Record‐breaking rainfall of 524.1 mm in 24 hr occurred in the coastal metropolitan city of Guangzhou, China, during 6–7 May 2017 and caused devastating flooding. Observation analysis and a nested very large eddy simulation (VLES) with Weather Research and Forecasting (WRF) model were conducted to investigate various factors that contributed to the heavy rainfall, including synoptic weather pattern, topographic effects, cold pool, and urban effects. First, the warm and moist southerly flow in the lower troposphere over the trumpet‐shaped topography of the Pearl River Delta continuously provided fuel for the development of the severe rainfall. Consequently, the southerly flow from the sea in the south strengthened with the development of the convection. Meanwhile, the precipitation‐produced weak cold pool supported a stationary outflow boundary, where new convective cells were continuously initiated and drifted downstream. The interaction between the cold outflows and the warm moist southerly flows in the lower troposphere formed a back‐building convective system, which produced local persistent heavy rainfall that lasted for more than 5 hr and reached record levels. Sensitivity experiments in which the urban area was removed from the model indicate that the urban forcing affected the timing and location of convective initiation and helped concentrate the maximum rain core. The nested WRF‐LES successfully simulated this heavy rainfall, and the model's advantages are noted for forecasting such local severe weather.