hadoop - who communicates with the namenode in yarn?

Question

since the jobTracker in MapReduce1 is replaced by the Application Master and Resouce Manager in Yarn I wonder who is communication in Yarn with the namenode to find out where the data is stored in the different datanodes? Is the Application Master doing so?

Answer 1

在 YARN 中，每个应用程序ApplicationMaster负责从Namenode. 稍后，当任务尝试在分配的节点上执行时，YarnChild从 HDFS 获取相应的拆分。