phoenix 索引实践

准备工作

创建测试表

CREATE TABLE my_table (
        rowkey VARCHAR NOT NULL PRIMARY KEY,
        v1 VARCHAR,
         v2 VARCHAR,
         v3 VARCHAR
     );

    UPSERT INTO my_table values('','value1','value2','value3');
    UPSERT INTO my_table values('','value1','value2','value3');
    UPSERT INTO my_table values('','value1','value2','value3');
    UPSERT INTO my_table values('','value1','value2','value3');
    UPSERT INTO my_table values('','value1','value2','value3');

开启索引支持

HBase --> 配置 --> 高级 --> 搜索 hbase-site.xml。
    在服务端添加下面配置：

    <property>
      <name>hbase.regionserver.wal.codec</name>
      <value>org.apache.hadoop.hbase.regionserver.wal.IndexedWALEditCodec</value>
    </property>

在这里插入图片描述

创建索引

全局索引

全局索引适合读多写少的场景。如果使用全局索引，读数据基本不损耗性能，所有的性能损耗都来源于写数据。数据表的添加、删除和修改都会更新相关的索引表（数据删除了，索引表中的数据也会删除；数据增加了，索引表的数据也会增加）。

注意:

对于全局索引在默认情况下，在查询语句中检索的列如果不在索引表中，Phoenix不会使用索引表将，除非使用hint。

创建全局索引

    CREATE INDEX my_index ON my_table ( v3 );

查看效果

 0: jdbc:phoenix:> select v3 from my_table where v3 = '';
    +--------------+
    |      V3      |
    +--------------+
    | 13000010030  |
    +--------------+
    1 row selected (2.155 seconds)
    0: jdbc:phoenix:> select * from my_table where v3 = '';
    +-------------------+------+--------+--------------+
    |      ROWKEY       |  V1  |   V2   |      V3      |
    +-------------------+------+--------+--------------+
    | 77a9ede22e169683  | aaa| bbb| 13000010030  |
    +-------------------+------+--------+--------------+
    1 row selected (2.337 seconds)
    0: jdbc:phoenix:> CREATE INDEX my_index ON my_table ( v3 );
    1,076,190 rows affected (33.875 seconds)
    0: jdbc:phoenix:> select * from my_table where v3 = '';
    +-------------------+------+--------+--------------+
    |      ROWKEY       |  V1  |   V2   |      V3      |
    +-------------------+------+--------+--------------+
    | 77a9ede22e169683  | aaa| bbb| 13000010030  |
    +-------------------+------+--------+--------------+
    1 row selected (3.296 seconds)
    0: jdbc:phoenix:> select v3 from my_table where v3 = '';
    +--------------+
    |      V3      |
    +--------------+
    | 13000010030  |
    +--------------+
    1 row selected (0.02 seconds)

本地索引

本地索引适合写多读少的场景，或者存储空间有限的场景。和全局索引一样，Phoenix也会在查询的时候自动选择是否使用本地索引。本地索引因为索引数据和原数据存储在同一台机器上，避免网络数据传输的开销，所以更适合写多的场景。由于无法提前确定数据在哪个Region上，所以在读数据的时候，需要检查每个Region上的数据从而带来一些性能损耗。

注意:

对于本地索引，查询中无论是否指定hint或者是查询的列是否都在索引表中，都会使用索引表。

创建本地索引

CREATE LOCAL INDEX LOCAL_IDEX ON my_table(v3);

