CDH安装Phoenix

1.文档编写目的

---

昨天我们简单介绍了一下Apache Phoenix，请参考Cloudera Labs中的Phoenix。今天我们主要讲述如何在CDH中安装配置Phoenix，并会做一些使用示例。

• 内容概述

1.安装及配置Phoenix

2.Phoenix的基本操作

3.使用Phoenix bulkload数据到HBase

4.使用Phoenix从HBase中导出数据到HDFS

• 测试环境

1.CDH5.11.2

2.RedHat7.2

3.Phoenix4.7.0

• 前置条件

1.CDH集群正常

2.HBase服务已经安装并正常运行

3.测试csv数据已准备

4.Redhat7中的httpd服务已安装并使用正常

2.在CDH集群中安装Phoenix

---

1.到Cloudera官网下载Phoenix的Parcel，注意选择与操作系统匹配的版本，因为本次测试使用的是Redhat7，所以选择后缀名为el7的文件。下载地址为：

http://archive.cloudera.com/cloudera-labs/phoenix/parcels/latest/

具体需要下载的三个文件地址为：

http://archive.cloudera.com/cloudera-labs/phoenix/parcels/latest/CLABS_PHOENIX-4.7.0-1.clabs_phoenix1.3.0.p0.000-el7.parcel
http://archive.cloudera.com/cloudera-labs/phoenix/parcels/latest/CLABS_PHOENIX-4.7.0-1.clabs_phoenix1.3.0.p0.000-el7.parcel.sha1
http://archive.cloudera.com/cloudera-labs/phoenix/parcels/latest/manifest.json

2.将下载好的文件发布到httpd服务，可以用浏览器打开页面进行测试。

[ec2-user@ip-172-31-22-86 phoenix]$ pwd
/var/www/html/phoenix
[ec2-user@ip-172-31-22-86 phoenix]$ ll
total 192852
-rw-r--r-- 1 root root        41 Jun 24  2016 CLABS_PHOENIX-4.7.0-1.clabs_phoenix1.3.0.p0.000-el7.parcel.sha1
-rw-r--r-- 1 root root 197466534 Jun 24  2016 CLABS_PHOENIX-4.7.0-1.clabs_phoenix1.3.0.p0.000-el7.parcel
-rw-r--r-- 1 root root      4687 Jun 24  2016 manifest.json
[ec2-user@ip-172-31-22-86 phoenix]$ 

3.从Cloudera Manager点击“Parcel”进入Parcel管理页面

点击“配置”，输入Phoenix的Parcel包http地址。

点击“保存更改“回到Parcel管理页面，发现CM已发现Phoenix的Parcel。

点击“下载”->“分配”->“激活”

4.回到CM主页，发现HBase服务需要部署客户端配置以及重启

重启HBase服务

安装完成。

3.如何在CDH集群中使用Phoenix

3.1Phoenix的基本操作

---

进入Phoenix的脚本命令目录

[ec2-user@ip-172-31-22-86 bin]$ cd /opt/cloudera/parcels/CLABS_PHOENIX/bin
[ec2-user@ip-172-31-22-86 bin]$ ll
total 16
-rwxr-xr-x 1 root root 672 Jun 24  2016 phoenix-performance.py
-rwxr-xr-x 1 root root 665 Jun 24  2016 phoenix-psql.py
-rwxr-xr-x 1 root root 668 Jun 24  2016 phoenix-sqlline.py
-rwxr-xr-x 1 root root 674 Jun 24  2016 phoenix-utils.py

使用Phoenix登录HBase

[ec2-user@ip-172-31-22-86 bin]$ ./phoenix-sqlline.py
Zookeeper not specified. 
Usage: sqlline.py   
Example: 
 1. sqlline.py localhost:2181:/hbase 
 2. sqlline.py localhost:2181:/hbase ../examples/stock_symbol.sql

