getWorkflowState() stuck at RUNNING despite workflow completion (state sync bug)

[leblancmeneses]

Dapr Workflow State Not Syncing After whenAny Resolution

Expected Behavior

When a workflow completes after whenAny([waitForExternalEvent, createTimer]) resolves, getWorkflowState() should return COMPLETED status within 1-2 seconds.

Actual Behavior

The workflow completes internally (Dapr logs show Orchestration completed with status COMPLETED), but getWorkflowState() continues returning RUNNING indefinitely.

Steps to Reproduce the Problem

1. Clone reproduction repository: dapr-whenanybug-repro.zip

   cd dapr-whenanybug-repro

2. Deploy to Kubernetes cluster with Dapr installed:

   # Build and load image
   TAG="test-$(date +%Y%m%d-%H%M%S)"
   docker build -t localhost:5001/dapr-whenanybug-integrated:${TAG} .
   kind load docker-image localhost:5001/dapr-whenanybug-integrated:${TAG}
   
   # Update deployment with image tag
   sed -i "s|image: localhost:5001/dapr-whenanybug-integrated:.*|image: localhost:5001/dapr-whenanybug-integrated:${TAG}|" k8s/deployment.yaml
   
   # Deploy
   kubectl apply -f k8s/rbac.yaml
   kubectl apply -f k8s/service.yaml
   kubectl apply -f k8s/deployment.yaml

3. Watch the logs:

   kubectl logs -f -l app=whenanybug-integrated -c integrated

4. Observe output:

   • ✅ K8s job created by activity
   • ✅ Job calls HTTP endpoint which raises event
   • ✅ Workflow logs: [Workflow] whenAny resolved!
   • ✅ Dapr logs: Orchestration completed with status COMPLETED
   • ❌ Status check: getWorkflowState() returns RUNNING for 10+ seconds

Environment

• Dapr version: 1.16.8
• @dapr/dapr version: 3.3.0
• Kubernetes version: 1.32
• State store: PostgreSQL with actorStateStore: "true"

Code Pattern

async function* jobWaitWorkflow(ctx, input) {
  // Call activity to create K8s job
  const result = yield ctx.callActivity('createJob', jobName);

  // Wait for external event OR timeout
  const eventTask = ctx.waitForExternalEvent(`job-done:${jobName}`);
  const timerTask = ctx.createTimer(30);
  const winner = yield ctx.whenAny([eventTask, timerTask]);

  // BUG: Workflow completes here but state never syncs
  return winner === eventTask ? 'completed' : 'timeout';
}

The K8s job calls back via HTTP, which raises the event using client.raiseEvent(). The workflow resolves internally but client.getWorkflowState() never reflects the completion.

[olitomlinson]

@leblancmeneses

getWorkflowState() returns RUNNING for 10+ seconds

does it eventually go to a completed state after waiting longer than 10 seconds?

[leblancmeneses]

No, the workflow never completes - getWorkflowState() is permanently stuck at RUNNING. After 5 full minutes of polling every second (300 checks), the status never transitioned to COMPLETED