Open madhurrya opened 4 months ago
Noticing this again https://dev.mod.spiff.status.im/extensions/recent-error-events
This seems to be related to the comment here https://github.com/sartography/spiff-arena/issues/1861#issuecomment-2206467374.
@jasquat above error is in this model. What exactly needs to change there? https://dev.mod.spiff.status.im/process-models/manage-talents:talent-acquisition-from-job-requisition-to-hiring:job-requisition:request-new-role
@madhurrya I'm not sure. My comment was based on what @essweine was saying in slack. This issue could be unrelated and could actually be caused by too many tasks over time and eventually causes the recursion in spiff to topple over.
Unfortunately the diagram for the process instances in question no longer exists so we can't check it but the instances are quite old with a lot of tasks - I saw 22,000+ events on one. I think the main takeaway is to make sure tasks do not have multiple inputs to them going forward.
Also maybe one of the new changes in SpiffWorkflow added a recursion path that did not exist before?
Looping back directly to a task without an intervening exclusive gateway often causes this error. As for volume of tasks, the relevant metric is the depth in the tree of a particular task (though of course typically the probability of a deeply nested task goes up as the number of tasks increases). If you can get a serialized version of one of the failing processes, I can take a look at it.
I tried to get the serialized json but the process I was running - poetry run python bin/get_bpmn_json_for_process_instance.py 35667
- was killed. I'm not sure we can export the instance in its entirety.
Noticed the 'maximum recursion depth exceeded' error in these instances
https://dev.mod.spiff.status.im/i/37138 https://dev.mod.spiff.status.im/i/36091 https://dev.mod.spiff.status.im/i/36229
these instances are quite old. there is a nonzero possibility that a fix occurred in the interim, though we don't have a specific hypothesis about what that might have been.
stack trace from one of the errors:
UPDATE: elizabeth pointed out that if you have two sequence flows pointing to the same task (where you should have an exclusive gateway), this can happen. https://github.com/sartography/SpiffWorkflow/issues/394 is related.