I use this tool for collecting performance metrics from HBase to Graphite, via StatsD.
It works by polling HBase's web API for JMX data, extracts interesting values, and sends them to StatsD.

Basic configuration

Configuration is done via client.cfg.
Copy and customize client.cfg.example.

Usage

Install the statsd pip package.
Invoke main.py, like:

python main.py

Measured metrics

Assuming cluster name (defined in client.cfg) "CLUSTERNAME"

Master servers
Assuming table "TABLENAME" and region "REGIONNAME"

• CLUSTERNAME.tables.TABLENAME.regioncount
• CLUSTERNAME.tables.TABLENAME.REGIONNAME.readRate
• CLUSTERNAME.tables.TABLENAME.REGIONNAME.writeRate

Region servers
Assuming server name (defaults to server FQDN with . replaced to -) "SERVERNAME"

• CLUSTERNAME.SERVERNAME.compactionQueueSize
• CLUSTERNAME.SERVERNAME.flushQueueSize
• CLUSTERNAME.SERVERNAME.readReqeustsCount
• CLUSTERNAME.SERVERNAME.regionCount
• CLUSTERNAME.SERVERNAME.running-tasks
• CLUSTERNAME.SERVERNAME.readRate
• CLUSTERNAME.SERVERNAME.writeRate

Finding it in Graphite

Because I'm using StatsD, the metric CLUSTERNAME.SERVERNAME.compactionQueueSize can be found as stats.gauges.hbase.CLUSTERNAME.SERVERNAME.compactionQueueSize.