HBase 连接器

Flink-1.12

官方文档

source 表

简述

离线批的方式读取数据

示例

• HBase 中 column families 必须声明为 Row 类型，可以不用定义所有column families 和 qualifier names。 e.g. base_info 映射为 HBase 的 column family name，dt、age、behavior 映射为 HBase 的 qualifier name。
• 必须设置：SET ‘stream_mode’ = ‘false’;
• 以元数据的方式配置krb认证连接时，需要在set中配置相关的文件信息
• ddl 方式

SET 'stream_mode' = 'false';

CREATE TABLE source (
    user_id VARCHAR,
    base_info row<age VARCHAR>
) WITH (
    'connector' = 'hbase',
    'connector.version' = '2.2.1',
    'table-name' = 'user_dim',
    'zookeeper.quorum' = 'a1.service.163.org:2181,a2.service.163.org:2181',
    'zookeeper.znode.parent' = '/hbase'
);

• 元数据方式

SET 'stream_mode' = 'false';

set 'user_dim.connector.version' = '1.4.3';
set 'user_dim.format.schema' = 'ROW<user_id VARCHAR, base_info ROW<age VARCHAR>>';
set 'user_dim.core-site' = 'core-site.xml',
set 'user_dim.hdfs-site' = 'hdfs-site.xml',
set 'user_dim.hbase-site' = 'hbase-site.xml',
set 'user_dim.krb.conf' = 'krb5.conf',
set 'user_dim.krb.keytab' = 'hbase.keytab',
set 'user_dim.krb.principal' = 'hbase/sloth-commerce-test1.jd.163.org@BDMS.163.COM',
set 'user_dim.format.schema' = 'ROW<user_id VARCHAR, base_info ROW<age VARCHAR>>';

set 'user_pvuv_sink.connector.version' = '1.4.3';
set 'user_pvuv_sink.format.schema' = 'ROW<user_id VARCHAR, base_info ROW<age VARCHAR>>';

insert into hbase_docker_one.`default`.user_pvuv_sink
select a.user_id, ROW(age) as base_info
from  hbase_docker_one.`default`.user_dim a;

维表

• ddl

-- source
CREATE TABLE user_log (
    user_id VARCHAR,
    item_id VARCHAR,
    category_id VARCHAR,
    behavior VARCHAR,
    ts VARCHAR
) WITH (
    'connector' = 'memory',
    'table.name' = 'user_log'
);

--dim
CREATE TABLE dim_table (
    user_id VARCHAR,
    base_info row<age VARCHAR>
) WITH (
    'connector' = 'hbase',
    'connector.version' = '2.2.1',
    'table-name' = 'user_dim',
    'zookeeper.quorum' = 'ddbadmin.service.163.org:2181',
    'zookeeper.znode.parent' = '/hbase',
    'lookup.cache.type' = 'ALL'
    ,'lookup.async' = 'true'
);

-- sink
CREATE TABLE pvuvage_sink (
    dt VARCHAR,
    age INT,
    pv BIGINT,
    uv BIGINT,
    PRIMARY KEY(dt,age) NOT ENFORCED
) WITH (
    'connector' = 'jdbc',
    'url' = 'jdbc:mysql://10.122.173.167:3306/flink-test',
    'table-name' = 'pvuv_age_sink_hbase',
    'username' = 'sys',
    'password' = 'netease'
);

INSERT INTO pvuvage_sink
select d.ts as dt, d.age, count(*) as pv, count(distinct d.user_id) as uv
from (
SELECT
  u.ts,
  cast(w.base_info.age as int)  as age,
  u.user_id,
  u.behavior
FROM (select user_id,item_id,category_id,behavior,ts,PROCTIME() as proc from user_log) as u
join dim_table for system_time as of u.proc as w
on u.user_id = w.user_id
where w.base_info.age > 10
) as d
GROUP BY d.ts, d.age;

• 元数据方式 因元数据中心不支持 row 形式的 schema 配置，需要在set中定义schema

set 'user_dim.connector.version' = '2.2.1';
set 'user_dim.lookup.async' = 'true';
set 'user_dim.format.schema' = 'ROW<user_id VARCHAR, base_info ROW<age VARCHAR>>';

-- source
CREATE TABLE user_log (
    user_id VARCHAR,
    item_id VARCHAR,
    category_id VARCHAR,
    behavior VARCHAR,
    ts VARCHAR
) WITH (
    'connector' = 'memory',
    'table.name' = 'user_log'
);

-- sink
CREATE TABLE pvuvage_sink (
    dt VARCHAR,
    age INT,
    pv BIGINT,
    uv BIGINT,
    PRIMARY KEY(dt,age) NOT ENFORCED
) WITH (
    'connector' = 'jdbc',
    'url' = 'jdbc:mysql://10.122.173.167:3306/flink-test',
    'table-name' = 'pvuv_age_sink_hbase',
    'username' = 'sys',
    'password' = 'netease',
    'sink.buffer-flush.max-rows' = '1'
);

