How to approach online surveys when you have no clear hypothesis to test?

Let's say you want to redesign a website that is being used by thousands of people for the last few years. To do so, you need to collect data about how people use this particular website and you decide that your first step in understanding user behavior is to use an online survey.

What exactly should you be asking when you don't have any hypothesis to test? When don't have previous data, you're not aware of usability problems and when you just need to collect as much data as possible about user behavior? What questions should you ask in such situations?