Fatigue life of filled natural rubbers used in isolators is very dispersive owing to many uncertainties such as loads and temperature. Conventional crack initiation methods and crack propagation methods to determine fatigue life of rubbers are insufficiently to characterize the phenomenon; thus, probabilistic methods are desired in practical engineering. Based on the test data of uniaxial tensile fatigue experiment of rubbers, the cumulative distribution function and probability density function of residual fatigue life corresponding to two typical probability distribution models of three parameters Weibull distribution（3 PWD） and logarithmic normal distribution（LND） were derived and applied to the prediction of residual life reliability, in order to determine the minimum required under different probability distribution model of sample number and working condition number. The results showed that 3 PWD was more suitable to characterize the probability distribution of rubber fatigue life. The minimum values of the number of rubber fatigue test samples and the number of working conditions are distributed between 4 and 6. The log mean should be preferred for subsequent fatigue analysis when determining the median life. The research can provide guidance for the anti-fatigue design and test of rubber vibration isolators.