Installing Hbase on Local Machine
Prerequisites:
1) Requires Java 6
2) Setup Hadoop on the node where HBase will be installed
3) The latest version of HBase can be downloaded from Hbase Releases
There are 3 ways in which Hbase can be installed.
For all the installation modes update conf/hbase-env.sh to point to the correct JAVA HOME location, also uncomment the same.
Standalone Mode:
In this Hbase will not use Hadoop as a file system, but will
use system's default file system. Just change one configuration,
hbase.rootdir in hbase-site.xml. If this is not changed then the
data will be stored in its default directory i.e. in tmp folder and data
will be lost on system reboot. Following configuration can be used:
Pseudo Distributed Mode: Hbase
will run on a single node only but will use Hadoop as its underlying
filesystem. It will use zookeeper also. This should be tried for
development enviornment. Following steps must be performed so set up Hbase in this mode:Update entry in hosts file: In etc/hosts file remove following entry
"127.0.1.1”
Delete this line. If this not removed then when Hbase Master tries to connect to HDFS it will find two IP address which are mapped for your system. And will most probably pick the second one i.e. 127.0.1.1 , at that port it can't find HDFS it will throw exceptions ConnectException
Note:
In Linux enviornment everytime we reboot our system the entries in
Hosts file change to their default values. So change the values
everytime.
Update config file (hbase-site.xml):
1) Set hbase.rootdir property to use HDFS file system, its value should be same as fs.default.name found in Hadoop core-site.xml. A sample property:
E.g: hdfs://NAMENODE_SERVER:PORT/HBASE_ROOTDIR
If any specific directory is mentioed in root directory (in our example Hbase directory). Then we need to create this directory in our HDFS. Run the below command from Hadoop/bin installation and give the newly created folder appropriate rights.
sudo -u
Also set hbase.cluster.distributed in config to true
Set base.tmp.dir: To a custom directory.
Update regionservers (in conf directory):
Set your ip or system name
Update regionservers (in conf directory):
Set your ip or system name
Note:
Make sure in mapred-site.xml of Hadoop all the address should be on IP
address instead of localhost. If this is not changed then running
HRegion services/APIs will throw port bind exception
Changes to be made on Hadoop
conf/hdfs-site.xml:
dfs.data.dir: Comma separated list of paths on the local filesystem of a DataNode where it should store its blocks.
Note: If this parameter is not set then its default value is the tmp directory. So on reboot all HBase data will be lost.
Run HBase
· Start Hadoop. If running HBase in Pseudo Distributed/Distributed mode start Hadoop first. From Hadoop installation directory run command:
bin/start-all.sh
· Start hbase. From Hbase installation directory run command
bin/start-hbase.sh
If run successfully this should show output like:
starting master, logging to /home/Installations/Hbase/hbase-0.90.3/bin/../logs/hbase-sasomani-master-impetus-1025.out
· Check HBase is up and running
Open up the HBase Web UI
· Go to HBase Shell: From Hbase installation directory type command
bin/hbase shell
· Create table in HBase: From Hbase shell type command on hbase (main)
create 'myLittleHBaseTable1', 'myLittleColumn'
To check on console
scan 'myLittleHBaseTable1'
Tips and TricksIf while starting, the error is encountered as “org.apache.hadoop.hdfs.protocol.ClientProtocol version mismatch”, the solution is to put the hadoop core jar in hbase/lib directory.Sometime, the master would not start throwing the exception of “Unable to connect”. To resolve the same, make the below entry in the /etc/hosts file