查看效果

0: jdbc:phoenix:> select * from my_table where v3 = '';
    +-------------------+------+--------+--------------+
    |      ROWKEY       |  V1  |   V2   |      V3      |
    +-------------------+------+--------+--------------+
    | 77a9ede22e169683  | aaa| bbb| 13000010030  |
    +-------------------+------+--------+--------------+
    1 row selected (3.545 seconds)
    0: jdbc:phoenix:> select v3 from my_table where v3 = '';
    +--------------+
    |      V3      |
    +--------------+
    | 13000010030  |
    +--------------+
    1 row selected (2.946 seconds)
    0: jdbc:phoenix:> CREATE LOCAL INDEX LOCAL_IDEX ON my_table(v3);
    1,076,190 rows affected (24.67 seconds)
    0: jdbc:phoenix:> select * from my_table where v3 = '';
    +-------------------+------+--------+--------------+
    |      ROWKEY       |  V1  |   V2   |      V3      |
    +-------------------+------+--------+--------------+
    | 77a9ede22e169683  | aaa| bbb| 13000010030  |
    +-------------------+------+--------+--------------+
    1 row selected (0.055 seconds)
    0: jdbc:phoenix:> select v3 from my_table where v3 = '';
    +--------------+
    |      V3      |
    +--------------+
    | 13000010030  |
    +--------------+
    1 row selected (0.013 seconds)

覆盖索引

覆盖索引是把原数据存储在索引数据表中，这样在查询时不需要再去HBase的原表获取数据就，直接返回查询结果。

注意：

查询是 select 的列和 where 的列都需要在索引中出现。

创建覆盖索引

CREATE INDEX my_index ON my_table ( v2,v3 ) INCLUDE ( v1 );

添加索引后提升到毫秒级

0: jdbc:phoenix:> select * from my_table where v3 = '' and v2 = '北京顺义';
    +-------------------+-----+-------+--------------+
    |      ROWKEY       | V1  |  V2   |      V3      |
    +-------------------+-----+-------+--------------+
    | 3f65283ed7553909  | wenyuan  | ccc| 13308117837  |
    +-------------------+-----+-------+--------------+
    1 row selected (2.42 seconds)
    0: jdbc:phoenix:> CREATE INDEX my_index ON my_table (v2,v3) INCLUDE ( v1 );
    1,076,190 rows affected (47.432 seconds)
    0: jdbc:phoenix:> select * from my_table where v3 = '' and v2 = '北京顺义';
    +-------------------+-----+-------+--------------+
    |      ROWKEY       | V1  |  V2   |      V3      |
    +-------------------+-----+-------+--------------+
    | 3f65283ed7553909  | wenyuan| ccc| 13308117837  |
    +-------------------+-----+-------+--------------+
    1 row selected (0.031 seconds)

函数索引

从Phoenix4.3版本就有函数索引，特点是索引的内容不局限于列，能根据表达式创建索引。适用于对查询表时过滤条件是表达式。如果你使用的表达式正好就是索引的话，数据也可以直接从这个索引获取，而不需要从数据库获取。

创建索引

CREATE INDEX my_index ON my_table(substr(v3,1,9)) INCLUDE ( v1 );

查看效果

0: jdbc:phoenix:> select v1,substr(v3,1,9) from my_table where substr(v3,1,9) = '';
    +-----+-------------------+
    | V1  | SUBSTR(V3, 1, 9)  |
    +-----+-------------------+
    | wenyuan| 130000109         |
    +-----+-------------------+
    1 row selected (3.656 seconds)
    0: jdbc:phoenix:> select v1,v3 from my_table where substr(v3,1,9) = '';
    +-----+--------------+
    | V1  |      V3      |
    +-----+--------------+
    | wenyuan| 13000010979  |
    +-----+--------------+
    1 row selected (3.969 seconds)
    0: jdbc:phoenix:> CREATE INDEX my_index ON my_table(substr(v3,1,9)) INCLUDE ( v1 );
    1,076,190 rows affected (45.833 seconds)

