Congrats to all at Camunda, really interested to see how Zeebe evolves.
I have a couple of questions (my apologies I know these are probably all in the code, but I am haven’t done a deep dive yet)
In the Zeebe.io landing page there is a none start event where some person/application launches the workflow. I was wondering what about the workflow with a message or signal start event. If I publish a workflow definition, will it start to subscribe to interesting events to create a workflow instance or must a workflow instance be first created?
I see you have a dependency on camunda engine itself, does this mean that camunda engine is doing the workflow orchestration or have you converted the definition to a state machine within the broker?
For event correlation, do you expect that the participating microservices will have some knowledge of the process instances, or is it completely decoupled and the correlaton is handled in the broker somehow?
In the roadmap there is a mention of partitioning topics - just wondering why kafka is not being considered here?
Currently there are not message or signal start event’s implemented. Zeebe supports currently only None Start Event, Service Task and End event, but more will follow. To see what is next see your road map.
There is no dependency to the camunda engine. Right, the execution of a workflow is done by StreamProcessors, for example the WorkflowInstanceStreamProcessor, which is a state machine.
I can’t answer this yet, since currently our focus lies on other topics.
We tried to implement a self contained service with a minimum of dependencies. Besides that it is also a performance issue, since we can access the data on one node directly without another framework.
Hope this answer your questions to your satisfaction.