replied on August 13, 2014
Hi Rob, Abby,
Thanks again for your input.
I can summarise some possible scenarios that I have encountered:
- Outflow paths/conditions are incorrectly configured, resulting in a "failure" in the process model. This is a user configuration problem in the Process Modeler, but since the modeler validated correctly, it's difficult to detect early. In this case, it's too late and the form needs to be submitted again. This error is recorded in the Application Event Logs, perhaps it could actually be caught (like a Try-Catch or something) and the submission be held back...since it doesn't know which Outflow path to follow.
- I've had errors trying to save to Laserfiche on occasion. I can't recall the exact details, but I know that in some instances (Like the server being offline) I am able to restart the specific step that was suspended by viewing the Instance itself under "Results" section. It's not that obvious that an error occured at this point and requires manual intervention to fix, would be nice if we handle this "error" and send an email notification or something. In cases where it fails altogether, well, one needs to start over again.
- Made changes to the process modeler for one process (on one of the final steps), not realising that existing instances already in the middle of that process may be affected by that specific change. The result was that those instances failed and needed to start from scratch.
- I've seen this, but not sure about it causing total failure, but I have seen SMTP errors logged in the Event logs for forms. Just mentioning it, not sure how those errors are handled.
I will see if I can chat with my colleagues and get some more examples, and if need be, I will try and get the detailed information if you need it.
In general, I think having a "blanket Catch All" setting as opposed to individual "Try-Catch" type settings like in WF (for now), would be very useful. If the errors are being logged to the Application Event Logs, then clearly you're catching them somewhere. At the same time the error happens, perhaps we can have options to control the behaviour of the process. Maybe "Suspend" and not execute that specific step until it's been fixed in the Process Modeler. Having an email notification to the administrator would be useful too.
Cheers
Sheldon