    0: jdbc:phoenix:> select v1,v3 from my_table where substr(v3,1,9) = '';
    +-----+--------------+
    | V1  |      V3      |
    +-----+--------------+
    | wenyuan| 13000010979  |
    +-----+--------------+
    1 row selected (3.44 seconds)
    0: jdbc:phoenix:> select v1,v3,substr(v3,1,9) from my_table where substr(v3,1,9) = '';
    +-----+--------------+-------------------+
    | V1  |      V3      | SUBSTR(V3, 1, 9)  |
    +-----+--------------+-------------------+
    | wenyuan| 13000010979  | 130000109         |
    +-----+--------------+-------------------+
    1 row selected (3.327 seconds)
    0: jdbc:phoenix:> select v1,substr(v3,1,9) from my_table where substr(v3,1,9) = '';
    +-----+--------------------+
    | V1  | " SUBSTR(V3,1,9)"  |
    +-----+--------------------+
    | wenyuan  | 130000109          |
    +-----+--------------------+
    1 row selected (0.013 seconds)
    0: jdbc:phoenix:> select v1 from my_table where substr(v3,1,9) = '';
    +-----+
    | V1  |
    +-----+
    | wenyuan|
    +-----+
    1 row selected (0.011 seconds)

索引Building

同步索引

    CREATE INDEX ASYNC_IDX ON SCHEMA_NAME.TABLE_NAME(BASICINFO."s1",BASICINFO."s2") ;

创建同步索引超时怎么办？

在客户端配置文件hbase-site.xml中，把超时参数设置大一些，足够 Build 索引数据的时间。

 <property>
        <name>hbase.rpc.timeout</name>
        <value>60000000</value>
    </property>
    <property>
        <name>hbase.client.scanner.timeout.period</name>
        <value>60000000</value>
    </property>
    <property>
        <name>phoenix.query.timeoutMs</name>
        <value>60000000</value>
    </property>

异步索引

异步Build索引需要借助MapReduce，创建异步索引语法和同步索引相差一个关键字：ASYNC。

创建异步索引

        CREATE INDEX ASYNC_IDX ON SCHEMA_NAME.TABLE_NAME ( BASICINFO."s1", BASICINFO."s2" ) ASYNC;
    

运行MapReduce

执行MapReduce

        hbase org.apache.phoenix.mapreduce.index.IndexTool \
        --schema SCHEMA_NAME\
        --data-table TABLE_NAME\
        --index-table ASYNC_IDX \
        --output-path ASYNC_IDX_HFILES

日志：

Java HotSpot(TM) 64-Bit Server VM warning: Using incremental CMS is deprecated and will likely be removed in a future release
        SLF4J: Class path contains multiple SLF4J bindings.
        SLF4J: Found binding in [jar:file:/opt/cloudera/parcels/CDH-5.12.1-1.cdh5.12.1.p0.3/jars/phoenix-4.14.0-cdh5.12.2-client.jar!/org/slf4j/impl/StaticLoggerBinder.class]
        SLF4J: Found binding in [jar:file:/opt/cloudera/parcels/CDH-5.12.1-1.cdh5.12.1.p0.3/jars/slf4j-log4j12-1.7.5.jar!/org/slf4j/impl/StaticLoggerBinder.class]
        SLF4J: See http://www.slf4j.org/codes.html#multiple_bindings for an explanation.
        SLF4J: Actual binding is of type [org.slf4j.impl.Log4jLoggerFactory]
        19/05/22 15:38:41 INFO log.QueryLoggerDisruptor: Starting  QueryLoggerDisruptor for with ringbufferSize=8192, waitStrategy=BlockingWaitStrategy, exceptionHandler=org.apache.phoenix.log.QueryLoggerDefaultExceptionHandler@dd0c991...
        19/05/22 15:38:41 INFO query.ConnectionQueryServicesImpl: An instance of ConnectionQueryServices was created.

        ...

