agent/agent-framework
1
0
Fork 0
mirror of https://github.com/microsoft/agent-framework.git synced 2026-06-16 21:04:09 +08:00
Code Issues Packages Projects Releases Wiki Activity

.NET: Fix intermittent checkpoint-restore race in in-process workflow runs (#5134)

Browse Source 
* Improve workflow unit tests

* Update test name prefix for clarity.

* Update tests to surface any errors.

* fix check-point restore-time race in off-thread workflow event stream

* Fixes an intermittent checkpoint-restore race in in-process workflow runs.
This commit is contained in:
Peter Ibekwe
2026-04-15 21:20:45 -07:00
committed by GitHub
Unverified
parent 8f7fd9525d
commit 87a8fa2a9d
2 changed files with 48 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
dotnet/src/Microsoft.Agents.AI.Workflows/InProc/InProcessRunnerContext.cs
+6
View File
@@ -419,6 +419,12 @@ internal sealed class InProcessRunnerContext : IRunnerContext
.Select(id => this.EnsureExecutorAsync(id, tracer: null).AsTask())
.ToArray();
// Discard queued external deliveries from the superseded timeline so a runtime
// restore cannot apply stale responses after importing the checkpoint state.
while (this._queuedExternalDeliveries.TryDequeue(out _))
{
}
this._nextStep = new StepContext();
this._nextStep.ImportMessages(importedState.QueuedMessages);
dotnet/tests/Microsoft.Agents.AI.Workflows.UnitTests/CheckpointResumeTests.cs
+42
View File
@@ -279,6 +279,48 @@ public class CheckpointResumeTests
"the workflow should be able to continue after the runtime restore replay");
}
/// <summary>
/// Verifies that restoring a live run clears any queued external responses from the
/// superseded timeline before importing checkpoint state.
/// </summary>
[Fact]
internal async Task Checkpoint_Restore_ClearsQueuedExternalResponsesBeforeImportAsync()
{
Workflow workflow = CreateSimpleRequestWorkflow();
CheckpointManager checkpointManager = CheckpointManager.CreateInMemory();
InProcessExecutionEnvironment env = ExecutionEnvironment.InProcess_Lockstep.ToWorkflowExecutionEnvironment();
await using StreamingRun run = await env.WithCheckpointing(checkpointManager)
.RunStreamingAsync(workflow, "Hello");
(ExternalRequest pendingRequest, CheckpointInfo checkpoint) = await CapturePendingRequestAndCheckpointAsync(run);
await run.SendResponseAsync(pendingRequest.CreateResponse("World"));
await run.RestoreCheckpointAsync(checkpoint);
List<WorkflowEvent> restoredEvents = await ReadToHaltAsync(run);
ExternalRequest replayedRequest = restoredEvents.OfType<RequestInfoEvent>()
.Select(evt => evt.Request)
.Should()
.ContainSingle("the restored run should still be waiting for the checkpointed request")
.Subject;
restoredEvents.OfType<WorkflowErrorEvent>().Should().BeEmpty(
"a queued response from the superseded timeline should not be processed after restore");
RunStatus statusAfterRestore = await run.GetStatusAsync();
statusAfterRestore.Should().Be(RunStatus.PendingRequests,
"the restored run should remain pending until a post-restore response is sent");
await run.SendResponseAsync(replayedRequest.CreateResponse("Again"));
List<WorkflowEvent> completionEvents = await ReadToHaltAsync(run);
completionEvents.OfType<WorkflowErrorEvent>().Should().BeEmpty(
"the restored request should complete cleanly once a new response is provided");
RunStatus finalStatus = await run.GetStatusAsync();
finalStatus.Should().Be(RunStatus.Idle,
"the workflow should finish once the replayed request receives a fresh response");
}
/// <summary>
/// Verifies that a resumed parent workflow re-emits pending requests that originated in a subworkflow.
/// </summary>