需要指定Zookeeper

[ec2-user@ip-172-31-22-86 bin]$ ./phoenix-sqlline.py ip-172-31-21-45:2181:/hbase
...
sqlline version 1.1.8
0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> !tables
+------------+--------------+-------------+---------------+----------+------------+--------------------+
| TABLE_CAT  | TABLE_SCHEM  | TABLE_NAME  |  TABLE_TYPE   | REMARKS  | TYPE_NAME  | SELF_REFERENCING_C |
+------------+--------------+-------------+---------------+----------+------------+--------------------+
|            | SYSTEM       | CATALOG     | SYSTEM TABLE  |          |            |                    |
|            | SYSTEM       | FUNCTION    | SYSTEM TABLE  |          |            |                    |
|            | SYSTEM       | SEQUENCE    | SYSTEM TABLE  |          |            |                    |
|            | SYSTEM       | STATS       | SYSTEM TABLE  |          |            |                    |
|            |              | ITEM        | TABLE         |          |            |                    |
+------------+--------------+-------------+---------------+----------+------------+--------------------+
0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> 

创建一张测试表

注意：建表必须指定主键。

0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> create table hbase_test
. . . . . . . . . . . . . . . . . . . . . .> (
. . . . . . . . . . . . . . . . . . . . . .>     s1 varchar not null primary key,
. . . . . . . . . . . . . . . . . . . . . .>     s2 varchar,
. . . . . . . . . . . . . . . . . . . . . .>     s3 varchar,
. . . . . . . . . . . . . . . . . . . . . .>     s4 varchar
. . . . . . . . . . . . . . . . . . . . . .> );
No rows affected (1.504 seconds)

在hbase shell中进行检查

插入一行数据。注意：Phoenix中没有insert语法，用upsert代替。参考：http://phoenix.apache.org/language/index.html

0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> upsert into hbase_test values('1','testname','testname1','testname2');
1 row affected (0.088 seconds)
0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> select * from hbase_test;
+-----+-----------+------------+------------+
| S1  |    S2     |     S3     |     S4     |
+-----+-----------+------------+------------+
| 1   | testname  | testname1  | testname2  |
+-----+-----------+------------+------------+
1 row selected (0.049 seconds)
0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> 

在hbase shell中进行检查

删除这行数据，delete测试

0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> delete from hbase_test where s1='1';
1 row affected (0.018 seconds)
0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> select * from hbase_test;
+-----+-----+-----+-----+
| S1  | S2  | S3  | S4  |
+-----+-----+-----+-----+
+-----+-----+-----+-----+
No rows selected (0.045 seconds)
0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> 

在hbase shell中进行检查

更新数据测试，注意Phoenix中没有update语法，用upsert代替。插入多条数据需要执行多条upsert语句，没办法将所有的数据都写到一个“values”后面。

0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> upsert into hbase_test values('1','testname','testname1','testname2');
1 row affected (0.017 seconds)
0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> upsert into hbase_test values('2','testname','testname1','testname2');
1 row affected (0.007 seconds)
0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> upsert into hbase_test values('3','testname','testname1','testname2');
1 row affected (0.008 seconds)
0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> select * from hbase_test;
+-----+-----------+------------+------------+
| S1  |    S2     |     S3     |     S4     |
+-----+-----------+------------+------------+
| 1   | testname  | testname1  | testname2  |
| 2   | testname  | testname1  | testname2  |
| 3   | testname  | testname1  | testname2  |
+-----+-----------+------------+------------+
3 rows selected (0.067 seconds)
0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> upsert into hbase_test values('1','fayson','testname1','testname2');
1 row affected (0.009 seconds)
0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> select * from hbase_test;
+-----+-----------+------------+------------+
| S1  |    S2     |     S3     |     S4     |
+-----+-----------+------------+------------+
| 1   | fayson    | testname1  | testname2  |
| 2   | testname  | testname1  | testname2  |
| 3   | testname  | testname1  | testname2  |
+-----+-----------+------------+------------+
3 rows selected (0.037 seconds)
0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase>

在hbase shell中进行检查

批量更新测试，创建另外一张表hbase_test1，表结构与hbase_test一样，并插入五条，有两条是hbase_test中没有的（主键为4，5），有一条与hbase_test中的数据不一样（主键为1），有两条是完全一样（主键为2，3）。

