HBase操作

基本操作

创建表

Examples:

  hbase> create 't1', {NAME => 'f1', VERSIONS => 5}
  hbase> create 't1', {NAME => 'f1'}, {NAME => 'f2'}, {NAME => 'f3'}
  hbase> # The above in shorthand would be the following:
  hbase> create 't1', 'f1', 'f2', 'f3'
  hbase> create 't1', {NAME => 'f1', VERSIONS => 1, TTL => 2592000, BLOCKCACHE => true}
  hbase> create 't1', 'f1', {SPLITS => ['10', '20', '30', '40']}
  hbase> create 't1', 'f1', {SPLITS_FILE => 'splits.txt'}
  hbase> # Optionally pre-split the table into NUMREGIONS, using
  hbase> # SPLITALGO ("HexStringSplit", "UniformSplit" or classname)
  hbase> create 't1', 'f1', {NUMREGIONS => 15, SPLITALGO => 'HexStringSplit'}

create 'dbname:newsinfo_anticheat_user_tag_data', 'user', 'device'

获得表的描述

hbase(main):006:0> describe "dbname:newsinfo_anticheat_user_tag_data"  

插入几条记录

put 'dbname:newsinfo_anticheat_user_tag_data', '5a483b8769e9560001f9d1b9_20181224', 'user:phone', '190eb638-185d-3e58-a009-fcd69f67b8ac'
put 'dbname:newsinfo_anticheat_user_tag_data', '596f51ba4e94d9000170e1ff_20181224', 'user:phone', '17086385281'

查看所有数据scan

hbase(main):014:0> scan 'dbname:newsinfo_anticheat_user_tag_data'
ROW                                              COLUMN+CELL
596f51ba4e94d9000170e1ff_20181223               column=user:phone, timestamp=1545646705395, value=17086385281

596f51ba4e94d9000170e1ff_20181224               column=user:phone, timestamp=1545646716425, value=17086385281                      

获得数据 get

获得一行的所有数据

get 'dbname:newsinfo_anticheat_user_tag_data', '5a483b8769e9560001f9d1b9_20181224'

获得某行,某列族的所有数据

get 'dbname:newsinfo_anticheat_user_tag_data', '5a483b8769e9560001f9d1b9_20181224','user'

获得某行,某列族,某列的所有数据

get 'dbname:newsinfo_anticheat_user_tag_data', '5a483b8769e9560001f9d1b9_20181224','user:dt'

预分区

默认情况下,在创建HBase表的时候会自动创建一个region分区,当导入数据的时候,所有的HBase客户端都向这一个region写数据,直到这个region足够大了才进行切分。一种可以加快批量写入速度的方法是通过预先创建一些空的regions,这样当数据写入HBase时,会按照region分区情况,在集群内做数据的负载均衡。
命令方式:

create ‘t1’, ‘f1’, {NUMREGIONS => 15, SPLITALGO => ‘HexStringSplit’}
也可以使用api的方式:
bin/hbase org.apache.hadoop.hbase.util.RegionSplitter test_table HexStringSplit -c 10 -f info
参数:
test_table是表名
HexStringSplit 是split 方式
-c 是分10个region
-f 是family

这样就可以将表预先分为15个区,减少数据达到storefile 大小的时候自动分区的时间消耗,并且还有以一个优势,就是合理设计rowkey 能让各个region 的并发请求平均分配(趋于均匀) 使IO 效率达到最高,但是预分区需要将filesize 设置一个较大的值,设置哪个参数呢, hbase.hregion.max.filesize 这个值默认是10G 也就是说单个region 默认大小是10G,
这个参数的默认值在0.90 到0.92到0.94.3各版本的变化:256M--1G--10G
但是如果MapReduce Input类型为TableInputFormat 使用hbase作为输入的时候,就要注意了,每个region一个map,如果数据小于10G 那只会启用一个map 造成很大的资源浪费,这时候可以考虑适当调小该参数的值,或者采用预分配region的方式,并将检测如果达到这个值,再手动分配region。

HBase已有表与Phoenix映射

使用phoenix 视图方式映射

初始创建

查看HBASE 已有表dbname:newsinfo_anticheat_tag_data

hbase(main):003:0> scan ' dbname:newsinfo_anticheat_tag_data'
ROW                                              COLUMN+CELL
 596f51ba4e94d9000170e1ff_20181223               column=user:dt, timestamp=1545647095916, value=20181223
 596f51ba4e94d9000170e1ff_20181223               column=user:phone, timestamp=1545646705395, value=17086385281
 596f51ba4e94d9000170e1ff_20181224               column=user:dt,

phoenix 4.10 版本后,对列映射做了优化,采用一套新的机制,不在基于列名方式映射到 hbase。如果只做查询,强烈建议使用phoenix 视图方式映射,删除视图不影响 hbase 源数据,语法如下:

