I have a set of 7 vectors and want to create a new 7'th vector that preserves the correlation between the original 7. How?

To be more specific. I have seven vectors x1, x2, ..., x7, all of the being normal distributed (or very close).

I want to create a new vector to replace x7, however such that the correlation to the vectors x1, x2, ..., x6 (with their original values) still applies.

How?

If expanding this problem to let's say 50 vectors, and I want to do the same fore x50 I might see that the covariance matrix is not positive definite - not by much, but slightly. Can this be solved as well?