Starting and Stopping HBase

Use these instructions to start, stop, restart, rolling restart, or decommission HBase clusters or individual hosts.

Starting or Restarting HBase

You can start HBase hosts individually or as an entire cluster.

Starting or Restarting HBase Using Cloudera Manager

  1. Go to the HBase service.
  2. Click the Actions button and select Start.
  3. To restart a running cluster, click Actions and select Restart or Rolling Restart. A rolling restart, which restarts each RegionServer, one at a time, after a grace period. To configure the grace period, see Configuring the Graceful Shutdown Timeout Property.
  4. The Thrift service has no dependencies and can be restarted at any time. To stop or restart the Thrift service:
    • Go to the HBase service.
    • Select Instances.
    • Select the HBase Thrift Server instance.
    • Select Actions for Selected and select either Stop or Restart.

Starting or Restarting HBase Using the Command Line

If you need the ability to perform a rolling restart, Cloudera recommends managing your cluster with Cloudera Manager.

  1. To start a HBase cluster using the command line, start the HBase Master by using the sudo hbase-master start command on RHEL or SuSE, or the sudo hadoop-hbase-regionserver start command on Ubuntu or Debian. The HMaster starts the RegionServers automatically.
  2. To start a RegionServer manually, use the sudo hbase-regionserver start command on RHEL or SuSE, or the sudo hadoop-hbase-regionserver start command on Ubuntu or Debian. Running multiple RegionServer processes on the same host is not supported.
  3. The Thrift service has no dependencies and can be restarted at any time. To start the Thrift server, use the hbase-thrift start on RHEL or SuSE, or the hadoop-hbase-thrift start on Ubuntu or Debian.

Stopping HBase

You can stop a single HBase host, all hosts of a given type, or all hosts in the cluster.

Stopping HBase Using Cloudera Manager

  1. To stop or decommission a single RegionServer:
    1. Go to the HBase service.
    2. Click the Instances tab.
    3. From the list of Role Instances, select the RegionServer or RegionServers you want to stop or decommission.
    4. Select Actions for Selected and select either Decommission (Graceful Stop) or Stop.
      • Graceful Stop causes the regions to be redistributed to other RegionServers, increasing availability during the RegionServer outage. Cloudera Manager waits for an interval determined by the Graceful Shutdown timeout interval, which defaults to three minutes. If the graceful stop does not succeed within this interval, the RegionServer is stopped with a SIGKILL (kill -9) signal. Recovery will be initiated on affected regions.
      • Stop happens immediately and does not redistribute the regions. It issues a SIGTERM (kill -5) signal.
  2. To stop or decommission a single HMaster, select the Master and go through the same steps as above.
  3. To stop or decommission the entire cluster, select the Actions button at the top of the screen (not Actions for selected) and select Decommission (Graceful Stop) or Stop.

Stopping HBase Using the Command Line

  1. Shut down the Thrift server by using the hbase-thrift stop command on the Thrift server host. sudo service hbase-thrift stop
  2. Shut down each RegionServer by using the hadoop-hbase-regionserver stop command on the RegionServer host.
    sudo service hadoop-hbase-regionserver stop
  3. Shut down backup HMasters, followed by the main HMaster, by using the hbase-master stop command.
    sudo service hbase-master stop