INSERT INTO pvuvage_sink
select d.ts as dt, d.age, count(*) as pv, count(distinct d.user_id) as uv
from (
SELECT
  u.ts,
  cast(w.base_info.age as int)  as age,
  u.user_id,
  u.behavior
FROM (select user_id,item_id,category_id,behavior,ts,PROCTIME() as proc from user_log) as u
join hbase_docker_two.`default`.user_dim for system_time as of u.proc as w
on u.user_id = w.user_id
where w.base_info.age > 10
) as d
GROUP BY d.ts, d.age;

sink 表

• ddl

  -- sink
  CREATE TABLE pvuvage_sink (
    user_id VARCHAR,
    base_info row<age VARCHAR>
  ) WITH (
    'connector' = 'hbase',
    'connector.version' = '2.2.1',
    'zookeeper.quorum' = 'ddbadmin.service.163.org:2181',
    'zookeeper.znode.parent' = '/hbase',
    'table-name' = 'user_pvuv_sink',
    'sink.buffer-flush.max-size' = '10mb',
    'sink.buffer-flush.max-rows' = '1000',
    'sink.buffer-flush.interval' = '1s'
  );

• 元数据方式

  set 'sloth_ua_one.format.schema' = 'ROW<user_id VARCHAR, base_info ROW<item_id VARCHAR, category_id varchar, behavior varchar, ts VARCHAR>>';
  set 'sloth_ua_one.connector.version' = '2.2.1';
  -- source
  CREATE TABLE user_log (
    user_id VARCHAR,
    item_id VARCHAR,
    category_id VARCHAR,
    behavior VARCHAR,
    ts VARCHAR
  ) WITH (
    'connector' = 'memory',
    'table.name' = 'user_log'
  );
  INSERT INTO
  hbase_docker_two.`default`.sloth_ua_one
  SELECT
  user_id,
  ROW(item_id , category_id, behavior, ts) as base_info
  FROM
  user_log;

  With 参数 / set 参数

• 针对某个catalog设置参数的，需要在下表的参数前添加catalog名称。如上述的: user_dim.connector.version

Flink-1.10

• Flink HBase Connector 官方文档

  HBase Sink

  简述

  Easystream 支持输出到 HBase。

  示例

CREATE TABLE pvuvage_sink (
    hbase_rowkey_name VARCHAR,
    base_info row<dt VARCHAR, age INT, behavior VARCHAR>
) WITH (
    'connector.type' = 'hbase',
    'connector.version' = '1.4.3',
    'connector.property.version' = '2',
    'connector.zookeeper.quorum' = '*',
    'connector.zookeeper.znode.parent' = '*',
    'connector.table-name' = 'user_pvuv_sink',
    'connector.write.buffer-flush.max-size' = '10mb',
    'connector.write.buffer-flush.max-rows' = '1000',
    'connector.write.buffer-flush.interval' = '1s'
);

• HBase 中 column families 必须声明为 Row 类型，可以不用定义所有column families 和 qualifier names。 e.g. base_info 映射为 HBase 的 column family name，dt、age、behavior 映射为 HBase 的 qualifier name。

With 参数

HBase 维表

简述

Easystream 支持使用 HBase 表作为维表进行 join。

示例

CREATE TABLE user_log (
    user_id VARCHAR,
    item_id VARCHAR,
    category_id VARCHAR,
    behavior VARCHAR,
    ts TIMESTAMP
) WITH (
    'connector.type' = 'kafka',
    'connector.version' = 'universal',
    'connector.topic' = 'user_behavior',
    'connector.startup-mode' = 'latest-offset',
    'connector.properties.zookeeper.connect' = 'xxx',
    'connector.properties.bootstrap.servers' = 'xxx',
    'connector.properties.group.id' = 'xxx',
    'update-mode' = 'append',
    'format.type' = 'json',
    'format.derive-schema' = 'true'
);
--dim
CREATE TABLE dim_table (
    user_id VARCHAR,
    base_info row<age VARCHAR, tall varchar>,
    info row<gener VARCHAR, children varchar>
) WITH (
    'connector.type' = 'hbase',
    'connector.version' = '1.4.3',
    'connector.property.version' = '2',
    'connector.table-name' = 'user_dim',
    'connector.zookeeper.quorum' = '*',
    'connector.zookeeper.znode.parent' = '*',
    'connector.lookup.cache.type' = 'lru'
);

-- sink
CREATE TABLE pvuvage_sink (
    dt VARCHAR,
    age INT,
    zhibiao1 BIGINT,
    zhibiao2 BIGINT
) WITH (
    'connector.type' = 'jdbc',
    'connector.url' = 'jdbc:mysql://10.122.173.167:3306/flink-test',
    'connector.table' = 'pvuv_age_sink_hbase',
    'connector.username' = '*',
    'connector.password' = '*',
    'connector.write.flush.max-rows' = '1'
);

INSERT INTO pvuvage_sink
select DATE_FORMAT(d.ts, 'yyyy-MM-dd HH:00') as dt, d.age, sum(d.tall) as zhibiao1, sum(d.childNum) as zhibiao2
from (
SELECT
  u.ts,
  cast(w.age as int)  as age,
  cast(w.tall as bigint) as tall,
  cast(w.children as bigint) as childNum,
  u.user_id,
  u.behavior
FROM (select user_id,item_id,category_id,behavior,ts,PROCTIME() as proc from user_log) as u
left join dim_table for system_time as of u.proc as w
on u.user_id = w.user_id
where w.age > 10
) as d
GROUP BY DATE_FORMAT(d.ts, 'yyyy-MM-dd HH:00') , d.age;

• Flink HBase Connector 官方文档

• Flink 维表 join 官方文档

With 参数

• 带 krb 认证的配置同版本1.12