I’m getting literally mad… I see something like the waterfall BuildBot diagram from Google and I’d like to study what do they have behind…
In the previous 3 years I went from nothing to “structured, .ini based, subversioned tests” usable for functional and also for non-functional testing. But, still… there was a Fitnesse instance on its own, an HP Mercury Quality Center (aka “MQC”) on its own, Hudson on its own, and deployment… by hand with tar.gz or zip files. 😦
Having a commit on the svn, going into Hudson, triggering Junit testing then Capistrano and automated functional and non-functional tests, reporting all back to MQC and creating a nice performance report, being able to compare build by build… seems science fiction. I see people creating stuff in MQC, doing Jmeter tests (basically rewriting the testcases) or with Selenium or HP LoadRunner… then installing stuff manually, or in the best case with Webistrano … and some other testing with Fitnesse. A mess…

I’d like to know experiences, from other people out there.

I have driven the development of the HTTP Quality Assurance Toolkit. At least, with that, we were able to do only once the test definition and being able to do functional & non-functional testing, and also Nagios monitoring.

I’m going to speak at DevOpsDays.org with an iGNiTe on Friday… about LoadTesting… not sure if to review my slides.