Typical data in a microbiome study consist of the operational taxonomic unit (OTU) counts that have the characteristic of excess zeros. Zero-inflated and hurdle generalized linear mixed models (GLMM) have been proposed in the literature for modelling such zero-inflated data. Checking the goodness-of-fits of GLMM models in practice is still a daunting problem. The null distributions of traditional Pearson and deviance residuals are far from the claimed normal distribution. We propose to use randomized quantile residual. Our simulation studies show that randomized quantile residual has well-calibrated null distribution under the true model, and has great statistical power in detecting the model failure. We will also demonstrate RQR with a real microbiome dataset.