|Message:||We have some strange behaviour that we can't quite get to the bottom of and need some advice.
We're load testing a Domino HTTP application and are presently fronting it with webbooster. We're putting 160 users through and running for over 12 hours.
It's on a 4 way Win2k box. WebB is also on the same box. For some reason localhost doesn't work in booster.config and we use the real hostname.
LoadRunner tells us response is poor and yet when we look at the responses from the Domino Http access log it suggests that the domino response is very good.
We then looked at the webbooster logs.
In the compress log we see an OpenDocument of size 470,275 bytes being compressed down to 84,378. the compress log tells us WebServerTime of 2,719ms and a TotalTime of 2,797ms. So far so good. This suggests that the gzipping of the ~470k is very quick.
The booster log tells us another story and for the same OpenDocument it says that the time-taken is 49,469ms.
This 49secs sort of ties up with the Loadrunner time.
What could introduce this amount of latency ? When does the time-taken 'timer' start and finish?
The box under load is busy but I would expect see Domino suffer more than WebBooster. This type of behaviour is repeated throughout the load tests at various stages.
Any thoughts? Any status commands to key in the console? Everything looks ok but isn't. I've run the booster with verbose gc and everything looks good.
This is the reason I've been asking the questions re timeouts, threads etc.
I've even been setting the TCP Time_Wait down to 30secs.