This paper studies few-shot relation extraction, which aims at predicting the relation for a pair of entities in a sentence by training with a few labeled examples in each relation. To more effectively generalize to new relations/tasks, in this paper we study the relationships between different tasks and propose to leverage a global task graph. We propose a novel Bayesian meta-learning approach to effectively learn the posterior distributions of the prototype vectors of tasks, where the initial prior of the prototype vectors is parameterized with a graph neural network on the global task graph. Moreover, to effectively optimize the posterior distributions of the prototype vectors, we propose to use the stochastic gradient Langevin dynamic, which can be related to the MAML algorithm but is able to handle the uncertainty of the prototype vectors. The whole framework can be effectively and efficiently optimized in an end-to-end fashion. Experiments on two benchmark datasets prove the effectiveness of our proposed approach against competitive baselines in both the few-shot and zero-shot settings.