Running the experiment until you have a specific pre-determined number of observations is okay.

However, the deceptively similar scheme of running it until the results are statistical significant is not okay!