken);

            try
            {
                await using var scope = _scopeFactory.CreateAsyncScope();
                await workItem(scope.ServiceProvider, stoppingToken);
            }
            catch (Exception ex)
            {
                _logger.LogError(ex, "Error processing queued work item");
                // Handle poison message - retry, dead-letter, etc.
            }
        }
    }
}

// Task queue interface
public interface IBackgroundTaskQueue
{
    ValueTask QueueBackgroundWorkItemAsync(
        Func<IServiceProvider, CancellationToken, ValueTask> workItem);

    ValueTask<Func<IServiceProvider, CancellationToken, ValueTask>> DequeueAsync(
        CancellationToken cancellationToken);
}

Health Checks for Workers

Adding Health Check Endpoint

// Program.cs
var builder = Host.CreateApplicationBuilder(args);

builder.Services.AddHostedService<Worker>();

// Add health checks
builder.Services.AddHealthChecks()
    .AddCheck<WorkerHealthCheck>("worker_health")
    .AddResourceUtilizationHealthCheck();

// Add HTTP endpoint for health checks
builder.Services.AddHealthChecksUI();

// Or use simple TCP listener for Kubernetes
builder.Services.AddSingleton<TcpHealthProbeService>();
builder.Services.AddHostedService(sp => sp.GetRequiredService<TcpHealthProbeService>());

var host = builder.Build();
host.Run();

Custom Health Check

public class WorkerHealthCheck : IHealthCheck
{
    private readonly WorkerState _workerState;

    public WorkerHealthCheck(WorkerState workerState)
    {
        _workerState = workerState;
    }

    public Task<HealthCheckResult> CheckHealthAsync(
        HealthCheckContext context,
        CancellationToken cancellationToken = default)
    {
        if (_workerState.LastSuccessfulRun > DateTime.UtcNow.AddMinutes(-5))
        {
            return Task.FromResult(HealthCheckResult.Healthy(
                $"Last successful run: {_workerState.LastSuccessfulRun}"));
        }

        return Task.FromResult(HealthCheckResult.Unhealthy(
            $"No successful run since: {_workerState.LastSuccessfulRun}"));
    }
}

// Shared state
public class WorkerState
{
    public DateTime LastSuccessfulRun { get; set; } = DateTime.UtcNow;
    public bool IsProcessing { get; set; }
}

Graceful Shutdown Pattern

Proper Shutdown Handling

public class GracefulWorker : BackgroundService
{
    private readonly ILogger<GracefulWorker> _logger;
    private int _currentWorkItemId;

    protected override async Task ExecuteAsync(CancellationToken stoppingToken)
    {
        _logger.LogInformation("Worker starting");

        while (!stoppingToken.IsCancellationRequested)
        {
            _currentWorkItemId = GetNextWorkItemId();

            try
            {
                // Pass cancellation token to all async operations
                await ProcessWorkItemAsync(_currentWorkItemId, stoppingToken);
            }
            catch (OperationCanceledException) when (stoppingToken.IsCancellationRequested)
            {
                _logger.LogInformation(
                    "Shutdown requested, stopping after work item {Id}", _currentWorkItemId);
                break;
            }
        }
    }

    public override async Task StopAsync(CancellationToken cancellationToken)
    {
        _logger.LogInformation("Worker stopping gracefully");
        await base.StopAsync(cancellationToken);
        _logger.LogInformation("Worker stopped");
    }
}

Windows Service Deployment

Configuring as Windows Service

// Program.cs
var builder = Host.CreateApplicationBuilder(args);

builder.Services.AddWindowsService(options =>
{
    options.ServiceName = "My Worker Service";
});

builder.Services.AddHostedService<Worker>();

var host = builder.Build();
host.Run();

Project File Settings

<Project Sdk="Microsoft.NET.Sdk.Worker">
  <PropertyGroup>
    <TargetFramework>net8.0</TargetFramework>
    <RuntimeIdentifier>win-x64</RuntimeIdentifier>
    <PublishSingleFile>true</PublishSingleFile>
    <SelfContained>true</SelfContained>
  </PropertyGroup>
</Project>

Best Practices

1. Use BackgroundService as base class - Handles StartAsync/StopAsync boilerplate and cancellation management
2. Create scopes for scoped dependencies - Use IServiceScopeFactory to resolve scoped services like DbContext
3. Propagate cancellation tokens everywhere - Pass to all async methods for responsive shutdown
4. Wrap work in try-catch - Unhandled exceptions stop the service completely
5. Use PeriodicTimer for timed tasks - Cleaner than Task.Delay with proper cancellation support
6. Add health checks - Essential for Kubernetes liveness/readiness probes
7. Avoid blocking StartAsync - Long initialization delays other hosted services
8. Call base methods when overriding - Always call await base.StartAsync() and await base.StopAsync()
9. Publish as single file for Windows Service - Reduces deployment complexity and errors
10. Consider scaling requirements - Separate worker projects if independent scaling is needed

Anti-Patterns to Avoid

Anti-Pattern	Why It's Bad	Better Approach
Ad-hoc while(true) loops	No graceful shutdown, poor lifecycle	Use BackgroundService
Ignoring cancellation token	Ungraceful shutdown, resource leaks	Propagate token to all async calls
Injecting scoped services directly	Captive dependencies, memory leaks	Use IServiceScopeFactory
Unhandled exceptions in ExecuteAsync	Silently stops the worker	Wrap in try-catch, log, continue
Long-running StartAsync	Blocks other services from starting	Move work to ExecuteAsync
async void methods	Crashes process on exception	Use async Task
Missing health checks	No visibility into worker status	Implement IHealthCheck
Polling with tight loops	CPU waste, no responsiveness	Use PeriodicTimer or event-driven
Not overriding StopAsync	Missed cleanup opportunity	Override for graceful cleanup
Singleton DbContext	Not thread-safe, stale data	Create scopes per operation

Deliver

• well-behaved worker processes and hosted services
• predictable startup and shutdown behavior
• proper scoped dependency handling
• health checks for production observability
• retry and poison-message handling for queue work

Validate

• cancellation token propagated and shutdown honored
• scoped services resolved within proper scopes
• exception handling prevents service crashes
• health checks report accurate worker status
• runtime behavior visible through logs or telemetry
• no blocking calls in async context