0: jdbc:phoenix:dsrv2.heracles.sohuno.com,dme>
use "dbname";
create view "newsinfo_anticheat_user_tag_data"("ROW" varchar primary key, "user"."dt" varchar , "user"."phone" varchar) ;

把HBASE中的ROW当作主键
表名和列族以及列名需要用双引号括起来,因为HBase是区分大小写的,如果不用双引号括起来的话Phoenix在创建表的时候会自动将小写转换为大写字母

Hbase新增列后重新映射

Hbase shell

新添加列user.did_count
put 'dbname:test2', '5a483b8769e9560001f9d1b9_20181224', 'user:did_count', '100'

Phoneix sql

#删除视图
drop view "newsinfo_anticheat_user_tag_data";

#重新创建视图,加上新增列
use "dbname";
create view "newsinfo_anticheat_user_tag_data"("ROW" varchar primary key, "user"."dt" varchar , "user"."phone" varchar, "user"."did_count" varchar) ;

重新查询有了新数据

使用phoenix 表方式映射

创建映射表

必须要表映射,需要禁用列映射规则(会降低查询性能),如下:

use "dbname";
create table "newsinfo_anticheat_user_tag_data"("ROW" varchar primary key, "user"."dt" varchar , "user"."phone" varchar)  column_encoded_bytes=0;

注意:删除映射表时,hbase对应数据也会被删除,慎用删除表操作!!!

Phoneix二级索引

创建二级索引

create index my_index2 on MY_TABLE(V1) include(v2);

HBase增加列

put 'dbname:test2', '5a483b8769e9560001f9d1b9_20181224', 'user:did_count', '100'

Phoneix导出csv文件

[@dudbname103113.heracles.sohuno.com ~]$ /opt/work/phoenix-4.13.1-HBase-1.3/bin/sqlline.py --help
usage: sqlline.py [-h] [-v VERBOSE] [-c COLOR] [-fc FASTCONNECT]
                  [zookeepers] [sqlfile]

Launches the Apache Phoenix Client.

positional arguments:
  zookeepers            The ZooKeeper quorum string
  sqlfile               A file of SQL commands to execute

optional arguments:
  -h, --help            show this help message and exit
  -v VERBOSE, --verbose VERBOSE
                        Verbosity on sqlline.
  -c COLOR, --color COLOR
                        Color setting for sqlline.
  -fc FASTCONNECT, --fastconnect FASTCONNECT
                        Fetch all schemas on initial connection

编辑一个sqlfile,写入待查询sql语句,如下所示:

[@dudbname103113.heracles.sohuno.com ~]$ cat /home/dbname/data/dev_xdf/phoneix/sqlfile.txt
use "dbname";

select * from "newsinfo_anticheat_blacklist_data" limit 5;

执行命令导出数据到csv文件,如下:

/opt/work/phoenix-4.13.1-HBase-1.3/bin/sqlline.py dsrv2.heracles.sohuno.com,dmeta2.heracles.sohuno.com,drm2.heracles.sohuno.com,dmeta1.heracles.sohuno.com /home/dbname/data/dev_xdf/phoneix/sqlfile.txt >> /home/dbname/data/dev_xdf/phoneix/sqlout.csv

查看导出csv文件内容:

[@dudbname103113.heracles.sohuno.com ~]$ cat /home/dbname/data/dev_xdf/phoneix/sqlout.csv
997/997 (100%) Done
+---------------------------+-----+---------------------------------------+-----------+
|          rowkey           | dt  |                  did                  |  did_dt   |
+---------------------------+-----+---------------------------------------+-----------+
| 51bd61df  |     | 76b46d96-a986-36b6-864c-4558d250e6ad  | 20190114  |
| 58a272ec427ad50001849b2b  |     | dd4768f4-5928-37bc-93d9-d700595434a1  | 20190129  |
| 58c246850c0e580001f4104a  |     | 9a79728e-3d0c-3f39-bad4-40a093ec27ea  | 20190122  |
| 596628720c0e5800018c8d2c  |     | f92979768431f3bc6dcf352ac67e5e5d      | 20190129  |
| 5969a1a5c6e6dd0001a27f35  |     | bb60dc61-69ec-3eaa-9ade-772122c8ac88  | 20190129  |
+---------------------------+-----+---------------------------------------+-----------+

