I have been asked by support to trace the processes for an infrequent bug we experience. Sometimes when doing an AP check run, upon printing checks, epicor freezes up completely and has to be closed before anything else can be done on that PC. I have traced a “good” check run, but I can’t get the timing right to capture the trace when the issue randomly decides to occur.
To that end, I am considering leaving a trace running on one of my user’s PCs. That way if the error does crop up, she won’t forget to have turned on tracing. My concern is with the trace log file size.
Does the trace log keep itself within a certain file size? Is this settable? Is it a bad idea to leave trace running for a long time?
Since we only run checks once a week, we only have on chance a week to see if the error will occur.
Does the Trace log file get appended to, overwritten, or a new one created each time the client is launched (with tracing enabled)?
If it is either of the later two, ask the user to close Epicor and then relaunch it right before doing the check run (like if they could remember that, they would just enable logging!)
Some spit-balling …
Is it possible to have a BPM control the trace logging? It wouldn’t need to be so fancy as to turn it off, but just make sure it is enabled via one of the BO’s called early in the Check printing run.
Some other debugging ideas…
Take a back up of the DB the night before the day checks are run and restore it to a Test environment. If the crash happens in Production, see if it happens in Test. If it does, restore Test again, and try running a subset of the checks. Maybe this could identify the offending records (assuming it is repeatable - which it might not be as you said the client PC has to have Epicor closed and reopened and then it works)
And one last thing. I’ve seen errant report entries in the Sys Monitor seem to crash a users client program. If it crashes during printing, check the SysMon (from another computer) to see if the report process started. Do this before the crashed client does anything else - including closing the client.
What about a simple BPM that reminds the user to turn on trace logging when they create a new payment group? Or if they aren’t to be trusted, turn of their menu access to payment entry so that they they can’t even open it until an admin is alerted.