Agent determination problem
I have a workflow that has been in production for almost a year. We are in the middle of integration testing for another rollout and I'm getting agent determination errors in this workflow. There are three steps involved. The first two use the exact same rule. The first of these two works fine. When it tries to run the rule again for the second step, it gets no agent found. If I test the rule it works. If I restart the workflow, it works. When it moves on to the third work item using a different rule, it errors out no agent found. When I test this rule, it works. When I restart at this step, it also works.
This is a new integration client that was copied from production. I've performed the standard workflow customization like I always do. I've run several other workflows and only notice the agent determination problems in this one. The workflow hasn't changed in months although there have been some user security changes...I'm not sure how that could really affect the agent determination during the creation of work items.
One thing that might be important...userA created the first workitem and the agent determination resolves to userB. When userB executes the first work item and completes, the second work item should also resolve to userB using the exact same rule. This fails. I restarted it using my account and it resolves to userB. UserB executes work item #2 and work item #3 should resolve to userC but the agent determation fails again. I don't know whether the work item agent is "used" to perform the creation of the next workitem or not but this is the only difference between the first two work items and one works while the other does not.
It's almost like userA can perform the agent determination but userB cannot. I can't manually execute the agent determination as userB so I ran se37 as myself in debug and changed SYST-UNAME to userB and it worked fine and returned userB as the agent. That would seem to eliminate security as an issue with the agent determination.
Are you absolutely sure that it is not an authorisation issue? Can you give SAP_ALL (and the possible structural authorisations) to all the users that are involved in the process? (I don't think that changing sy-uname in debug mode to another user really is a correct way to try to prove whether this is an authorisation issue.)
The agent determination of the following step is executed by the previous WF agent, if there is no background steps between. So if the previous agent doesn't have enought authorisations, the agent determination will fail.