Database Reference
In-Depth Information
The Start/stop HBase cluster
Starting or stopping an HBase cluster is a very common task that is performed by
any administrator; any kind of coniguration change for tuning is a very common
scenario. The dependent systems such as HDFS and ZooKeeper are assumed to be
running before HBase starts/stops. Apache HBase provides the following script
iles under the $HBASE_HOME/bin directory to be used for the start/stop operation:
hbase-daemon.sh : This ile is used to start/stop/restart a speciic daemon
as follows:
$HBASE_HOME/bin/hbase-daemon.sh [start/stop/restart]
[regionserver/master]
This command needs to be manually executed on each box.
hbase-daemons.sh : This ile is used to start/stop/restart all the daemons
as follows:
$HBASE_HOME/bin/hbase-daemons.sh --hosts regionserversfile [start/
stop/restart]
This script ile wraps the hbase-daemon.sh ile and is mainly used to restart
all the services on any cluster. This script also requires a password-less SSH
from the host running the script to all target hosts.
start-hbase.sh : This script ile typically runs from the master server to
start the complete HBase cluster. It wraps the hbase-daemons.sh ile to
perform this operation. The following command is an example:
$HBASE_HOME/bin/start-hbase.sh
stop-hbase.sh : Similar to start-hbase.sh , it stops the complete
HBase cluster.
Adding nodes
As the application goes on and the workload increases, the size of the cluster
increases as well. Assuming that the Hadoop datanode and region server share
the same physical boxes, the following steps are performed to add a node:
1.
Start the datanode service.
2.
Start the RegionServer service and verify it.
3.
Run the "balancer" command on the master server after opening the HBase
shell. This will move some regions from all existing RegionServers to the
newly added RegionServer and also balance the workload across the cluster.
 
Search WWH ::




Custom Search