Markov Chain Monte Carlo (MCMC) is an inherently serial algorithm that is widely used to provide samples from the posterior distribution. Although several techniques have been developed to generate MCMC samples in parallel, the question of efficient use of massively parallel computing to generate samples from many dimensional and (possibly) multimodal posterior is still on a great interest. During the presentation, a method of parallelizing MCMC sampling in which the parameter space is divided by multiple subspaces that are sampled on different computing nodes without the need for communication will be presented. Samples from different subspaces are then reweighted and normalized by using the Adaptive Harmonic Mean Integration (AHMI) algorithm. As an example of this parallelization idea, the MCMC simulation of the proton bunch parameters in the AWAKE experiment will be discussed.