Lifecycle of Worker Roles (RoleEntryPoint)

In this post we will talk about 3 methods that are available in the RoleEntryPoint class for Web and Worker Roles (especially Worker Roles)

• OnStart
• Run
• OnStop

This 3 methods are called in different moment of the lifecycle of a Worker Role. Each of them has a clear scope and output.

OnStart
This method is called automatically when the instance of Worker Role is initialized. It is used to initialized the context or prepare the instance before executing the task.
This method returns a bool value that it is used to tell the system if the initialization was made with success or not. On the happy case the return value should be true. If an error occurs or the initialization fails, that the false value should be returned. When OnStart methods returns false, the Run method will not be called and the instance will be 'restarted' - No other methods will be called.

Run
The Run method is used to execute the batch operation or the logic that is necessary to be executed by the Worker Role. We should 'block' the code to not exist from this method as long as we want to keep our instance alive.
This method is called only when OnStart methods returns true. When we exit from the Run method, the OnStop method will be called and worker role will be restarted.

OnStop
This method is used to make the cleanup sequence of our system. For example deallocate resources or commit some changes or logs.
This method is called only when the no exception is triggered by Run or OnStart method.

Below you can find the lifecycle flow.

Let's take a look on the following code:

    public class WorkerRole : RoleEntryPoint
    {
        private readonly CancellationTokenSource cancellationTokenSource = new CancellationTokenSource();
        private readonly ManualResetEvent runCompleteEvent = new ManualResetEvent(false);

        public override void Run()
        {
            Logger.Debug(LoggingAction.StartingWorkerRole, "Foo.Workers.SU is running");

            try
            {
                RunAsync(cancellationTokenSource.Token).Wait();
            }
            finally
            {
                runCompleteEvent.Set();
            }
        }

        public override bool OnStart()
        {
            bool reFoolt = base.OnStart();

            Logger.Debug(LoggingAction.StartingWorkerRole, "Foo.Workers.SU has been started");

            ObjectFactory.InitializeContainer(new DependencyInstaller());

            var rp = ObjectFactory.GetObject<IPRP>();            
            rp.StartListeningAsync().Wait();            

            return reFoolt;
        }

        public override void OnStop()
        {
            Logger.Debug(LoggingAction.StoppingWorkerRole, "Foo.Workers.SU is stopping");

            cancellationTokenSource.Cancel();
            runCompleteEvent.WaitOne();

            base.OnStop();

            Logger.Debug(LoggingAction.StoppingWorkerRole, "Foo.Workers.SU has stopped");
        }

        private async Task RunAsync(CancellationToken cancellationToken)
        {
            while (!cancellationToken.IsCancellationRequested)
            {                
                await Task.Delay(1000);
            }
        }
    }

Looking over the above code, we observe that no logic is done on Rum method. Not only we initialize the object factory but we also start listening to a queue in the OnStart method. Because of this the Run method is not used and our Worker Role will be in a 'transition' phase all the time.
What we should keep in OnStart method? The initialization of object factory can be keeped in this method. Also, if we would have some initialization on the logger side, that we should keep it in this place.
What we should not keep in OnStart method? We should move the start listening to our our queue from OnStart method to Run. That is a piece of code that is long running and should be in the Run.

    public class WorkerRole : RoleEntryPoint
    {
        private readonly CancellationTokenSource cancellationTokenSource = new CancellationTokenSource();
        private readonly ManualResetEvent runCompleteEvent = new ManualResetEvent(false);

        public override void Run()
        {
            Logger.Debug(LoggingAction.StartingWorkerRole, "Foo.Workers.SU is running");

            try
            {
                RunAsync(cancellationTokenSource.Token).Wait();
            }
            finally
            {
                runCompleteEvent.Set();
            }
        }

        public override bool OnStart()
        {
            bool reFoolt = base.OnStart();
      
            Logger.Debug(LoggingAction.StartingWorkerRole, "Foo.Workers.SU has been started");
      
      try
      {
        ObjectFactory.InitializeContainer(new DependencyInstaller());            
      }
      catch(Exception ex)
      {
        Logger.Exception(ex);
        reFoolt = false;
      }

            return reFoolt;
        }
    
    public void Run()
    {
      var rp = ObjectFactory.GetObject<IPRP>();            
            rp.StartListeningAsync().Wait();            
    }
    

        public override void OnStop()
        {
            Logger.Debug(LoggingAction.StoppingWorkerRole, "Foo.Workers.SU is stopping");

            cancellationTokenSource.Cancel();
            runCompleteEvent.WaitOne();

            base.OnStop();

            Logger.Debug(LoggingAction.StoppingWorkerRole, "Foo.Workers.SU has stopped");
        }

        private async Task RunAsync(CancellationToken cancellationToken)
        {
            while (!cancellationToken.IsCancellationRequested)
            {                
                await Task.Delay(1000);
            }
        }
    }

In conclusion, there are two important things that we should remember:

• OnStart should only initialize and configure our system
• Run method is used for processing and log running tasks