Each version of MR3 has three releases which use different versions of Hadoop for their compilation. For MR3 1.1, we have the following releases:

  • master uses Hadoop 3.1 as its base. It can be used on Hadoop 3.1 and later.
  • base2.8 uses Hadoop 2.8 as its base. It can be used on Hadoop 2.8 and later.
  • base2.7 uses Hadoop 2.7 as its base. It can be used on Hadoop 2.7 and later.

Note that base2.7 can be used not only on Hadoop 2.7 but also on higher versions such as Hadoop 2.8 and 3.1. Here is the guideline for choosing an MR3 release:

Hive on Hadoop on Kubernetes
Hive 2 base2.7 base2.7
Hive 3 base2.7, base2.8, master master
Hive 4 master master
  • For running Hive 2 on MR3, the user should choose base2.7, whether on Hadoop or on Kubernetes.
  • For running Hive 3 on MR3 on Hadoop, the user can use base2.7, base2.8, or master.
  • For running Hive 3 on MR3 on Kubernetes, we recommend master.
  • For running Hive 4 on MR3, we recommend master, both on Hadoop or on Kubernetes.

Running Hive 3 on MR3 on Cloudera CDH gives a special case:

  • On Cloudera CDH, the user should choose base2.7.

Download an MR3 release and uncompress it. We rename the new directory to mr3-run.

$ wget https://github.com/mr3project/mr3-release/releases/download/v1.1/mr3-1.1-base2.7.tar.gz
$ gunzip -c mr3-1.1-base2.7.tar.gz | tar xvf -;
$ mv mr3-1.1-base2.7 mr3-run
$ wget https://github.com/mr3project/mr3-release/releases/download/v1.1/mr3-1.1-base2.8.tar.gz
$ gunzip -c mr3-1.1-base2.8.tar.gz | tar xvf -;
$ mv mr3-1.1-base2.8 mr3-run
$ wget https://github.com/mr3project/mr3-release/releases/download/v1.1/mr3-1.1-master.tar.gz
$ gunzip -c mr3-1.1-master.tar.gz | tar xvf -;
$ mv mr3-1.1-master mr3-run