add server timeout functionality plus docker test

DotPulsar.sln

@@ -26,6 +26,8 @@ Project("{2150E333-8FDC-42A3-9474-1A3956D46DE8}") = "Solution Items", "Solution
 EndProject
 Project("{9A19103F-16F7-4668-BE54-9A1E7A4F7556}") = "Processing", "samples\Processing\Processing.csproj", "{CC1494FA-4EB5-4DB9-8BE9-0A6E8D0D963E}"
 EndProject
+Project("{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}") = "DotPulsar.Consumer", "tests\DotPulsar.Consumer\DotPulsar.Consumer.csproj", "{36E6E6EF-A471-4AE4-B696-1C9DAAFA2770}"
+EndProject
 Global
 	GlobalSection(SolutionConfigurationPlatforms) = preSolution
 		Debug|Any CPU = Debug|Any CPU
@@ -56,6 +58,10 @@ Global
 		{CC1494FA-4EB5-4DB9-8BE9-0A6E8D0D963E}.Debug|Any CPU.Build.0 = Debug|Any CPU
 		{CC1494FA-4EB5-4DB9-8BE9-0A6E8D0D963E}.Release|Any CPU.ActiveCfg = Release|Any CPU
 		{CC1494FA-4EB5-4DB9-8BE9-0A6E8D0D963E}.Release|Any CPU.Build.0 = Release|Any CPU
+		{36E6E6EF-A471-4AE4-B696-1C9DAAFA2770}.Debug|Any CPU.ActiveCfg = Debug|Any CPU
+		{36E6E6EF-A471-4AE4-B696-1C9DAAFA2770}.Debug|Any CPU.Build.0 = Debug|Any CPU
+		{36E6E6EF-A471-4AE4-B696-1C9DAAFA2770}.Release|Any CPU.ActiveCfg = Release|Any CPU
+		{36E6E6EF-A471-4AE4-B696-1C9DAAFA2770}.Release|Any CPU.Build.0 = Release|Any CPU
 	EndGlobalSection
 	GlobalSection(SolutionProperties) = preSolution
 		HideSolutionNode = FALSE
@@ -66,6 +72,7 @@ Global
 		{14934BED-A222-47B2-A58A-CFC4AAB89B49} = {E7106D0F-B255-4631-9FB8-734FC5748FA9}
 		{6D44683B-865C-4D15-9F0A-1A8441354589} = {E7106D0F-B255-4631-9FB8-734FC5748FA9}
 		{CC1494FA-4EB5-4DB9-8BE9-0A6E8D0D963E} = {E7106D0F-B255-4631-9FB8-734FC5748FA9}
+		{36E6E6EF-A471-4AE4-B696-1C9DAAFA2770} = {E1C932A9-6D4C-4DDF-8922-BE7B71F12F1C}
 	EndGlobalSection
 	GlobalSection(ExtensibilityGlobals) = postSolution
 		SolutionGuid = {88355922-E70A-4B73-B7F8-ABF8F2B59789}

src/DotPulsar/Abstractions/IPulsarClientBuilder.cs

@@ -53,6 +53,16 @@ public interface IPulsarClientBuilder
     /// </summary>
     IPulsarClientBuilder KeepAliveInterval(TimeSpan interval);
 
+    /// <summary>
+    /// The maximum amount of time to wait without receiving any message from the server at
+    /// which point the connection is assumed to be dead or the server is not responding.
+    /// As we are sending pings the server should respond to those at a minimum within this specified timeout period.
+    /// Once this happens the connection will be torn down and all consumers/producers will enter
+    /// the disconnected state and attempt to reconnect
+    /// The default is 60 seconds.
+    /// </summary>
+    IPulsarClientBuilder ServerResponseTimeout(TimeSpan interval);
+
     /// <summary>
     /// Set the listener name. This is optional.
     /// </summary>

src/DotPulsar/Internal/Connection.cs

@@ -33,11 +33,15 @@ public sealed class Connection : IConnection
     private readonly IAuthentication? _authentication;
     private int _isDisposed;
 
