Phoenix4.8整合Spark

本文使用Spark shel操作phoenix（参考官方文档http://phoenix.apache.org/phoenix_spark.html，解决了官方文档的一些问题）

通过phoenix提供的包来操作hbase比直接使用jdbc的方式更高效（因为比jdbc的方式更细粒度化的调度）

软件环境：

Phoenix4.8

HBase1.1.2

Spark1.5.2/1.6.3/2.x

1、在phoenix中建表

CREATE TABLE TABLE1 (ID BIGINT NOT NULL PRIMARY KEY, COL1 VARCHAR);
UPSERT INTO TABLE1 (ID, COL1) VALUES (1, 'test_row_1');
UPSERT INTO TABLE1 (ID, COL1) VALUES (2, 'test_row_2');

1、启动spark-shelll

spark-shell  --jars /opt/phoenix4.8/phoenix-spark-4.8.0-HBase-1.1.jar,/opt/phoenix4.8/phoenix-4.8.0-HBase-1.1-client.jar

2、使用DataSource API，load为DataFrame

import org.apache.phoenix.spark._

val df = sqlContext.load("org.apache.phoenix.spark", Map("table" -> "TABLE1", "zkUrl" -> "192.38.0.231:2181"))
df.filter(df("COL1") === "test_row_1" && df("ID") === 1L).select(df("ID")).show

如果spark2.x版本，使用如下方式（前提是修改了phoenix-spark模块的代码，使之兼容spark2.x，重新编译）：

import org.apache.phoenix.spark._

val df = spark.read.format("org.apache.phoenix.spark").options(Map("table" -> "TABLE1", "zkUrl" -> "192.38.0.231:2181")).load
df.filter(df("COL1") === "test_row_1" && df("ID") === 1L).select(df("ID")).show

使用spark-sql方式如下

spark-sql --jars /opt/phoenix4.8/phoenix-spark-4.8.0-HBase-1.1.jar,/opt/phoenix4.8/phoenix-4.8.0-HBase-1.1-client.jar

CREATE TABLE spark_ph
USING org.apache.phoenix.spark
OPTIONS (
  table "TABLE1",
  zkUrl "192.38.0.231:2181"
);

3、使用Configuration类，load为DataFrame

import org.apache.hadoop.conf.Configuration
import org.apache.phoenix.spark._

val configuration = new Configuration()

val df = sqlContext.phoenixTableAsDataFrame("TABLE1", Array("ID", "COL1"), conf = configuration)

df.show

4、使用Zookeeper URL ，load为RDD

import org.apache.phoenix.spark._

//将ID和COL1两列，加载为一个RDD
val rdd = sc.phoenixTableAsRDD("TABLE1", Seq("ID", "COL1"), zkUrl = Some("192.38.0.231:2181"))

rdd.count()

val firstId = rdd.first()("ID").asInstanceOf[Long]
val firstCol = rdd.first()("COL1").asInstanceOf[String]

5、通过Spark向Phoenix中写入数据（RDD方式）

phoenix中创建表output_test_table：
CREATE TABLE OUTPUT_TEST_TABLE (id BIGINT NOT NULL PRIMARY KEY, col1 VARCHAR, col2 INTEGER);

import org.apache.phoenix.spark._

val sc = new SparkContext("local", "phoenix-test")
val dataSet = List((1L, "1", 1), (2L, "2", 2), (3L, "3", 3))

sc.parallelize(dataSet).saveToPhoenix("OUTPUT_TEST_TABLE", Seq("ID","COL1","COL2"), zkUrl = Some("192.38.0.231:2181"))

0: jdbc:phoenix:localhost> select * from output_test_table;
+-----+-------+-------+
| ID  | COL1  | COL2  |
+-----+-------+-------+
| 1   | 1     | 1     |
| 2   | 2     | 2     |
| 3   | 3     | 3     |
+-----+-------+-------+
3 rows selected (0.168 seconds)

6、通过Spark向Phoenix中写入数据（DataFrame方式）

phoenix中创建两张表：

CREATE TABLE INPUT_TABLE (id BIGINT NOT NULL PRIMARY KEY, col1 VARCHAR, col2 INTEGER);
upsert into input_table values(1,'col1',1);
upsert into input_table values(2,'col2',2);
CREATE TABLE OUTPUT_TABLE (id BIGINT NOT NULL PRIMARY KEY, col1 VARCHAR, col2 INTEGER);

import org.apache.phoenix.spark._
import org.apache.spark.sql.SaveMode

val df = sqlContext.load("org.apache.phoenix.spark", Map("table" -> "INPUT_TABLE", "zkUrl" -> "192.38.0.231:2181"))

df.save("org.apache.phoenix.spark", Map("table" -> "OUTPUT_TABLE", "zkUrl" -> "192.38.0.231:2181"))

spark2.x向phoenix加载数据

df.write.format("org.apache.phoenix.spark").options( Map("table" -> "OUTPUT_TABLE", "zkUrl" -> "192.38.0.231:2181")).mode("overwrite").save

0: jdbc:phoenix:localhost> select * from output_table;
+-----+-------+-------+
| ID  | COL1  | COL2  |
+-----+-------+-------+
| 1   | col1  | 1     |
| 2   | col2  | 2     |
+-----+-------+-------+
2 rows selected (0.092 seconds)

示例：PageRank

1）下载数据enron.csv.gz

2）Phoenix中创建两张表：

CREATE TABLE EMAIL_ENRON(MAIL_FROM BIGINT NOT NULL, MAIL_TO BIGINT NOT NULL CONSTRAINT pk PRIMARY KEY(MAIL_FROM, MAIL_TO));
CREATE TABLE EMAIL_ENRON_PAGERANK(ID BIGINT NOT NULL, RANK DOUBLE CONSTRAINT pk PRIMARY KEY(ID));

3）加载数据到phoenix

gunzip enron.csv.gz 
./psql.py -t EMAIL_ENRON localhost /opt/enron.csv 

4）在spark中计算pagerange，结果保存到phoenix中

import org.apache.spark.graphx._
import org.apache.phoenix.spark._
val rdd = sc.phoenixTableAsRDD("EMAIL_ENRON", Seq("MAIL_FROM", "MAIL_TO"), zkUrl=Some("localhost"))
val rawEdges = rdd.map{ e => (e("MAIL_FROM").asInstanceOf[VertexId], e("MAIL_TO").asInstanceOf[VertexId]) }
val graph = Graph.fromEdgeTuples(rawEdges, 1.0)
val pr = graph.pageRank(0.001) 
pr.vertices.saveToPhoenix("EMAIL_ENRON_PAGERANK", Seq("ID", "RANK"), zkUrl = Some("localhost"))

5）查看计算结果

0: jdbc:phoenix:localhost> select * from email_enron_pagerank order by rank desc limit 5;
+-------+---------------------+
|  ID   |        RANK         |
+-------+---------------------+
| 5038  | 497.2989872977676   |
| 273   | 117.18141799210386  |
| 140   | 108.63091596789913  |
| 458   | 107.2728800448782   |
| 588   | 106.11840798585399  |
+-------+---------------------+
5 rows selected (0.283 seconds)