Hi,
I’m looking into Camunda as a solution to orchestrate our micro-services architecture. I would, therefore, make heavy use of the external task activities.
The fetchAndLock API, with its timeouts and retries, seems to handle all but one error conditition: what if there are no workers fetching a topic? How long will messages be stacked in this queue? Is there a way to setup an initial timeout for a task, so that it raises an incident if no one fetches it?
Best regards
Carlos Jourdan
Hi Carlos,
you can model this behavior in your process. For example, add a timer boundary event to the service task. After the given time is reached, you can leave the task, send a notification or create an incident.
Does this help you?
Best regards,
Philipp
Hi Philipp,
Technically that would work, but I think it would pollute a lot our workflows.
Our applications would look a lot like this example from Bernd Rücker’s blog:
If I have to set up a timer boundary event for each external task it will be a lot of boilerplate work. So, if there is no functionallity for this in the framework yet, I guess I’ll prefer to set up some sort of monitoring on the outset to make sure there are no lingering tasks…
Cheers
Hi @carlosjourdan,
currently, there is no built-in monitoring for this. So, you must build something yourself.
Best regards,
Philipp
Hi @carlosjourdan,
you can use a custom parse-listener to add a timer job for each external task in the background without modeling it.
Here is an example, that you have to adopt to your use case: https://github.com/camunda/camunda-bpm-examples/tree/master/process-engine-plugin/bpmn-parse-listener-on-user-task
Does it help you further?
Cheers, Ingo
That looks promissing. Thanks.
I have similar requirements as @carlosjourdan and got the BPMN Parse Listener to work for ServiceTasks. But how do I concretely implement the timeout behavior for a ServiceTask?
Thanks,
Wilko
