Skip to main content
Version: 2.1.1

Hbase

Description​

Use hbase-connectors to output data to Hbase , Hbase (>=2.1.0) and Spark (>=2.0.0) version compatibility depends on hbase-connectors . The hbase-connectors in the official Apache Hbase documentation is also one of the Apache Hbase Repos.

tip

Engine Supported and plugin name

  • Spark: Hbase
  • Flink

Options​

nametyperequireddefault value
hbase.zookeeper.quorumstringyes
catalogstringyes
staging_dirstringyes
save_modestringnoappend
hbase.*stringno

hbase.zookeeper.quorum [string]​

The address of the zookeeper cluster, the format is: host01:2181,host02:2181,host03:2181

catalog [string]​

The structure of the hbase table is defined by catalog , the name of the hbase table and its namespace , which columns are used as rowkey, and the correspondence between column family and columns can be defined by catalog hbase table catalog

staging_dir [string]​

A path on HDFS that will generate data that needs to be loaded into hbase . After the data is loaded, the data file will be deleted and the directory is still there.

save_mode [string]​

Two write modes are supported, overwrite and append . overwrite means that if there is data in the hbase table , truncate will be performed and then the data will be loaded.

append means that the original data of the hbase table will not be cleared, and the load operation will be performed directly.

hbase.* [string]​

Users can also specify multiple optional parameters. For a detailed list of parameters, see Hbase Supported Parameters.

If these non-essential parameters are not specified, they will use the default values given in the official documentation.

common options [string]​

Sink plugin common parameters, please refer to Sink Plugin for details

Examples​

 hbase {
source_table_name = "hive_dataset"
hbase.zookeeper.quorum = "host01:2181,host02:2181,host03:2181"
catalog = "{\"table\":{\"namespace\":\"default\", \"name\":\"customer\"},\"rowkey\":\"c_custkey\",\"columns\":{\"c_custkey\":{\"cf\":\"rowkey\", \"col\":\"c_custkey\", \"type\":\"bigint\"},\"c_name\":{\"cf\":\"info\", \"col\":\"c_name\", \"type\":\"string\"},\"c_address\":{\"cf\":\"info\", \"col\":\"c_address\", \"type\":\"string\"},\"c_city\":{\"cf\":\"info\", \"col\":\"c_city\", \"type\":\"string\"},\"c_nation\":{\"cf\":\"info\", \"col\":\"c_nation\", \"type\":\"string\"},\"c_region\":{\"cf\":\"info\", \"col\":\"c_region\", \"type\":\"string\"},\"c_phone\":{\"cf\":\"info\", \"col\":\"c_phone\", \"type\":\"string\"},\"c_mktsegment\":{\"cf\":\"info\", \"col\":\"c_mktsegment\", \"type\":\"string\"}}}"
staging_dir = "/tmp/hbase-staging/"
save_mode = "overwrite"
}

This plugin of Hbase does not provide users with the function of creating tables, because the pre-partitioning method of the hbase table will be related to business logic, so when running the plugin, the user needs to create the hbase table and its pre-partition in advance; for rowkey Design, catalog itself supports multi-column combined rowkey="col1:col2:col3" , but if there are other design requirements for rowkey , such as add salt , etc., it can be completely decoupled based on the transform plugin pair rowkey is modified.