-    public Connection(IPulsarStream stream, TimeSpan keepAliveInterval, IAuthentication? authentication)
+    public Connection(
+        IPulsarStream stream,
+        TimeSpan keepAliveInterval,
+        TimeSpan serverResponseTimeout,
+        IAuthentication? authentication)
     {
         _lock = new AsyncLock();
         _channelManager = new ChannelManager();
-        _pingPongHandler = new PingPongHandler(this, keepAliveInterval);
+        _pingPongHandler = new PingPongHandler(this, keepAliveInterval, serverResponseTimeout);
         _stream = stream;
         _authentication = authentication;
     }
@@ -294,6 +298,11 @@ private async Task Send(BaseCommand command, CancellationToken cancellationToken
     }
 
     public async Task ProcessIncommingFrames(CancellationToken cancellationToken)
+    {
+        await Task.WhenAny(ProcessIncommingFramesImpl(cancellationToken), _pingPongHandler.ServerNotResponding).ConfigureAwait(false);
+    }
+
+    public async Task ProcessIncommingFramesImpl(CancellationToken cancellationToken)
     {
         await Task.Yield();
 

src/DotPulsar/Internal/ConnectionPool.cs

@@ -38,6 +38,7 @@ public sealed class ConnectionPool : IConnectionPool
     private readonly string? _listenerName;
     private readonly TimeSpan _keepAliveInterval;
     private readonly IAuthentication? _authentication;
+    private readonly TimeSpan _serverResponseTimeout;
 
     public ConnectionPool(
         CommandConnect commandConnect,
@@ -47,6 +48,7 @@ public ConnectionPool(
         TimeSpan closeInactiveConnectionsInterval,
         string? listenerName,
         TimeSpan keepAliveInterval,
+        TimeSpan serverResponseTimeout,
         IAuthentication? authentication)
     {
         _lock = new AsyncLock();
@@ -59,6 +61,7 @@ public ConnectionPool(
         _cancellationTokenSource = new CancellationTokenSource();
         _closeInactiveConnections = CloseInactiveConnections(closeInactiveConnectionsInterval, _cancellationTokenSource.Token);
         _keepAliveInterval = keepAliveInterval;
+        _serverResponseTimeout = serverResponseTimeout;
         _authentication = authentication;
     }
 
@@ -159,7 +162,8 @@ private async ValueTask<Connection> GetConnection(PulsarUrl url, CancellationTok
     private async Task<Connection> EstablishNewConnection(PulsarUrl url, CancellationToken cancellationToken)
     {
         var stream = await _connector.Connect(url.Physical).ConfigureAwait(false);
-        var connection = new Connection(new PulsarStream(stream), _keepAliveInterval, _authentication);
+        var connection = new Connection(new PulsarStream(stream), _keepAliveInterval, _serverResponseTimeout,
+            _authentication);
         DotPulsarMeter.ConnectionCreated();
         _connections[url] = connection;
         _ = connection.ProcessIncommingFrames(_cancellationTokenSource.Token).ContinueWith(t => DisposeConnection(url));

src/DotPulsar/Internal/DotPulsarMeter.cs

@@ -29,6 +29,7 @@ public static class DotPulsarMeter
     private static int _numberOfReaders;
     private static int _numberOfConsumers;
     private static int _numberOfProducers;
+    private static int _numberOfServerTimeouts;
 #pragma warning restore IDE0044
 #pragma warning restore IDE0079
     private static readonly Histogram<double> _producerSendDuration;
@@ -42,6 +43,7 @@ static DotPulsarMeter()
         _ = Meter.CreateObservableGauge("dotpulsar.reader.count", GetNumberOfReaders, "{readers}", "Number of readers");
         _ = Meter.CreateObservableGauge("dotpulsar.consumer.count", GetNumberOfConsumers, "{consumers}", "Number of consumers");
         _ = Meter.CreateObservableGauge("dotpulsar.producer.count", GetNumberOfProducers, "{producers}", "Number of producers");
+        _ = Meter.CreateObservableGauge("dotpulsar.server.timeout.count", GetNumberOfProducers, "{servertimeout}", "Number of times server stopped responding");
         _producerSendDuration = Meter.CreateHistogram<double>("dotpulsar.producer.send.duration", "ms", "Measures the duration for sending a message");
         _consumerProcessDuration = Meter.CreateHistogram<double>("dotpulsar.consumer.process.duration", "ms", "Measures the duration for processing a message");
     }
@@ -54,6 +56,7 @@ static DotPulsarMeter()
 
