On Fri, Dec 7, 2012 at 7:51 PM, sebb <[email protected]> wrote: > On 7 December 2012 18:34, Philippe Mouawad <[email protected]> > wrote: > > Starting with 1000 threads with sometime pages taking 200 Ko, that can > > take much useless memory particularly for previous sample result which is > > not used 90% of time. > > Isn't the previous sample result needed for Assertions and Post-Processors? > Or is that done separately? > > You are right. Stupid idea of mine :-)
> > Shouldn't we give it a try ? > > My concern is that this could be a cause of subtle failures if used > inappropriately, and could cause unnecessary support requests that > take a long time to investigate and answer - i.e. we could be causing > a lot of support work. > > > On Friday, December 7, 2012, sebb wrote: > > > >> On 7 December 2012 17:15, Philippe Mouawad <[email protected] > <javascript:;>> > >> wrote: > >> > Hello, > >> > What about implementing a kind of agressive mode for High Load Tests. > >> > > >> > It would be a Config Element that would change slightly JMeter > behaviour: > >> > > >> > - Nullify previous SampleResult > >> > - Nullify JMeterContext#previousSampler > >> > - Clean response data as soon as sampling and post/processor , > >> > assertions have been done > >> > >> I think that would stop the listeners from working. > >> > >> > > >> > Other field of improvements would be to work on HashTree size and > >> cloning. > >> > > >> > What are your thoughts ? > >> > >> Unless the test plan includes a listener that saves copies of every > >> sample result, then at most two samples will be saved (current and > >> previous) per thread. > >> > >> Is there really a problem with memory usage for high load tests - > >> unless the samples are huge? > >> > >> The risk is that clearing the data will cause subtle test failures > >> which are very hard to debug. > >> > >> > Regards > >> > Philippe > >> > > > > > > -- > > Cordialement. > > Philippe Mouawad. > -- Cordialement. Philippe Mouawad.
