I have played around with this. I am using 2 WebTest client machines to reduce the overhead, each of them starting 16 instances of WebTest with the same script. In config.xml I specify to save a summary but not the results (writing a html file for each page seems to slow execution down a bit) . Then I end up with 16 directories on each of the two "clients", each directory containing a WebTestReport.xml. The following grep/cut works in bash (I tried on Kubuntu, but should work on other Linux flavours as well) and basically extracts the description and the time in milliseconds for all steps that have a description.
for f in */WebTestReport.xml; do xml2 < $f | tr '\n' ';'|sed -e "s#/summary/testresult/results/step;#\n#g"|grep description|grep -v Implementation|sed -e "s#/summary/testresult/results/step/##g"|tr '=' ';'|cut -d ';' -f 2,4|sort; done|sort Then I do some aggregation/averaging in a spreadsheet. Are others using a similar approach? cheers Ivan Marc Guillemot wrote: > > Hi, > > until yesterday I was convinced that WebTest was not an appropriate tool > for load testing. A mail from Serban Balamaci on htmlunit's user mailing > list seems to show that I was perhaps wrong: > -- View this message in context: http://www.nabble.com/Load-test-with-WebTest-possible--tp8648127p20319554.html Sent from the WebTest mailing list archive at Nabble.com. _______________________________________________ WebTest mailing list [email protected] http://lists.canoo.com/mailman/listinfo/webtest
