Why we are configuring mapred job tracker in YARN

Question

With the introduction of YARN JobTracker and TaskTracker was replaced with Resource Manager and NodeManager. But, in some of the tutorials for installing Hadoop 2.6.0 or 2.7.0 or 2.8, I have noticed that they are still setting mapreduce.framework.name property to yarn. Besides this, they are also setting mapred.job.tracker property.

I checked the official doc and I found the the description for mapred.job.tracker property as

"The host and port that the MapReduce job tracker runs at. If "local", then jobs are run in-process as a single map and reduce task."

Now, I am curious and confused at the same time about the fact that why we need to configure JobTracker in case of YARN?

Please let me know the reason behind this.

Answer 1

I really dont know the reason behind setting job tracker in YARN w.r.t. The tutorials that you have visited. To be honest, I think they have not read about YARN and are just following their old habit of configuring job tracker. One other plausible reason for doing so to run the hadoop without YARN, just for sake of back up (may be - just a hunch).

Strictly speaking, the JobTracker and TaskTracker does not come with YARN. Even if add theses properties, they will be ignored. In fact the properties has been changed. Here are the new properties w.r.t. Resource manager (JobTracker) and node manager (TaskTracker):

yarn.resourcemanager.address to replace mapred.jobtracker.address.