Kind of a newbie question but one that's been stumping me...
I'm noticing when I set the number of mappers to use (i.e. -m in CLI Sqoop command) that isn't the number of mappers my cluster is using to execute the Sqoop job.
For instance, I just ran an import with -m 8 specified and the number of mappers used is 4 until the import hits 50% and then drops to 3 for the rest of the import process (my cluster has 16 available map slots; no other concurrent jobs processing).
I read through the "Controlling Parallelism" piece of the user guide but didn't seem to find my answer.
Thanks in advanced for the help,
Abraham Elmahrek 2013-12-13, 22:53