I love vol. X of Landau&Lifshitz. It even contains the Schwinger-Keldysh real-time formalism in a later chapter, but it also starts out from the basic classical version and is the only book which gets the detailed-balance principle right. It's not necessary for the S-matrix to be symmetric under parity and time reversal, but unitarity is sufficient. Of course, the Boltzmann equation, the H theorem etc. works also for the weak interaction, which has a unitary S-matrix but is neither P nor T invariant.
For the relativistic Boltzmann equation, a marvelous book is
C. Cercignani, G. M. Kremer, The Relativistic Boltzmann Equation: Theory and Applications, Birkhäuser Verlag 2002.
For a derivation from quantum many-body theory (non-relativistic case), using the Keldysh real-time formalism a very good (even pedagogical) treatment can be found in
P. Danielewicz, Quantum Theory of Nonequilibrium Processes I, Ann. Phys., 152 (1984), p. 239.
http://dx.doi.org/10.1016/0003-4916(84)90092-7
P. Danielewicz, Quantum Theory of Nonequilibrium Processes II. Application to Nuclear Collisions, Ann. Phys., 152 (1984), p. 305–326.
http://dx.doi.org/10.1016/0003-4916(84)90093-9
Another standard review article is
W. Botermans and R. Malfliet, Quantum transport theory of nuclear matter, Phys. Rept., 198 (1990), p. 115–194.
http://dx.doi.org/10.1016/0370-1573(90)90174-Z