I have a simulation that I generated via the Recorder. In one of the requests, it references a .txt file in the user-files directory (for the body of the request).
Because I’m stupid, I’ve since moved the files around all over the place. In the latest directory where I put everything, I didn’t copy the user-files. So the .txt file was missing and not found. I noticed a bunch of exceptions being logged, but I assumed that they’d just count as failed requests and that Gatling would move on to the next request.
Now that I’m looking at the numbers more closely, it actually appears that the exception may have been doing something much more sinister… perhaps causing the request to get pawned off to another Akka Actor, and thus issued again?
In any case, my total request counts look very odd (they are much, much higher than what I’d expect if I’d had 0 failures).
Is there a known/expected behavior for the case where a user-file is missing?
Which request count are you talking about? The one in Gatling, or one computed server side?
The only sane explanation would be that you use tryMax. Could you share your simulation, please?
I’m referring to the total request count that shows up in the Gatling HTML. In some runs where I was missing one of the text files, it shows up as approximately 8x the expected value. Once I add the text file back into place, I’ve never seen that total request count differ from my expected value at all.
The simulation is straight out of a Recorder session with no modifications, so I definitely didn’t add in a tryMax. You can see it here:
I am, however, wrapping it with that fancy ChainBuilder stuff that you helped me figure out in a previous e-mail thread:
I have absolutely no idea about what might happen here.
Will have to dig if I find time…
Cool, thanks. Not urgent as the fix is pretty simple on my end: “Don’t forget the user-files, dummy!”
However I am pretty well convinced at this point that Very Bad Things happen if you screw that up like I did.