Could LR actually be causing server failures/error
I'm working on testing a Web system here and using LR 11 with TruClient. We switched to TC as the UI has a lot of Ajax.
While working on this test I am frequently running into a scenario in which the app server WebShere (a layer behind the Web server) is crashing and needs to be manually restarted by a developer. I have duplicated this many times using a single user in VUGen telling the script to run for say 20 iterations.
We have have WebSeal -> IIS -> WebSphere as the hierarchy. And we have a custom java app from our vendor running on WebSphere.
There are the usual think times and sync points so the script is not racing along too fast.
We have not been able to consistently duplicate the problem manually, outside of LR/VUGen.
There are concerns that LR is actually causing the problem. But yesterday I ran a 25 user test for 30 mins that ran to completion without a single occurrence of the error. I would guess that if a single LR user is causing the error, 25 would cause it much more often?
Is it conceivable that a LR TruClient script could actually be causing the app server to crash, not with load, but just by being run with a single user? Something about the script itself rather than load?
I'm skeptical as the Web browser never talks to the part that crashes directly (the app server), it only converses with the Web server.
Re: Could LR actually be causing server failures/error
As a first step, increase all timeouts of TC script (including object identification timeouts) to the maximum (I believe it is 1000 sec) and then execute the script to verify the problematic object first.
Then try to work with properties used in object along with the arguments specified.
Finally play with arguments first and then with properties to identify the property - argument pair.
Use same with manual steps to replicate the issue.