Phoneix查询

0: jdbc:phoenix:dsrv2.heracles.sohuno.com,dme> select * from "device_tag_data" where "dt"='20190210' and TO_NUMBER("user_count")>10 LIMIT 2;
+------------------------------------------------+-----------+-------------+----------+------------+--------------+-----------+--------------+-----+
|                      ROW                       |    dt     | user_count  | fenshen  | gyroscope  | android_ver  | hotcloud  | ad_exposure  | ad_ |
+------------------------------------------------+-----------+-------------+----------+------------+--------------+-----------+--------------+-----+
| 03215fc6-5520-3e49-b1e5-f6d72024fa62_20190210  | 20190210  | 11          | 1        |            |              | 1         |              |     |
| 0a340e16-f0dd-3be0-8632-8ac1895d2e6a_20190210  | 20190210  | 12          | 1        |            | 8.1.0        | 1         |              |     |
+------------------------------------------------+-----------+-------------+----------+------------+--------------+-----------+--------------+-----+

写hbase

Spark Bulkload方式

def main(args: Array[String]): Unit = {

  //日期定义
  val today = args(0)
  val hbaseTabName=args(1)
  val hdfsTmpPath = args(2)

  //创建SparkSession
  val sparkconf = new SparkConf().setAppName("UserMetrics")
.set("spark.serializer", "org.apache.spark.serializer.KryoSerializer")
  val sc = new SparkContext(sparkconf)
  val spark = SparkSessionSingleton.getInstance(sc.getConf)

  //配置hbase参数
  val conf = new Configuration()
  conf.set("hbase.zookeeper.quorum", "zklist")
  conf.set("hbase.zookeeper.property.clientPort", "2181")
  conf.set("zookeeper.znode.parent", "/hbase ")
  conf.set(TableOutputFormat.OUTPUT_TABLE, hbaseTabName)

  //获取rdd数据
  val device _rdd = get_device_rdd(spark,today)

  //spark bulkload导用户指标数据入hbase
  spark_bulkload_to_hbase(device_rdd,conf,hdfsTmpPath)

  sc.stop()
}

def get_device_rdd(spark: SparkSession,today: String) :RDD[(ImmutableBytesWritable, Put)] = {
  val sql = ""
  val data = spark.sql(sql)

  val ad_exposure_click_rdd = data.rdd.map(record => {
    val did = record.getString(0)
    val rowkey = did+"_"+today
    val put = new Put(Bytes.toBytes(rowkey))

    try{
      put.addColumn(Bytes.toBytes("device"), Bytes.toBytes("dt"), Bytes.toBytes(today))
      val user_count = record.get(1)
      put.addColumn(Bytes.toBytes("device"), Bytes.toBytes("user_count"), Bytes.toBytes(user_count.toString))
    }catch {
      case e: Exception => println(s"${e.printStackTrace()}")
    }
    (new ImmutableBytesWritable, put)
  })
  ad_exposure_click_rdd
}

/**
  * spark bulkload导dataframe数据入hbase
  */
def spark_direct_bulkload_to_hbase(rdd: RDD[(ImmutableBytesWritable, Put)], conf: Configuration,path: String): Unit ={
  val job = Job.getInstance(conf)
  job.setOutputKeyClass(classOf[ImmutableBytesWritable])
  job.setOutputValueClass(classOf[Result])
  job.setOutputFormatClass(classOf[TableOutputFormat[ImmutableBytesWritable]])

  //save to hbase hfile
  job.getConfiguration.set("mapred.output.dir", path)
  rdd.saveAsNewAPIHadoopDataset(job.getConfiguration)
}

/**
  * spark 先生成hfiles在调用 bulkload导数据入hbase
  */
def spark_bulkload_to_hbase(rdd: RDD[(ImmutableBytesWritable, Put)], conf: Configuration,path: String,hbaseTabName: String): Unit ={
  val myTable = new HTable(conf, hbaseTabName)
  // Save Hfiles on HDFS
  rdd.saveAsNewAPIHadoopFile(path, classOf[ImmutableBytesWritable], classOf[Result], classOf[TableOutputFormat[ImmutableBytesWritable]], conf)

  //Bulk load Hfiles to Hbase
  val bulkLoader = new LoadIncrementalHFiles(conf)
  bulkLoader.doBulkLoad(new Path(path), myTable)
}

Hbase API方式

data.rdd.foreachPartition(
  partitionRecords => {
    val conn = getHBaseConn(hbaseTabName) // 获取Hbase连接
    val tName = TableName.valueOf(hbaseTabName)
    val table = conn.getTable(tName)
    partitionRecords.foreach(record => {

      val userid = record.getString(0)
      val rowkey = userid + "_" + today
      val put = new Put(Bytes.toBytes(rowkey))

      try {
        put.addColumn(Bytes.toBytes("user"), Bytes.toBytes("dt"), Bytes.toBytes(today))

        val ad_exposure = record.get(1)
        if (ad_exposure != None && ad_exposure != null) {
          put.addColumn(Bytes.toBytes("user"), Bytes.toBytes("ad_exposure"), Bytes.toBytes(ad_exposure.toString))
        }
        val ad_click_rate = record.get(2)
        if (ad_click_rate != None && ad_click_rate != null) {
          put.addColumn(Bytes.toBytes("user"), Bytes.toBytes("ad_click_rate"), Bytes.toBytes(ad_click_rate.toString))
        }
        Try(table.put(put)).getOrElse(table.close())//将数据写入HBase,若出错关闭table
      } catch {
        case e: Exception => println(s"================================ ${e.printStackTrace()}")
      }
    })
    table.close()//分区数据写入HBase后关闭连接
    conn.close()
  })