0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> create table hbase_test1
. . . . . . . . . . . . . . . . . . . . . .> (
. . . . . . . . . . . . . . . . . . . . . .>     s1 varchar not null primary key,
. . . . . . . . . . . . . . . . . . . . . .>     s2 varchar,
. . . . . . . . . . . . . . . . . . . . . .>     s3 varchar,
. . . . . . . . . . . . . . . . . . . . . .>     s4 varchar
. . . . . . . . . . . . . . . . . . . . . .> );
No rows affected (1.268 seconds)
0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> 
0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> upsert into hbase_test1 values('1','fayson','testname1','testname2');
1 row affected (0.031 seconds)
0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> upsert into hbase_test1 values('2','testname','testname1','testname2');
1 row affected (0.006 seconds)
0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> upsert into hbase_test1 values('3','testname','testname1','testname2');
1 row affected (0.005 seconds)
0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> upsert into hbase_test1 values('4','testname','testname1','testname2');
1 row affected (0.005 seconds)
0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> upsert into hbase_test1 values('5','testname','testname1','testname2');
1 row affected (0.007 seconds)
0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> select * from hbase_test1;
+-----+-----------+------------+------------+
| S1  |    S2     |     S3     |     S4     |
+-----+-----------+------------+------------+
| 1   | fayson    | testname1  | testname2  |
| 2   | testname  | testname1  | testname2  |
| 3   | testname  | testname1  | testname2  |
| 4   | testname  | testname1  | testname2  |
| 5   | testname  | testname1  | testname2  |
+-----+-----------+------------+------------+
5 rows selected (0.038 seconds)

批量更新，我们用hbase_test1中的数据去更新hbase_test。

0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> upsert into hbase_test select * from hbase_test1;
5 rows affected (0.03 seconds)
0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> select * from hbase_test;
+-----+-----------+------------+------------+
| S1  |    S2     |     S3     |     S4     |
+-----+-----------+------------+------------+
| 1   | fayson    | testname1  | testname2  |
| 2   | testname  | testname1  | testname2  |
| 3   | testname  | testname1  | testname2  |
| 4   | testname  | testname1  | testname2  |
| 5   | testname  | testname1  | testname2  |
+-----+-----------+------------+------------+
5 rows selected (0.039 seconds)
0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase>

批量更新发现对于已有的数据，如果值不一样，会覆盖，对于相同的数据会保持不变，对于没有的数据会直接作为新的数据插入。

3.2使用Phoenix bulkload数据到HBase

---

准备需要批量导入的测试数据，这里使用TPC_DS的item表数据。

[ec2-user@ip-172-31-22-86 ~]$ ll item.dat
-rw-r--r-- 1 root root 28855325 Oct  3 10:23 item.dat
[ec2-user@ip-172-31-22-86 ~]$ head -1 item.dat
1|AAAAAAAABAAAAAAA|1997-10-27||Powers will not get influences. Electoral ports should show low, annual chains. Now young visitors may pose now however final pages. Bitterly right children suit increasing, leading el|27.02|23.23|5003002|exportischolar #2|3|pop|5|Music|52|ableanti|N/A|3663peru009490160959|spring|Tsp|Unknown|6|ought|

因为Phoenix的bulkload只能导入csv，所以我们先把该数据的分隔符修改为逗号，并且后缀名改为.csv

[ec2-user@ip-172-31-22-86 ~]$ sed -i 's/|/,/g' item.dat
[ec2-user@ip-172-31-22-86 ~]$ mv item.dat item.csv
[ec2-user@ip-172-31-22-86 ~]$ ll item.csv 
-rw-r--r-- 1 ec2-user ec2-user 28855325 Oct  3 10:26 item.csv
[ec2-user@ip-172-31-22-86 ~]$ head -1 item.csv 
1,AAAAAAAABAAAAAAA,1997-10-27,,Powers will not get influences. Electoral ports should show low, annual chains. Now young visitors may pose now however final pages. Bitterly right children suit increasing, leading el,27.02,23.23,5003002,exportischolar #2,3,pop,5,Music,52,ableanti,N/A,3663peru009490160959,spring,Tsp,Unknown,6,ought,

上传该文件到HDFS

[ec2-user@ip-172-31-22-86 ~]$ hadoop fs -mkdir /fayson
[ec2-user@ip-172-31-22-86 ~]$ hadoop fs -put item.csv /fayson
[ec2-user@ip-172-31-22-86 ~]$ hadoop fs -ls /fayson
Found 1 items
-rw-r--r--   3 ec2-user supergroup   28855325 2017-10-03 10:28 /fayson/item.csv
[ec2-user@ip-172-31-22-86 ~]$ 

