Created attachment 38323 [details] Test jmx file In v5.4.3, using a standard thread group I've been successfully using CSV Data Set Config with a CSV of 100 users for login/password in my HTTP requests. I'm now trying v5.5 and the Open Model Thread Group, but the users are not all being read from the CSV, it seems to skip users out of the list of 100. The skipping seems random, such as users 1-4 are read then 5 is skipped, then 6-15 are read, but 16-18 are skipped, and so on. I've also tried the CSV Data Set Config with recycle on end of file, and can see that if when it does recycle it again skips some users (not the same ones). So by the end of a test, some users have been used 3 times, some 1, some 5, etc. Works fine in v5.5 using a standard thread group, but when I try the Open Model the behaviour is as above. I have simplified my test right down to a single request, and removed any confidential data. In the attached jmx there is a "Thread Group" and a "Open Model Thread Group". Easiest to see the issue by disabling one of these and running the test, monitoring the Summary Report. There you can see the label and the username that is used. "Thread Group" goes through the csv in sequence, but "Open Model Thread Group" goes through in sequence skipping random users. I have also attached a test csv file for the usernames.
Created attachment 38324 [details] test user list csv
Thanks for reporting. Open Model Thread Group indeed initializes config elements twice. The workaround is to wrap the contents of the group in a Simple Controller. In other words, Open Model Thread Group { SimpleController { CSV Data Set; HTTP request } }.
Thank you. The workaround is very simple to do and seems to work perfectly.
I've filed a PR with the fix: https://github.com/apache/jmeter/pull/717
(In reply to Vladimir Sitnikov from comment #4) > I've filed a PR with the fix: https://github.com/apache/jmeter/pull/717 Thank you :-)
This issue has been migrated to GitHub: https://github.com/apache/jmeter/issues/5682