It is the result of the 'speculative execution' feature of Hadoop Map/Reduce. Nothing really to worry about, but you may want to have control over it (JobConf has some methods that set it to true/false per phase or overall).
2011/4/21 ajing.wang <[email protected]>: > Hi ,guys. > I'm confused with mapred's running state. see the table of job running > state. > > > KindTotal Tasks(successful+failed+killed)Successful tasksFailed tasksKilled > tasksStart TimeFinish Time > Setup110021-四月-2011 17:56:5221-四月-2011 17:56:53 (1sec) > Map14120221-四月-2011 17:56:5521-四月-2011 17:58:59 (2mins, 4sec) > Reduce880021-四月-2011 17:57:1321-四月-2011 17:59:04 (1mins, 51sec) > Cleanup110021-四月-2011 17:58:5521-四月-2011 17:58:56 (1sec) > > > > why map process have 2 killed tasks. and what factor make hadoop kill the > task? > Thanks. > jbm. -- Harsh J