     public static void ConnectionCreated() => Interlocked.Increment(ref _numberOfConnections);
     public static void ConnectionDisposed() => Interlocked.Decrement(ref _numberOfConnections);
+    public static void ServerTimedout() => Interlocked.Decrement(ref _numberOfServerTimeouts);
     private static int GetNumberOfConnections() => Volatile.Read(ref _numberOfConnections);
 
     public static void ReaderCreated() => Interlocked.Increment(ref _numberOfReaders);
@@ -67,6 +70,7 @@ static DotPulsarMeter()
     public static void ProducerCreated() => Interlocked.Increment(ref _numberOfProducers);
     public static void ProducerDisposed() => Interlocked.Decrement(ref _numberOfProducers);
     private static int GetNumberOfProducers() => Volatile.Read(ref _numberOfProducers);
+    private static int GetNumberOfServerTimeouts() => Volatile.Read(ref _numberOfServerTimeouts);
 
 
     public static bool MessageSentEnabled => _producerSendDuration.Enabled;

src/DotPulsar/Internal/PingPongHandler.cs

@@ -25,29 +25,37 @@ public sealed class PingPongHandler : IAsyncDisposable
 {
     private readonly IConnection _connection;
     private readonly TimeSpan _keepAliveInterval;
+    private readonly TimeSpan _serverResponseTimeout;
     private readonly Timer _timer;
     private readonly CommandPing _ping;
     private readonly CommandPong _pong;
     private long _lastCommand;
+    private readonly TaskCompletionSource<object> _serverNotRespondingTcs;
 
-    public PingPongHandler(IConnection connection, TimeSpan keepAliveInterval)
+    public PingPongHandler(IConnection connection,
+        TimeSpan keepAliveInterval,
+        TimeSpan serverResponseTimeout)
     {
         _connection = connection;
         _keepAliveInterval = keepAliveInterval;
+        _serverResponseTimeout = serverResponseTimeout;
         _timer = new Timer(Watch);
         _timer.Change(_keepAliveInterval, TimeSpan.Zero);
         _ping = new CommandPing();
         _pong = new CommandPong();
         _lastCommand = Stopwatch.GetTimestamp();
+        _serverNotRespondingTcs = new TaskCompletionSource<object>();
     }
 
+    public Task ServerNotResponding => _serverNotRespondingTcs.Task;
+
     public bool Incoming(BaseCommand.Type commandType)
     {
         Interlocked.Exchange(ref _lastCommand, Stopwatch.GetTimestamp());
 
         if (commandType == BaseCommand.Type.Ping)
         {
-            Task.Factory.StartNew(() => SendPong());
+            Task.Factory.StartNew(SendPong);
             return true;
         }
 
@@ -61,13 +69,24 @@ private void Watch(object? state)
             var lastCommand = Interlocked.Read(ref _lastCommand);
             var now = Stopwatch.GetTimestamp();
             var elapsed = TimeSpan.FromSeconds((now - lastCommand) / Stopwatch.Frequency);
+
+            if (elapsed > _serverResponseTimeout)
+            {
+                DotPulsarMeter.ServerTimedout();
+                _serverNotRespondingTcs.SetResult(new object());
+                return;
+            }
+
             if (elapsed >= _keepAliveInterval)
             {
                 Task.Factory.StartNew(() => SendPing());
                 _timer.Change(_keepAliveInterval, TimeSpan.Zero);
             }
             else
-                _timer.Change(_keepAliveInterval.Subtract(elapsed), TimeSpan.Zero);
+            {
+                var result = _keepAliveInterval.Subtract(elapsed);
+                _timer.Change(result, TimeSpan.Zero);
+            }
         }
         catch
         {

src/DotPulsar/Internal/PulsarClientBuilder.cs

@@ -27,6 +27,7 @@ public sealed class PulsarClientBuilder : IPulsarClientBuilder
     private readonly List<IHandleException> _exceptionHandlers;
     private EncryptionPolicy? _encryptionPolicy;
     private TimeSpan _keepAliveInterval;
+    private TimeSpan _serverResponseTimeout;
     private string? _listenerName;
     private TimeSpan _retryInterval;
     private Uri _serviceUrl;
@@ -51,6 +52,7 @@ public PulsarClientBuilder()
 
