Service Task with Async Continuation Never Executes

We’ve experienced this issue - can I ask - what are the downsides of setting jobExecutorDeploymentAware? It sounds from earlier discussion by @StephenOTT and @Webcyberrob that the REST deployment option requires some work to be a first class citizen? REST is certainly the more desirable deployment option for us - fits better with current deployment tools and pipelines. We are also planning to deploy Camunda via Docker containers.

My best understanding after reading the thread is that if deployment aware is switched off (which it seems is a requirement when deploying by REST), Camunda engines will need to be restarted to become aware of the new deployments?

Also, how does one go about

I’m struggling to find a description of this in the documentation.

Just wanting to follow up on this - I’ve taken another look at the jobExecutorDeploymentAware documentation and am needing to clarify:

  • Can anyone confirm that jobExecutorDeploymentAware must be set to false if using REST API deployments? That seems to be the case from this thread.
  • And therefore that it is important to maintain homogeneous clusters when deploying via REST?

Are there any other implications I should be aware of?

Thanks.

Hi @Ettery and everyone with the same problem,

I had the same error, and yes, if using REST API deployments, you must set this flag to false. There’s the explanation I received: The registration is always a local and in-memory state inside an engine. For deployments made via the REST API, the engine will not recreate that in-memory registration after restarting. So that the Job Executor won’t pick up any jobs to execute.

Sorry to resurrect an old thread, but wanted to communicate that we got burned by this situation today, as well. It took awhile to find this solution and I think it should be highlighted more clearly in the documentation. The Job Executor section describes why the flag should be True in a heterogeneous environment, but it does not at all mention the possible adverse affects of having it set to True in a container implementation using REST API deployments.

Would it make sense to have this value defaulted to False in the published container distribution?

2 Likes

It took me a few hours of sleep as well. My case was very similar:

When does It accurs:

  • Deployment with REST API
  • Most cases when your Application restart.

How to solve:
1- Verify your bpm-platform.xml file with this tag:
<property name="jobExecutorDeploymentAware">false</property>
2 - Redeploy your APP (Docker, whatever)