Merge branch 'main' into wangbill/largepayloadv2

Author: YunchuWang

CHANGELOG.md

@@ -1,5 +1,8 @@
 # Changelog
 
+## v1.14.0
+- Add RestartAsync API Support at DurableTaskClient ([#456](https://github.com/microsoft/durabletask-dotnet/pull/456))
+
 ## v1.13.0
 - Add orchestration execution tracing ([#441](https://github.com/microsoft/durabletask-dotnet/pull/441))
 

eng/targets/Release.props

@@ -17,7 +17,7 @@
   </PropertyGroup>
 
   <PropertyGroup>
-    <VersionPrefix>1.13.0</VersionPrefix>
+    <VersionPrefix>1.14.0</VersionPrefix>
     <VersionSuffix></VersionSuffix>
   </PropertyGroup>
 

src/Client/Core/DurableTaskClient.cs

@@ -399,6 +399,47 @@ public virtual Task<PurgeResult> PurgeAllInstancesAsync(
         throw new NotSupportedException($"{this.GetType()} does not support purging of orchestration instances.");
     }
 
+    /// <summary>
+    /// Restarts an orchestration instance with the same or a new instance ID.
+    /// </summary>
+    /// <remarks>
+    /// <para>
+    /// This method restarts an existing orchestration instance. If <paramref name="restartWithNewInstanceId"/> is <c>true</c>,
+    /// a new instance ID will be generated for the restarted orchestration. If <c>false</c>, the original instance ID will be reused.
+    /// </para><para>
+    /// The restarted orchestration will use the same input data as the original instance. If the original orchestration
+    /// instance is not found, an <see cref="ArgumentException"/> will be thrown.
+    /// </para><para>
+    /// Note that this operation is backend-specific and may not be supported by all durable task backends.
+    /// If the backend does not support restart operations, a <see cref="NotSupportedException"/> will be thrown.
+    /// </para>
+    /// </remarks>
+    /// <param name="instanceId">The ID of the orchestration instance to restart.</param>
+    /// <param name="restartWithNewInstanceId">
+    /// If <c>true</c>, a new instance ID will be generated for the restarted orchestration.
+    /// If <c>false</c>, the original instance ID will be reused.
+    /// </param>
+    /// <param name="cancellation">
+    /// The cancellation token. This only cancels enqueueing the restart request to the backend.
+    /// Does not abort restarting the orchestration once enqueued.
+    /// </param>
+    /// <returns>
+    /// A task that completes when the orchestration instance is successfully restarted.
+    /// The value of this task is the instance ID of the restarted orchestration instance.
+    /// </returns>
+    /// <exception cref="ArgumentException">
+    /// Thrown if an orchestration with the specified <paramref name="instanceId"/> was not found. </exception>
+    /// <exception cref="InvalidOperationException">
+    /// Thrown when attempting to restart an instance using the same instance Id
+    /// while the instance has not yet reached a completed or terminal state. </exception>
+    /// <exception cref="NotSupportedException">
+    /// Thrown if the backend does not support restart operations. </exception>
+    public virtual Task<string> RestartAsync(
+        string instanceId,
+        bool restartWithNewInstanceId = false,
+        CancellationToken cancellation = default)
+        => throw new NotSupportedException($"{this.GetType()} does not support orchestration restart.");
+
     // TODO: Create task hub
 
     // TODO: Delete task hub

src/Client/Grpc/GrpcDurableTaskClient.cs

@@ -395,6 +395,42 @@ public override Task<PurgeResult> PurgeAllInstancesAsync(
         return this.PurgeInstancesCoreAsync(request, cancellation);
     }
 
+    /// <inheritdoc/>
+    public override async Task<string> RestartAsync(
+        string instanceId,
+        bool restartWithNewInstanceId = false,
+        CancellationToken cancellation = default)
+    {
+        Check.NotNullOrEmpty(instanceId);
+        Check.NotEntity(this.options.EnableEntitySupport, instanceId);
+
+        var request = new P.RestartInstanceRequest
+        {
+            InstanceId = instanceId,
+            RestartWithNewInstanceId = restartWithNewInstanceId,
+        };
+
+        try
+        {
+            P.RestartInstanceResponse result = await this.sidecarClient.RestartInstanceAsync(
+                request, cancellationToken: cancellation);
+            return result.InstanceId;
+        }
+        catch (RpcException e) when (e.StatusCode == StatusCode.NotFound)
+        {
+            throw new ArgumentException($"An orchestration with the instanceId {instanceId} was not found.", e);
+        }
+        catch (RpcException e) when (e.StatusCode == StatusCode.FailedPrecondition)
+        {
+            throw new InvalidOperationException($"An orchestration with the instanceId {instanceId} cannot be restarted.", e);
+        }
+        catch (RpcException e) when (e.StatusCode == StatusCode.Cancelled)
+        {
+            throw new OperationCanceledException(
+                $"The {nameof(this.RestartAsync)} operation was canceled.", e, cancellation);
+        }
+    }
+
     static AsyncDisposable GetCallInvoker(GrpcDurableTaskClientOptions options, out CallInvoker callInvoker)
     {
         if (options.Channel is GrpcChannel c)

src/Client/OrchestrationServiceClientShim/ShimDurableTaskClient.cs

@@ -254,6 +254,59 @@ public override async Task<OrchestrationMetadata> WaitForInstanceStartAsync(
         }
     }
 