         _exceptionHandlers = new List<IHandleException>();
         _keepAliveInterval = TimeSpan.FromSeconds(30);
+        _serverResponseTimeout = TimeSpan.FromSeconds(60);
         _retryInterval = TimeSpan.FromSeconds(3);
         _serviceUrl = new Uri($"{Constants.PulsarScheme}://localhost:{Constants.DefaultPulsarPort}");
         _clientCertificates = new X509Certificate2Collection();
@@ -96,6 +98,12 @@ public IPulsarClientBuilder KeepAliveInterval(TimeSpan interval)
         return this;
     }
 
+    public IPulsarClientBuilder ServerResponseTimeout(TimeSpan interval)
+    {
+        _serverResponseTimeout = interval;
+        return this;
+    }
+
     public IPulsarClientBuilder ListenerName(string listenerName)
     {
         _listenerName = listenerName;
@@ -165,7 +173,8 @@ public IPulsarClient Build()
 
         var exceptionHandlers = new List<IHandleException>(_exceptionHandlers) { new DefaultExceptionHandler(_retryInterval) };
         var exceptionHandlerPipeline = new ExceptionHandlerPipeline(exceptionHandlers);
-        var connectionPool = new ConnectionPool(_commandConnect, _serviceUrl, connector, _encryptionPolicy.Value, _closeInactiveConnectionsInterval, _listenerName, _keepAliveInterval, _authentication);
+
+        var connectionPool = new ConnectionPool(_commandConnect, _serviceUrl, connector, _encryptionPolicy.Value, _closeInactiveConnectionsInterval, _listenerName, _keepAliveInterval, _serverResponseTimeout, _authentication);
         var processManager = new ProcessManager(connectionPool);
 
         return new PulsarClient(connectionPool, processManager, exceptionHandlerPipeline, _serviceUrl);

tests/DotPulsar.Consumer/DotPulsar.Consumer.csproj

@@ -0,0 +1,26 @@
+<Project Sdk="Microsoft.NET.Sdk">
+
+    <PropertyGroup>
+      <TargetFramework>net6.0</TargetFramework>
+      <ImplicitUsings>enable</ImplicitUsings>
+      <OutputType>Exe</OutputType>
+      <Nullable>enable</Nullable>
+      <PublishSingleFile>true</PublishSingleFile>
+      <DebugType>embedded</DebugType>
+      <RuntimeIdentifier>linux-x64</RuntimeIdentifier>
+    </PropertyGroup>
+
+    <ItemGroup>
+      <ProjectReference Include="..\..\src\DotPulsar\DotPulsar.csproj" />
+    </ItemGroup>
+
+  <Target Name="PublishConsumer" AfterTargets="Build">
+    <Exec Command="dotnet publish $(ProjectDir) -r linux-x64 -o $(ProjectDir)..\DotPulsar.Tests\$(OutDir) --no-build"
+          ConsoleToMSBuild="true" IgnoreExitCode="true" IgnoreStandardErrorWarningFormat="true">
+      <Output TaskParameter="ConsoleOutput" PropertyName="ConsoleOutput" />
+      <Output TaskParameter="ExitCode" PropertyName="ExitCode" />
+    </Exec>
+
+    <Error Condition="'$(ExitCode)' != '0'" Text="$(ConsoleOutput)"  />
+  </Target>
+</Project>

