Hadoop series-job submission process on Yarn

Insert picture description here

Submission process

  1. Assignment submission
  • ClientCall the job.waitForCompletionmethod to submit MapReduce jobs to the entire cluster.
  • ClientTo ResourceManagera job application id.
  • RMClientThe path and job submitted to the returned resource id.
  • ClientSubmit the jar package, slice information, and xml configuration file to the specified path.
  • ClientAfter the submission is completed, MRapply for operation mrApplicationMaster.
  1. Job initialization
  • RMAfter receiving Clientthe request, convert the job into a Task and submit it to the task queue.
  • A certain person is free to NodeManagerreceive the job.
  • NMCreate Containerand apply for operation mrApp.
  • NMDownload the Clientsubmitted resources.
  1. Task Assignment
  • mrAppRMRun multiple MapTasktasks to request .
  • RMAssign the mrAppsubmitted tasks to multiple NMand NMstart them separately MapTask.
  • MapTaskAfter running, run mrAppto the RMrequest ReduceTask.
  • RduceTaskFrom the MapTaskacquired data the respective partition.
  • After the program is complete, MRthe RMcancellation of their own.

Guess you like

Origin blog.csdn.net/Cxf2018/article/details/109840837