+    /// <inheritdoc/>
+    public override async Task<string> RestartAsync(
+        string instanceId,
+        bool restartWithNewInstanceId = false,
+        CancellationToken cancellation = default)
+    {
+        Check.NotNullOrEmpty(instanceId);
+        cancellation.ThrowIfCancellationRequested();
+
+        // Get the current orchestration status to retrieve the name and input
+        OrchestrationMetadata? status = await this.GetInstanceAsync(instanceId, getInputsAndOutputs: true, cancellation);
+
+        if (status == null)
+        {
+            throw new ArgumentException($"An orchestration with the instanceId {instanceId} was not found.");
+        }
+
+        bool isInstaceNotCompleted = status.RuntimeStatus == OrchestrationRuntimeStatus.Running ||
+                                    status.RuntimeStatus == OrchestrationRuntimeStatus.Pending ||
+                                    status.RuntimeStatus == OrchestrationRuntimeStatus.Suspended;
+
+        if (isInstaceNotCompleted && !restartWithNewInstanceId)
+        {
+            throw new InvalidOperationException($"Instance '{instanceId}' cannot be restarted while it is in state '{status.RuntimeStatus}'. " +
+                   "Wait until it has completed, or restart with a new instance ID.");
+        }
+
+        // Determine the instance ID for the restarted orchestration
+        string newInstanceId = restartWithNewInstanceId ? Guid.NewGuid().ToString("N") : instanceId;
+
+        OrchestrationInstance instance = new()
+        {
+            InstanceId = newInstanceId,
+            ExecutionId = Guid.NewGuid().ToString("N"),
+        };
+
+        // Use the original serialized input directly to avoid double serialization
+        // TODO: OrchestrationMetada doesn't have version property so we don't support version here.
+        // Issue link: https://github.com/microsoft/durabletask-dotnet/issues/463
+        TaskMessage message = new()
+        {
+            OrchestrationInstance = instance,
+            Event = new ExecutionStartedEvent(-1, status.SerializedInput)
+            {
+                Name = status.Name,
+                OrchestrationInstance = instance,
+            },
+        };
+
+        await this.Client.CreateTaskOrchestrationAsync(message);
+        return newInstanceId;
+    }
+
     [return: NotNullIfNotNull("state")]
     OrchestrationMetadata? ToMetadata(Core.OrchestrationState? state, bool getInputsAndOutputs)
     {

src/Grpc/orchestrator_service.proto

@@ -482,6 +482,15 @@ message PurgeInstancesResponse {
     google.protobuf.BoolValue isComplete = 2;
 }
 
+message RestartInstanceRequest {
+    string instanceId = 1;
+    bool restartWithNewInstanceId = 2;
+}
+
+message RestartInstanceResponse {
+    string instanceId = 1;
+}
+
 message CreateTaskHubRequest {
     bool recreateIfExists = 1;
 }
@@ -682,6 +691,9 @@ service TaskHubSidecarService {
     // Rewinds an orchestration instance to last known good state and replays from there.
     rpc RewindInstance(RewindInstanceRequest) returns (RewindInstanceResponse);
 
+    // Restarts an orchestration instance.
+    rpc RestartInstance(RestartInstanceRequest) returns (RestartInstanceResponse);
+
     // Waits for an orchestration instance to reach a running or completion state.
     rpc WaitForInstanceStart(GetInstanceRequest) returns (GetInstanceResponse);
 

src/Grpc/versions.txt

@@ -1,2 +1,2 @@
-# The following files were downloaded from branch main at 2025-08-08 16:46:11 UTC
-https://raw.githubusercontent.com/microsoft/durabletask-protobuf/e88acbd07ae38b499dbe8c4e333e9e3feeb2a9cc/protos/orchestrator_service.proto
+# The following files were downloaded from branch main at 2025-09-10 22:50:45 UTC
+https://raw.githubusercontent.com/microsoft/durabletask-protobuf/985035a0890575ae18be0eb2a3ac93c10824498a/protos/orchestrator_service.proto

test/Client/Core.Tests/DependencyInjection/DefaultDurableTaskClientBuilderTests.cs