tests/DotPulsar.Consumer/Program.cs

@@ -0,0 +1,95 @@
+namespace DotPulsar.Consumer;
+
+using Abstractions;
+using Extensions;
+using Internal;
+using System.Net.Sockets;
+
+public static class Program
+{
+    public const string ConnectedAwaitingDisconnection = "Successfully connected, awaiting disconnection";
+    public const string DisconnectedAwaitingReconnection = "Successfully disconnected, awaiting reconnection";
+    public const string Reconnected = "Successfully reconnected";
+
+    public static async Task<int> Main(string[] args)
+    {
+        Console.WriteLine($"Running with url {args[0]}");
+        var client = CreateClient(args[0], args[1]);
+
+        var testRunId = Guid.NewGuid().ToString("N");
+
+        var topic = $"persistent://public/default/consumer-tests-{testRunId}";
+        await using var consumer = client.NewConsumer(Schema.ByteArray)
+            .ConsumerName($"consumer-{testRunId}")
+            .InitialPosition(SubscriptionInitialPosition.Earliest)
+            .SubscriptionName($"subscription-{testRunId}")
+            .StateChangedHandler(changed => Console.WriteLine($"Consumer state changed to {changed.ConsumerState}"))
+            .Topic(topic)
+            .Create();
+
+        var timeout = Task.Delay(TimeSpan.FromSeconds(30));
+        var result = await Task.WhenAny(timeout, consumer.StateChangedTo(ConsumerState.Active).AsTask());
+
+        if (result == timeout)
+        {
+            Console.WriteLine("Timed out waiting for active status");
+            return -1;
+        }
+
+        Console.WriteLine(ConnectedAwaitingDisconnection);
+        timeout = Task.Delay(TimeSpan.FromSeconds(30));
+        result = await Task.WhenAny(timeout, consumer.StateChangedTo(ConsumerState.Disconnected).AsTask());
+
+        if (result == timeout)
+        {
+            Console.WriteLine("Timed out waiting for disconnected status");
+            return -2;
+        }
+
+        Console.WriteLine(DisconnectedAwaitingReconnection);
+        timeout = Task.Delay(TimeSpan.FromSeconds(30));
+        result = await Task.WhenAny(timeout, consumer.StateChangedTo(ConsumerState.Active).AsTask());
+
+        if (result == timeout)
+        {
+            Console.WriteLine("Timed out waiting for reconnected status");
+            return -3;
+        }
+
+        Console.WriteLine(Reconnected);
+        return 0;
+    }
+
+    private static IPulsarClient CreateClient(string url, string token)
+        => PulsarClient
+            .Builder()
+            .Authentication(AuthenticationFactory.Token(_ => ValueTask.FromResult(token)))
+            .KeepAliveInterval(TimeSpan.FromSeconds(5))
+            .ServerResponseTimeout(TimeSpan.FromSeconds(10))
+            .ExceptionHandler(new TestExceptionHandler())
+            .ServiceUrl(new Uri(url))
+            .Build();
+}
+
+internal class TestExceptionHandler : IHandleException
+{
+    private readonly IHandleException _inner = new DefaultExceptionHandler(TimeSpan.FromSeconds(5));
+
+    public ValueTask OnException(ExceptionContext exceptionContext)
+    {
+        Console.WriteLine("Exception occurred: {0}", exceptionContext.Exception.Message);
+
+        // This occurs when reconnecting the docker network, it takes some time for the alias to be reestablished
+        if (exceptionContext.Exception is SocketException se && se.Message.Contains("Name or service not known"))
+        {
+            exceptionContext.ExceptionHandled = true;
+            exceptionContext.Result = FaultAction.Retry;
+        }
+        else
+        {
+            _inner.OnException(exceptionContext);
+        }
+
+        return ValueTask.CompletedTask;
+    }
+}

tests/DotPulsar.Tests/DotPulsar.Tests.csproj

@@ -23,6 +23,7 @@
 
   <ItemGroup>
     <ProjectReference Include="..\..\src\DotPulsar\DotPulsar.csproj" />
+    <ProjectReference Include="..\DotPulsar.Consumer\DotPulsar.Consumer.csproj" />
   </ItemGroup>
 
   <ItemGroup>

tests/DotPulsar.Tests/IntegrationCollection.cs

@@ -18,3 +18,6 @@ namespace DotPulsar.Tests;
 
 [CollectionDefinition("Integration")]
 public class IntegrationCollection : ICollectionFixture<IntegrationFixture> { }
+
+[CollectionDefinition("KeepAlive")]
+public class KeepAliveCollection : ICollectionFixture<KeepAliveFixture> { }