        19/05/22 15:41:19 INFO index.IndexTool: Loading HFiles from INDEX_PERSONAS_TAG_HFILES/MY_SCHEMA.INDEX_PERSONAS_TAG
        19/05/22 15:41:19 WARN mapreduce.LoadIncrementalHFiles: Skipping non-directory hdfs://bigdata-dev-41:8020/user/root/INDEX_PERSONAS_TAG_HFILES/MY_SCHEMA.INDEX_PERSONAS_TAG/_SUCCESS
        19/05/22 15:41:19 INFO hfile.CacheConfig: CacheConfig:disabled
        19/05/22 15:41:19 INFO mapreduce.LoadIncrementalHFiles: Trying to load hfile=hdfs://bigdata-dev-41:8020/user/root/INDEX_PERSONAS_TAG_HFILES/MY_SCHEMA.INDEX_PERSONAS_TAG/0/e1f766365b4f4c7cb6cfc6e0d18328b8 first=0\x0010\x00\xE4\xB8\x9A\xE4\xB8\xBB\x000\x000\x0010\x000\x00\xE6\xAD\xA3\xE5\xB8\xB8\xE4\xB8\x9A\xE4\xB8\xBB\x001001.99\x000\x001\x003\x00\xE8\x80\x81\xE5\xAE\xA2\xE6\x88\xB7\x00\xE6\x9C\xAA\xE7\x9F\xA5\x0042471415705946377 last=2\x009\x00\xE7\xA7\x9F\xE5\xAE\xA2\x002\x002\x009\x002\x00\xE9\x95\xBF\xE6\x9C\x9F\xE4\xB8\x8D\xE4\xBA\xA4\xE7\x89\xA9\xE4\xB8\x9A\xE7\xAE\xA1\xE7\x90\x86\xE8\xB4\xB9\x00988.56\x000\x001\x004\x00\xE6\x9C\xAA\xE7\x9F\xA5\x00\xE5\x9C\x9F\xE8\xB1\xAA\x0044ff3613003558171
        19/05/22 15:41:20 INFO index.IndexToolUtil:  Updated the status of the index INDEX_PERSONAS_TAG to ACTIVE

遇到问题

Error: Could not find or load main class org.apache.phoenix.mapreduce.index.IndexTool

解决办法

将 phoenix-4.14.0-cdh5.12.2-client.jar 包复制到 hbase 的 lib 目录下

[root@node00 ~]# cd /opt/cloudera/parcels/
        [root@node00 parcels]# cd APACHE_PHOENIX/lib/phoenix
        [root@node00 phoenix]# cp phoenix-4.14.0-cdh5.12.2-client.jar /opt/cloudera/parcels/CDH/jars/
        [root@node00 phoenix]# cd /opt/cloudera/parcels/CDH/lib/hbase/lib/
        [root@node00 lib]# ln -s ../../../jars/phoenix-4.14.0-cdh5.12.2-client.jar phoenix-4.14.0-cdh5.12.2-client.jar

索引用法总结

Phoenix 的二级索引主要有两种，即全局索引和本地索引。
全局索引适合读多写少的场景，如果使用全局索引，读数据基本不损耗性能，所有的性能损耗都来源于写数据。
本地索引适合写多读少的场景，或者存储空间有限的场景。

索引定义完之后，一般来说，Phoenix自己会判定使用哪个索引更加有效。
但是，全局索引必须是查询语句中所有列都包含在全局索引中，它才会生效。

索引为：

create index my_index on my_table (v3);

select v1 from my_table where v3 = '';

上面语句怎样才能使用索引呢？

有以下三种方法使它使用索引：

使用覆盖索引

    CREATE INDEX cover_index ON my_table(v3) INCLUDE (v1);

查看效果

    0: jdbc:phoenix:> select v1 from my_table where v3 = '';
    +------+
    |  V1  |
    +------+
    | wenyuan|
    +------+
    1 row selected (0.01 seconds)

使用 Hint 强制索引

SELECT /*+ INDEX(my_table my_index) */ v1 FROM my_table WHERE v3 = '';

查看效果

    0: jdbc:phoenix:> SELECT /*+ INDEX(my_table my_index) */ v1 FROM my_table WHERE v3 = '';
    +------+
    |  V1  |
    +------+
    | wenyuan|
    +------+
    1 row selected (0.044 seconds)

使用本地索引

    CREATE LOCAL INDEX local_index on my_table (v3);

查看效果

0: jdbc:phoenix:> select v1 from my_table where v3 = '';
    +------+
    |  V1  |
    +------+
    | wenyuan|
    +------+
    1 row selected (0.025 seconds)