Authentication Problem Load Runner 7.0.2
When running 10 virtual users (using web_set_user), the Loadrunner logs indicate that 10 unique user ids are logging on to the windows 2000 web server. However, the IIS log indicate that some users are logging on multiple times and some users that have Loadrunner logs don't have entries in the IIS log at all.
If I'm running 10 users, I always get 10 entries in the IIS logs but not with the user ids that I expect to see.
If I run the same test multiple times, you get different results. For example, the first time I run the test I will get 2 entries for user2 and no entries for user3. The next time I run the script I may get no entries for user2 and 3 entries for user3.
Has anyone else experienced this?
TIA for your help.
Re: Authentication Problem Load Runner 7.0.2
First, I assume that you're using a parameter file and that your data assignment method ("Select next row") = unique?
If those settings are correct, are you running the WinInet or the sockets replay engine? WinInet is required for some things, but I had a similar "flaky" experience before and Merc's advice was to use the sockets replay engine instead of WinInet.
Here was my experience:
Load Runner 6.0 - parameter file with user ID, password, and mainframe access ID, set to unique for the first param, and "use same row" for the other params. User ID and password for the web server authentication were accurate, but it was associating the wrong mainframe access ID, so requests were failing at the mainframe (network traces from Load Runner system to web server identified LR machine was sending it incorrectly).
Merc's recommendation was to upgrade to 6.5 and use the sockets replay engine (if I recall correctly 6.0 offered only WinInet). At the time (mid 2000), Microsoft's knowledge base articles were very clear that Wininet was not designed to be run as a server-type process, and was not considered thread-safe under certain conditions such as intense load. I think their tone has softened some since then.
I decided to play it extremely safe by using the sockets replay engine and running the vusers as processes versus threads. The problem didn't reoccur after that. While the readme data on the WinInet versus sockets usage is not very detailed, Merc's support knowledge base has some decent writeups on the best times to use each, just keyword search on WinInet.
Hope this helps