This Ruby script is aimed to be used in hbase shell. Just write that file in your current working directory and execute it with hbase shell lr.rb

You'll then have a new list_regions(table) function which give you SERVERNAME, REGIONNAME, STARTKEY, ENDKEY, SIZE(MB), OCCUPANCY(%), REQUESTS The occupancy is computed over the specific region size of the table or the default parameter.

This script was defined to merge tiny regions in HBase : when dealing with a high throughput writing environment, regions can easily multiply even if you sized regions accordingly. In my case I did calculate 20 GB region size for having around 250 regions. I wasn't expecting this number getting much more than the double with splitting, so I was surprised when regions exceeded 1300 !

Then, having a look on regions, I realized lot of them were tiny regions, so I decided to merge them to get a more decent regions number.

The idea is to launch the script with ./hbase_region_repartition.sh TABLE_NAME THRESHOLD MAX_REGIONS

THRESHOLD means we want to merge adjacent regions not exceeding THRESHOLD percent of region size defined MAX_REGIONS means that we don't want to merge more than MAX_REGIONS at a time

This is used for backuping all Hadoop configurations with Ambari API. You can for example use this script before an Ambari migration so that you'll be able to compare a before/after configs with diff to see if some parameters has been adjusted/added/removed. Just modify the parameters in the script and launch it without parameters.

You can easily export a cluster structure with the Ambari API (blueprints). When you want to submit a blueprint for creating a new cluster, you got to provide a host mapping json file which fits the blueprint hostgroups definition. The goal of this script is to recreate the actual host mapping json file, if you want to recreate the same cluster on the same hosts for some reason.