How much testing is too much?

We all want to achieve a high level of coverage to give us confidence in the code, but at what levels should we focus our efforts, and how much? In particular, how should we focus the efforts between Unit tests, Integration tests, and Functional tests?

Focusing too much time and attention in just one of these areas can have detrimental effects. Yet, taking the time too get extensive code coverage at all levels may be costly and cause delayed releases and/or sales.

So where do we draw the line? How much testing is too much?