@@ -144,6 +144,14 @@ public override Task<OrchestrationMetadata> WaitForInstanceStartAsync(
         {
             throw new NotImplementedException();
         }
+
+        public override Task<string> RestartAsync(
+            string instanceId, 
+            bool restartWithNewInstanceId = false, 
+            CancellationToken cancellation = default)
+        {
+            throw new NotImplementedException();
+        }
     }
 
     class CustomDataConverter : DataConverter

test/Client/Core.Tests/DependencyInjection/DurableTaskClientBuilderExtensionsTests.cs

@@ -174,6 +174,14 @@ public override Task<OrchestrationMetadata> WaitForInstanceStartAsync(
         {
             throw new NotImplementedException();
         }
+
+        public override Task<string> RestartAsync(
+            string instanceId, 
+            bool restartWithNewInstanceId = false, 
+            CancellationToken cancellation = default)
+        {
+            throw new NotImplementedException();
+        }
     }
 
     class GoodBuildTargetOptions : DurableTaskClientOptions

test/Client/OrchestrationServiceClientShim.Tests/ShimDurableTaskClientTests.cs

@@ -326,6 +326,79 @@ public async Task ScheduleNewOrchestrationInstance_IdProvided_TagsProvided()
         await this.RunScheduleNewOrchestrationInstanceAsync("test", "input", options);
     }
 
+    [Theory]
+    [InlineData(false)]
+    [InlineData(true)]
+    public async Task RestartAsync_EndToEnd(bool restartWithNewInstanceId)
+    {
+        string originalInstanceId = "test-instance-id";
+        string orchestratorName = "TestOrchestrator";
+        object input = "test-input";
+        string serializedInput = "\"test-input\"";
+
+        // Create a completed orchestration state
+        Core.OrchestrationState originalState = CreateState(input, "test-output");
+        originalState.OrchestrationInstance.InstanceId = originalInstanceId;
+        originalState.Name = orchestratorName;
+        originalState.OrchestrationStatus = Core.OrchestrationStatus.Completed;
+
+        // Setup the mock to return the original orchestration state
+        this.orchestrationClient
+            .Setup(x => x.GetOrchestrationStateAsync(originalInstanceId, false))
+            .ReturnsAsync(new List<Core.OrchestrationState> { originalState });
+
+        // Capture the TaskMessage for verification becasue we will create this message at RestartAsync.
+        TaskMessage? capturedMessage = null;
+        this.orchestrationClient
+            .Setup(x => x.CreateTaskOrchestrationAsync(It.IsAny<TaskMessage>()))
+            .Callback<TaskMessage>(msg => capturedMessage = msg)
+            .Returns(Task.CompletedTask);
+
+        string restartedInstanceId = await this.client.RestartAsync(originalInstanceId, restartWithNewInstanceId);
+
+        if (restartWithNewInstanceId)
+        {
+            restartedInstanceId.Should().NotBe(originalInstanceId);
+        }
+        else
+        {
+            restartedInstanceId.Should().Be(originalInstanceId);
+        }
+
+        // Verify that CreateTaskOrchestrationAsync was called
+        this.orchestrationClient.Verify(
+            x => x.CreateTaskOrchestrationAsync(It.IsAny<TaskMessage>()),
+            Times.Once);
+
+        // Verify the captured message details
+        capturedMessage.Should().NotBeNull();
+        capturedMessage!.Event.Should().BeOfType<ExecutionStartedEvent>();
+        
+        var startedEvent = (ExecutionStartedEvent)capturedMessage.Event;
+        startedEvent.Name.Should().Be(orchestratorName);
+        startedEvent.Input.Should().Be(serializedInput);
+        // TODO: once we support version at ShimDurableTaskClient, we should check version here. 
+        startedEvent.OrchestrationInstance.InstanceId.Should().Be(restartedInstanceId);
+        startedEvent.OrchestrationInstance.ExecutionId.Should().NotBeNullOrEmpty();
+        startedEvent.OrchestrationInstance.ExecutionId.Should().NotBe(originalState.OrchestrationInstance.ExecutionId);
+    }
+
+    [Fact]
+    public async Task RestartAsync_InstanceNotFound_ThrowsArgumentException()
+    {
+        string nonExistentInstanceId = "non-existent-instance-id";
+
+        // Setup the mock to client return empty orchestration state (instance not found)
+        this.orchestrationClient
+            .Setup(x => x.GetOrchestrationStateAsync(nonExistentInstanceId, false))
+            .ReturnsAsync(new List<Core.OrchestrationState>());
+
+        // RestartAsync should throw an ArgumentException since the instance is not found
+        Func<Task> restartAction = () => this.client.RestartAsync(nonExistentInstanceId);
+
+        await restartAction.Should().ThrowAsync<ArgumentException>()
+            .WithMessage($"*An orchestration with the instanceId {nonExistentInstanceId} was not found*");
+    }
 
     static Core.OrchestrationState CreateState(
         object input, object? output = null, DateTimeOffset start = default)