通过Phoenix创建item表，注意为了方便阅读，只创建了4个字段

0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> create table item
. . . . . . . . . . . . . . . . . . . . . .> (
. . . . . . . . . . . . . . . . . . . . . .>     i_item_sk varchar not null primary key,
. . . . . . . . . . . . . . . . . . . . . .>     i_item_id varchar,
. . . . . . . . . . . . . . . . . . . . . .>     i_rec_start_varchar varchar,
. . . . . . . . . . . . . . . . . . . . . .>     i_rec_end_date varchar
. . . . . . . . . . . . . . . . . . . . . .> );
No rows affected (1.268 seconds)
0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase>

执行bulkload命令导入数据

[ec2-user@ip-172-31-22-86 ~]$ HADOOP_CLASSPATH=/opt/cloudera/parcels/CDH/lib/hbase/hbase-protocol-1.2.0-cdh5.12.1.jar:/opt/cloudera/parcels/CDH/lib/hbase/conf hadoop jar /opt/cloudera/parcels/CLABS_PHOENIX/lib/phoenix/phoenix-4.7.0-clabs-phoenix1.3.0-client.jar org.apache.phoenix.mapreduce.CsvBulkLoadTool -t item -i /fayson/item.csv
17/10/03 10:32:24 INFO util.QueryUtil: Creating connection with the jdbc url: jdbc:phoenix:ip-172-31-21-45.ap-southeast-1.compute.internal,ip-172-31-22-86.ap-southeast-1.compute.internal,ip-172-31-26-102.ap-southeast-1.compute.internal:2181:/hbase;
...
17/10/03 10:32:24 INFO zookeeper.ZooKeeper: Initiating client connection, connectString=ip-172-31-21-45.ap-southeast-1.compute.internal:2181,ip-172-31-22-86.ap-southeast-1.compute.internal:2181,ip-172-31-26-102.ap-southeast-1.compute.internal:2181 sessionTimeout=60000 watcher=hconnection-0x7a9c0c6b0x0, quorum=ip-172-31-21-45.ap-southeast-1.compute.internal:2181,ip-172-31-22-86.ap-southeast-1.compute.internal:2181,ip-172-31-26-102.ap-southeast-1.compute.internal:2181, baseZNode=/hbase
17/10/03 10:32:24 INFO zookeeper.ClientCnxn: Opening socket connection to server ip-172-31-21-45.ap-southeast-1.compute.internal/172.31.21.45:2181. Will not attempt to authenticate using SASL (unknown error)
...
17/10/03 10:32:30 INFO mapreduce.Job: Running job: job_1507035313248_0001
17/10/03 10:32:38 INFO mapreduce.Job: Job job_1507035313248_0001 running in uber mode : false
17/10/03 10:32:38 INFO mapreduce.Job:  map 0% reduce 0%
17/10/03 10:32:52 INFO mapreduce.Job:  map 100% reduce 0%
17/10/03 10:33:01 INFO mapreduce.Job:  map 100% reduce 100%
17/10/03 10:33:01 INFO mapreduce.Job: Job job_1507035313248_0001 completed successfully
17/10/03 10:33:01 INFO mapreduce.Job: Counters: 50
...
17/10/03 10:33:01 INFO mapreduce.AbstractBulkLoadTool: Loading HFiles from /tmp/fef0045b-8a31-4d95-985a-bee08edf2cf9
...

在Phoenix中查询该表

0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> select * from item limit 10;
+------------+-------------------+----------------------+-----------------+
| I_ITEM_SK  |     I_ITEM_ID     | I_REC_START_VARCHAR  | I_REC_END_DATE  |
+------------+-------------------+----------------------+-----------------+
| 1          | AAAAAAAABAAAAAAA  | 1997-10-27           |                 |
| 10         | AAAAAAAAKAAAAAAA  | 1997-10-27           | 1999-10-27      |
| 100        | AAAAAAAAEGAAAAAA  | 1997-10-27           | 1999-10-27      |
| 1000       | AAAAAAAAIODAAAAA  | 1997-10-27           | 1999-10-27      |
| 10000      | AAAAAAAAABHCAAAA  | 1997-10-27           | 1999-10-27      |
| 100000     | AAAAAAAAAKGIBAAA  | 1997-10-27           | 1999-10-27      |
| 100001     | AAAAAAAAAKGIBAAA  | 1999-10-28           | 2001-10-26      |
| 100002     | AAAAAAAAAKGIBAAA  | 2001-10-27           |                 |
| 100003     | AAAAAAAADKGIBAAA  | 1997-10-27           |                 |
| 100004     | AAAAAAAAEKGIBAAA  | 1997-10-27           | 2000-10-26      |
+------------+-------------------+----------------------+-----------------+
10 rows selected (0.054 seconds)
0: jdbc:phoenix:ip-172-31-21-45:2181:/hbase> 