Hbase数据导出

Hbase shell

通过查询条件过滤导出

hbase(main):006:0> scan 'dbname:newsinfo_anticheat_blacklist_data",{COLUMNS => 'user:dt',LIMIT=>1}
ROW                                        COLUMN+CELL               59c925dbcb8e580001ddd21c                  column=user:dt, timestamp=1547805110793, value=20190114                              

导出到文件:

echo " scan 'dbname:newsinfo_anticheat_blacklist_data',{COLUMNS => 'user:dt',LIMIT=>1}" | hbase shell > ./hbase.csv

通过hive导出

有时候我们需要把已存在Hbase中的用户画像数据导到hive里面查询,也就是通过hive就能查到hbase里的数据。但是我又不想使用sqoop或者DataX等工具倒来倒去。这时候可以在hive中创建关联表的方式来查询hbase中的数据。
HBase中建表,然后Hive中建一个外部表,这样当Hive中写入数据后,HBase中也会同时更新。

用hive映射表访问hbase数据

在hbase中创建表后,我们只能在hbase shell中使用scan查询数据,这对于熟悉SQL的使用者不怎么习惯,不过我们可以在hive中创建与hbase表的映射来访问hbase表中的数据,例子如下:

1.这里hbase中的表dbname: device_tag_data已经存在

hbase(main):067:0> scan "dbname:device_tag_data",LIMIT=>1
ROW                                COLUMN+CELL
0000039f-2d6e-3140-bb97-0d7294cfa4fe_2019 column=device:dt, timestamp=1547637050603, value=20190114
0000039f-2d6e-3140-bb97-0d7294cfa4fe_2019 column=device:gyroscope, timestamp=1547631713244, value=103
0000039f-2d6e-3140-bb97-0d7294cfa4fe_2019 column=device:user_count, timestamp=1547631631653, value=1                                     0000039f-2d6e-3140-bb97-0d7294cfa4fe_2019 column=device:android_ver, timestamp=1547638332436, value=6.0

2.创建hive映射表关联hbase

CREATE EXTERNAL TABLE dbname.device_tag_data(
key string,
dt string,
user_count string,
fenshen string,
gyroscope string,
android_ver string,
hotcloud string,
ad_exposure string,
ad_click_rate string
)
STORED BY 'org.apache.hadoop.hive.hbase.HBaseStorageHandler'
WITH SERDEPROPERTIES
("hbase.columns.mapping" =
":key,device:dt,device:user_count,device:fenshen,device:gyroscope,device:android_ver,device:hotcloud,device:ad_exposure,device:ad_click_rate")
TBLPROPERTIES("hbase.table.name" = "dbname:device_tag_data");

主要是配置hbase.table.name和hbase.columns.mapping,一个是hbase表名,一个是hbase字段和hive字段的一一映射,然后就可以从hive中读写hbase数据。
注意hbase.columns.mapping后面的字段直接不能出现空格和换行。

3.通过hive查询数据

0: jdbc:hive2://10.31.103.113:10003/dbname> select * from dbname.device_tag_data where key='0000039f-2d6e-3140-bb97-0d7294cfa4fe_20190114';

注意:
这里我们访问的dbname.device_tag_data表是虚表,数据是存储在hbase中的。Hive 与HBase集成,直接从Hive里面连HBase的数据库进行查询,虽然没有做专门的Benchmark, 但总感觉直接对HBase进行查询操作不怎么靠谱,如果我们要频繁做很多类型的数据分析,那HBase的压力一定会倍增。为此我们可以再建立一个新的hive空表, 把查询出来的数据全部导入到新表当中,以后的所有数据分析操作在新表中完成。

4.创建hive表

CREATE TABLE dbname.device_tag_data2(
key string,
dt string,
user_count string,
fenshen string,
gyroscope string,
android_ver string,
hotcloud string,
ad_exposure string,
ad_click_rate string
)
ROW FORMAT DELIMITED FIELDS TERMINATED BY '\t'
STORED AS TEXTFILE
LOCATION 'hdfs://ns/user/dbname/hive/online/device_tag_data2';

5.将hbase中的表数据加载到本地表

INSERT OVERWRITE TABLE dbname.device_tag_data2 select * from dbname.device_tag_data;

至此大功告成!
以后所有复杂的数据查询和数据分析都可以在新hive表中完成。

03-22 20:36