In our standard FME Flow training course we have one excercise where Workspace A does some checks on user parameters and if they are succesful it runs Workspace B through an FMEFlowJobSubmitter.
One of our trainees accidentally set up the FMEFlowJobSubmitter to run Workspace A instead and when he published it to Flow he made a recursive loop: every time Workspace A ran, it queued Workspace A again. I was busy helping somebody else at that time so when I got around to his question I noticed it had ran 17000+ jobs in a matter of minutes. These were low-impact jobs, so it wasn’t an immediate problem, but it’s safe to say that this has the potential of completely overwhelming an FME Flow setup.
After talking to Safe we decided to post it as an idea here to see if others have come across this as well or have opinions about it.
The big question would be:
Should FME be able to detect these recursion risks?
And in a broader sense: have you had this happen to you? Or something similar that you think FME should be able to detect and warn you about?
