It is not a big problem. For macro measurements you have to focus beam to the center of cuvette. If you know focal distance of your lens, the focal point needs to be in the center of cuvette. If you are using micro-Raman it will be more efficient to take low magnification objective. You have to focus on the surface, find the band of interest and move focus down to the solution and monitor signal. You signal will increase until certain point and will decrease. So, you have to find point where signal is max. To find the surface of the liquids/solutions, you have to see back reflection of the laser beam. It should be a nice small laser spot on the surface.

a common approach is to use a cuvette and conduct Raman scattering at 90 degrees scattering configuration (as supposed to backscattering configuration often used for opaque materials such as semiconductors). You can tweak the optics and maximize the Raman peak counts for a known Raman line for the sample you are testing to ensure you are focusing on the right spot and your collection geometry is optimized.

People usually measure in round capillaries, but the curvature of the capillary's wall produces huge astigmatism and hence loss of signal due to smeared focus.

It is much better to use square capillaries. If you can't get one, just put a droplet of your fluid between two cover glasses and focus between them. Or you can use round capillary, but put a drop of water over it and attach the cover glass through it. In both cases you obtain a flat outer surface and several-fold improvement of signal. In addition to the above, if your sample stage is horisontal, you can obviously focus directly the free surface of your fluid. Finally, with modern spectrometers you can measure using backscattering geometry without any harm.