推测执行的不适应场景
?
在HADOOP里,如果一个任务运行比预期的慢,就会尽快检测和启动另一个相同的任务作为备份来执行相同的工作,虽然它会降低执行慢的任务执行失败带来的损失,但也会消耗更多的资源,执行重复的工作。有利有弊,可以选择使用。org.apache.hadoop.ipc.RemoteException: org.apache.hadoop.hdfs.protocol.AlreadyBeingCreatedException:?failed to create file ......_part-00000 for DFSClient_attempt_201205151343_1053807_r_000004_2 on client 10.249.38.46.because this file is already being create by ......
jobConf.setBoolean("mapred.reduce.tasks.speculative.execution",false);
jobConf.setBoolean("mapred.map.tasks.speculative.execution",false);