Running Flume on HBase
----------------------

1) Compile top-level flume
   $ ant

2) Copy your version of hbase-*.jar into the plugins/hbasesink/lib
   I tested against hbase-0.89.20100621+17.jar

3) Compile hbase-sink from this directory
   ../plugins/hbasesink$ ant

4) Successful compilation will produce "hbase_sink.jar" in ../plugins/hbasesink/

5) Add the plugin classes in flume/conf/flume-site.xml
<configuration>
  <property>
    <name>flume.plugin.classes</name>
    <value>com.cloudera.flume.hbase.Attr2HBaseEventSink</value>
    <description>Comma separated list of plugin classes</description>
  </property>
</configuration>


6) Include the jars in the "FLUME_CLASSPATH"
From the terminal where you would start the Flume nodes:
  $ export FLUME_CLASSPATH=/your_path/flume/plugins/hbasesink/hbase_sink.jar:/your_path/hbase-*/conf/
  Include all the jars that your plugin refers to

Here's how I've run flume against hbase in dev mode:

7) From your hbase-*, "bin/start-hbase.sh"

Note that starting hbase will also start ZooKeeper

8) using the hbase shell create a table for the flume sink to write events

$ bin/hbase shell
> create 't1', 'f1', 'f2'

9) Undertand the attr2hbase sink usage and configure the source:
   "usage: attr2hbase(\"table\" [,\"sysFamily\"[, \"writeBody\"[,\"attrPrefix\"[,\"writeBufferSize\"[,\"writeToWal\"]]]]])
   Refer parameter_mapping.html in /docs for more details.

10) start flume, I started a node with a rssatomSource and attr2hbase("t1","f1","f2:event","attr-prefix","10","false")'

> scan 't1'
(This is output of my rowkey and data)

hbase(main):002:0> scan 't1'
ROW                          COLUMN+CELL
 \x00\x00\xF6_\x0Fk\xF4\x80  column=f2:event, timestamp=1274227444388, value=hello
 \x00\x00\xF6_\x0Fk\xF4\x80  column=f1:host, timestamp=1274227444388, value=valhalla
 \x00\x00\xF6_\x0Fk\xF4\x80  column=f1:timestamp, timestamp=1274227444388, value=\x00\x00\x01\x28\xAD\xDF\xCA\
                             x7C
1 row(s) in 0.0550 seconds