For elastic constants you need to strain your crystal and calculate the total-energy response to the strain. The nature of the strain dictates the specific elastic constant or a linear combination of elastic constants. For VASP the structure is defined in the POSCAR file. For the specific strains you can use what has been published. For example for cubic: Phys. Rev. B 48, 5844 (1993); hexagonal: Phys. Rev. B 51, 17431 (1995); orthorhombic: Phys. Rev. B 66, 085113 (2002); monoclinic: Phys. Rev. B 79, 104110 (2009). Hope this helps.
You can also check the ElaStic code (http://exciting-code.org/elastic), which can generate suitable strains and analyze the results for any structure. There is no VASP interface in the official version, but you can look in their paper about it for more information Computer Physics Communications 184 (2013) 1861–1873, and you can also find out more about their method for generating strains in the original reference: Comput. Phys. Comm. 181 (3) (2010) 671–675.
Edit: Also beware that you need tighter than normal convergence parameters for accurate elastic constants. With VASP, I would go for
PREC = Accurate
ADDGRID = .True.
EDIFF = 1e-6
EDIFFG = -0.002
Remember to check your k-point convergence carefully.