在hbase shell中查询该表

hbase(main):002:0> scan 'ITEM', LIMIT => 10
ROW                         COLUMN+CELL                                                                 
 1                          column=0:I_ITEM_ID, timestamp=1507041176470, value=AAAAAAAABAAAAAAA         
 1                          column=0:I_REC_START_VARCHAR, timestamp=1507041176470, value=1997-10-27     
 1                          column=0:_0, timestamp=1507041176470, value=                                
 10                         column=0:I_ITEM_ID, timestamp=1507041176470, value=AAAAAAAAKAAAAAAA         
 10                         column=0:I_REC_END_DATE, timestamp=1507041176470, value=1999-10-27          
 10                         column=0:I_REC_START_VARCHAR, timestamp=1507041176470, value=1997-10-27     
 10                         column=0:_0, timestamp=1507041176470, value=                                
...
 100004                     column=0:I_REC_START_VARCHAR, timestamp=1507041176470, value=1997-10-27     
 100004                     column=0:_0, timestamp=1507041176470, value=                                
10 row(s) in 0.2360 seconds

入库条数检查

条数相等，全部入库成功。

3.3使用Phoenix从HBase中导出数据到HDFS

---

Phoenix还提供了使用MapReduce导出数据到HDFS的功能，以pig的脚本执行。首先准备pig脚本。

[ec2-user@ip-172-31-22-86 ~]$ cat export.pig 
REGISTER /opt/cloudera/parcels/CLABS_PHOENIX/lib/phoenix/phoenix-4.7.0-clabs-phoenix1.3.0-client.jar;
rows = load 'hbase://query/SELECT * FROM ITEM' USING org.apache.phoenix.pig.PhoenixHBaseLoader('ip-172-31-21-45:2181');
STORE rows INTO 'fayson1' USING PigStorage(',');

[ec2-user@ip-172-31-22-86 ~]$

执行该脚本

[ec2-user@ip-172-31-22-86 ~]$ pig -x mapreduce export.pig 
...
Counters:
Total records written : 102000
Total bytes written : 4068465
Spillable Memory Manager spill count : 0
Total bags proactively spilled: 0
Total records proactively spilled: 0

Job DAG:
job_1507035313248_0002

2017-10-03 10:45:38,905 [main] INFO  org.apache.pig.backend.hadoop.executionengine.mapReduceLayer.MapReduceLauncher - Success!

导出成功后检查HDFS中的数据

[ec2-user@ip-172-31-22-86 ~]$ hadoop fs -ls /user/ec2-user/fayson1
Found 2 items
-rw-r--r--   3 ec2-user supergroup          0 2017-10-03 10:45 /user/ec2-user/fayson1/_SUCCESS
-rw-r--r--   3 ec2-user supergroup    4068465 2017-10-03 10:45 /user/ec2-user/fayson1/part-m-00000
[ec2-user@ip-172-31-22-86 ~]$ hadoop fs -cat /user/ec2-user/fayson1/part-m-00000 | head -2
1,AAAAAAAABAAAAAAA,1997-10-27,
10,AAAAAAAAKAAAAAAA,1997-10-27,1999-10-27
cat: Unable to write to output stream.
[ec2-user@ip-172-31-22-86 ~]$ 

检查条数为10200与原始数据一致，全部导出成功。

4.总结

---

• 使用Cloudera提供的Phoenix Parcel，可以很方便的安装Phoenix。
• 使用Phoenix可以对HBase进行建表，删除，更新等操作，都是以大家熟悉的SQL方式操作。
• Phoenix提供了批量导入/导出数据的方式。批量导入只支持csv格式，分隔符为逗号。
• Phoenix中的SQL操作，可以马上同步到HBase，通过hbase shell检查都成功。
• 目前Cloudera官方提供的Phoenix版本较旧，为4.7.0，社区最新版本为4.11.0
• Phoenix提供的SQL语法较为简陋，没有insert/update，一律用upsert代替。
• 使用upsert插入数据时，只能一条一条插入，没法将全部字段值写到一个“values”后面。