Why camundaClient.newCompleteUserTaskCommand is requesting /v2/user-tasks/:taskId/completion

Camunda 8 Topics Discussion & Questions
2

Hi there! :waving_hand:

Great question about the API endpoint behavior. Let me clarify what’s happening here.

Why /v2/user-tasks instead of /v1/tasks?

The camundaClient.newCompleteUserTaskCommand is correctly using the Camunda 8 REST API v2 endpoint (/v2/user-tasks/:taskId/completion). This is the expected behavior for Camunda 8.9, and here’s why:

API Evolution in Camunda 8

  • v2 API (/v2/user-tasks) is the current standard for Camunda/Zeebe user tasks lifecycle operations (assign, unassign, update, complete)
  • v1 API (/v1/tasks) is the legacy Tasklist API that’s been deprecated since 8.8 and is scheduled for removal in 8.10

Key Differences:

  • v2: For Camunda user tasks (modeled as Camunda/Zeebe User Task elements)
  • v1: For legacy job worker-based user tasks

The Real Issue: 503 Service Unavailable

The problem isn’t the endpoint choice - it’s the 503 error indicating a connection issue with the brokers. This typically means:

  1. Gateway-to-Broker routing problem - The gateway can’t deliver the request to a broker/partition
  2. Partition awareness issue - The gateway isn’t aware of the partition where your task resides
  3. Configuration mismatch - Incorrect addresses or context paths

Troubleshooting Steps:

1. Check Gateway/Broker Health

# Check logs for partition/connection errors
docker logs <zeebe-broker>
docker logs <zeebe-gateway>

# Health check
curl http://localhost:9600/actuator/health

2. Verify Configuration

Make sure your Spring Boot configuration is correct:

camunda:
  client:
    mode: self-managed
    zeebe:
      rest-address: http://{gateway-host}:{gateway-rest-port}
      # Include context path if configured (e.g., /zeebe)

3. Common Fixes:

  • Restart the gateway if it lost partition awareness
  • Include context path in rest-address if your gateway is configured with one
  • Ensure version compatibility between your client (8.9.0-SNAPSHOT) and cluster
  • Check port configuration - Use gateway’s REST port (typically 8080), not Tasklist port (8082)

References:

The endpoint choice is correct - focus on resolving the broker connectivity issue. Let me know if you need help with any specific configuration details!