MapReduce health checks
There are many factors that impact the performance of a job or application submitted to the cluster. The important checks, which can help narrow down the bottlenecks and help in improving the performance, can be many, but the few important ones are as follows:
Health of JobTracker or the RM
Backlog of tasks in the cluster; make sure that the number of tasks does not cross the upper limit of the maximum tasks supported in the cluster
Localities of the tasks run to make sure that there is minimal across-rack traffic
Health of TaskTracker and other components like NodeManager depending upon the MR version
The above checks are very well-documented and talked about at the Cloudera website. Please read them for further understanding.
