Hadoop HBase

Comparison

RDBMS

HBase

Overview

Master

HBase Master Node orchestrates a cluster of one or more Region Servers

• Bootstrap a virgin install
• Assign regions to Region Servers
• Recover Region Server failure

Region Server

• Carry regions
• Respond to client read/write requests
• Inform Master Node the new region

Zookeeper

The authority on cluster state. The Zookeeper ensemble hosts vital:

• Location of the hbase:meta table
• Address of Master
• Host the assignment transaction state (for recovery purpose)

Client

When bootstrapping a client connection to HBase cluster, the client must be passed the location of Zookeeper ensemble. Thereafter, client navigates the ZooKeeper hierarchy to learn about cluster attributes  

hbase:meta

HBase keep a special catalog table named hbase:meta, within which it maintains the current list, state and locations of all user-space regions afloat on the cluster. Entries in hbase:meta are keyed by region name, which is 

• table_name,start_row,creation_time.MD5(table_namestart_rowcreation_time).

Fresh clients connect to the ZooKeeper cluster first to learn the location of hbase:meta. The client then does a lookup against the appropriate hbase:meta region to figure out the hosting user-space region and its location. Thereafter, the client interacts directly with the hosting region server.

As region transitions - split, disabled, enabled, deleted or redeployed, the catalog table is updated.

HBase Commands

By default, HBase writes to /${java.io.tmpdir}/hbase-${user.name}.${java.io.tmpdir} usually maps to /tmp, but you should configure HBase to use a more permanent location by setting hbase.tmp.dir in hbase-site.xml. In standalone mode, the HBase master, the regionserver, and a Zookeeper instance are all run in the same JVM.

• HBase Shell

%hbase shell

# Create a table with one column family (with default column family attributes)
hbase(main):001:0>create 'table_name', 'column_family_name'

# Display tables 
hbase(main):002:0>list

# Insert some data
hbase(main):003:0>put 'table_name' 'row1', 'column_family_name:column_name', 'value1'
hbase(main):004:0>put 'table_name' 'row2', 'column_family_name:column_name', 'value2'
hbase(main):005:0>put 'table_name' 'row3', 'column_family_name:column_name', 'value3'

# Get a particular row with the row key
hbase(main):006:0>get 'table_name' 'row1'

# List the table content
hbase(main):007:0>scan 'table_name'

Client

Java Client, REST, Thrift and MapReduce API.

Load Data

For a large dataset, first, copy the raw file into HDFS, and then run a MapReduce job that can read the input and write to HBase.

 

转载于:https://my.oschina.net/u/3551123/blog/1486627