We discuss the host galaxy metallicity distribution of all long gamma-ray bursts (GRBs) whose redshifts are known to be $< 0.4$, including newly obtained spectroscopic datasets of the host galaxies of GRB 060614, 090417B, and 130427A. We compare the metallicity distribution of the low-redshift sample to the model predictions, and constrain the relation between metallicity and GRB occurrence. We take account of spatial variation of metallicities among star forming regions within a galaxy. We found that the models, in which only low-metallicity stars produce GRBs with a sharp cutoff of GRB production efficiency around 12+log(O/H) $sim$ 8.3, can well reproduce the observed distribution, while the models with no metallicity dependence are not consistent with the observations. We also discuss possible sampling biases we may suffer by collecting long GRBs whose redshifts are known, presenting the photometric observations of the host galaxy of GRB 111225A at $z = 0.297$ whose redshift has been undetermined until $sim$ 2.3 years after the burst.