Let’s have 4-node Flowable cluster.
I want user tasks(processes) executed on nodes 1 and 2, and async service tasks (sub-processes) on nodes 3 and 4. So a user starts a process on node 1, enters some data, then async sub-process is executed on node 3 or 4 and the result is returned to node 1 for the user.
Is it possible with just Flowable configuration or I have to have an additional communication channel to start processes and pass data?
I have achieved something like this by disabling the job scheduler on some nodes using something like this: processEngineConfiguration.setAsyncExecutorActivate(false);
You will need to retrieve the result some way, because you synchronous call from the user will not wait for the async one to me done. Your user will get a response probably before the async job has executed.
We have a status indicator on our UI when the user loads the process. They can see what elements from the process have executed or are waiting.
Thanks for the response.
Completely disable async processing may be an interesting option, but I’ll need to check if we need one on the “user interface” node.
The “async” user scenario is expected: the document will be in the “external processing” state until the result is available, so user will see the read-only page.
I accomplished something similar by separating the process modeling across tiers using the EventRegistry and kafka to communicate between the two. Basically I have a web tier which defines business processes, and a worker node tier which also has business processes defined.
The web tier process controls the overall flow of a complex job and the worker nodes implement discrete tasks which are time consuming and need to occur separate from the web tier.
Communication between the tiers is handled using the EventRegistry and Kafka as the message broker. I found this easiest to implement by separating the databases the deployments use for the web and worker tier.
Not sure about the specifics of your situation, but this may be another avenue to consider.