These retries have been part of Load Tester since the beginning, but starting with 6.6 we turned on reporting for these events. We began reporting retry events after a customer told us that they needed the information and we could see no reason to hide it. You'll notice that the testcase continues to work normally after the retry and you should also see that the retry events don't show up in the load test report as failed performance goals.


If you try to run a load test based on a testcase with a replay that has one of these retries, you may get a warning that the replay wasn't successful. If the retry is the only reason for the warning, you may click through it.


The tool may also retry a request after a long period of time (often 120 seconds). For most people, the only concern here is that the user had to wait that long to get a response.


If you have the time and curiosity, you can try examining the traffic in wireshark. Sometimes the server is closing connections despite sending a Keep-Alive header, or there can be a race condition where the HTTP request from the client crosses the FIN packet from the server on the wire, or there may be a variety of other causes. Any of these can force a retry, which usually won't be noticeable to the user.