Spark系列-SparkSQL实战

• Spark系列-核心概念
• Spark系列-SparkSQL实战
• Spark性能优化指南——基础篇
• Spark性能优化指南——高级篇

之前系统的计算大部分都是基于Kettle + Hive的方式，但是因为最近数据暴涨，很多Job的执行时间超过了1个小时，即使是在优化了HiveQL的情况下也有超过30分钟，所以近期把计算引擎从Hive变更为Spark。

普通的简单Job就使用SparkSQL来计算，数据流是经过spark计算，把结果插入到Mysql中

在项目中新建三个类，第一个Logger类用于日志的输出

# coding=utf-8
import logging
from logging import handlers

class Logger(object):
    leven_relations = {
        'debug':logging.DEBUG,
        'info':logging.INFO,
        'warning': logging.WARNING,
        'error': logging.ERROR
    }

    def __init__(self, fileName, level='info', when='D', backCount=3, fmt='%(asctime)s - %(pathname)s[line:%(lineno)d] - %(levelname)s: %(message)s'):
        self.logger = logging.getLogger(fileName)
        format_str = logging.Formatter(fmt)
        self.logger.setLevel(self.leven_relations.get(level))
        #屏幕日志
        sh = logging.StreamHandler()
        sh.setFormatter(format_str)
        #文件日志
        th = handlers.TimedRotatingFileHandler(filename=fileName, when=when, backupCount=backCount, encoding='utf-8')
        th.setFormatter(format_str)
        self.logger.addHandler(th)
        self.logger.addHandler(sh)

第二个是SparkSQL公共类，引用的是pyspark

# coding=utf-8

from pyspark import SparkConf,SparkContext
from pyspark.sql import HiveContext

class SparkSqlCommon(object):
    sql_str = ''
    app_name = ''

    def __init__(self, sql, app_name):
        if sql is None:
            raise Exception('sql cannot be empty')
        self.sql_str = sql

        if app_name is None:
            raise Exception('app_name cannot be empty')
        self.app_name = app_name

    def execute(self):
        spark_conf = SparkConf().setAppName(self.app_name)
        spark_context = SparkContext(conf=spark_conf)
        spark_context.setLogLevel("INFO")
        hive_context = HiveContext(spark_context)
        result_rdd = hive_context.sql(self.sql_str)
        result = result_rdd.collect()
        return result

第三个是Mysql公共类,用于把计算结果落地到mysql

# coding=utf-8

import pymysql
from com.randy.common.Logger import Logger

class DatacenterCommon(object):
    sql_str = ''
    jdbcHost = ''
    jdbcPort = ''
    jdbcSchema = ''
    jdbcUserName = ''
    jdbcPassword = ''

    def __init__(self, sql_str,log, jdbcHost = '127.0.0.1', jdbcPort = 3306, jdbcSchema = 'demo', jdbcUserName = 'root', jdbcPassword = '123456'):
        if sql_str is None:
            raise Exception('sql_str cannot be empty')

        self.sql_str = sql_str
        self.jdbcHost = jdbcHost
        self.jdbcPort = jdbcPort
        self.jdbcSchema = jdbcSchema
        self.jdbcUserName = jdbcUserName
        self.jdbcPassword = jdbcPassword
        self.log = log


    def execute(self):
        db = pymysql.connect(host=self.jdbcHost,
                             port=self.jdbcPort,
                             user=self.jdbcUserName,
                             passwd=self.jdbcPassword,
                             db=self.jdbcSchema,
                             charset='utf8')
        try:
            db_cursor = db.cursor()
            db_cursor.execute(self.sql_str)
            db.commit()
        except Exception, e:
            self.log.logger.error('str(e):\t\t', str(e))
            db.rollback()

调用的客户端代码如下

# coding=utf-8
# !/usr/bin/python2.7

import datetime
from com.randy.spark.Logger import Logger
from com.randy.spark.SparkSqlCommon import SparkSqlCommon
from com.randy.spark.DatacenterCommon import DatacenterCommon




#需要修改，每个应用都不一样
app_name = 'demo1'

# SparkSql（不能以分号结尾）
select_sql = '''
                  SELECT count(*) from futures.account
'''

# Mysql
insert_sql = '''
            insert into demo.demo1(id) values({0});
'''


if __name__ == '__main__':
    currentDay = datetime.datetime.now().strftime('%Y%m%d')
    log = Logger('/home/python-big-data-job/log/' + app_name + "_" + str(currentDay) + '.log')
    log.logger.info("**************************start invoke {0},{1} *****************".format(app_name,currentDay))

    sparkSqlCommon = SparkSqlCommon(sql=select_sql,app_name=app_name)
    selectResult = sparkSqlCommon.execute()
    log.logger.info("sparkSqlCommon result:{0}".format(selectResult))
    if selectResult is None:
        log.logger.error("taojin_1 selectResult while is empty")
    else:
        insert_sql = insert_sql.format(selectResult[0][0])
        log.logger.info(insert_sql)
        datacenterCommon = DatacenterCommon(sql_str=insert_sql, log=log)
        datacenterCommon.execute()

        log.logger.info("**************************end invoke {0},{1} *****************".format(app_name, currentDay))

其中spark-submit提交代码如下：

sudo -u hdfs spark-submit --master local[*] --py-files='/home/python-big-data-job/com.zip,/home/python-big-data-job/pymysql.zip' /home/python-big-data-job/taojin/demo1.py

因为项目中使用到了本地文件，所有把三个公共类打包到了com.zip中作为依赖文件

其中pymysql.zip是pymysql的源码文件，因为我在过程中发现了ImportError: No module named pymysql

但是集群已经使用pip安装了pymysql，没有找到有效解决办法，按照https://zhuanlan.zhihu.com/p/43434216和https://www.cnblogs.com/piperck/p/10121097.html都无效，最终只能把pymysql以依赖文件的方式打包

其中使用yarn cluster部署也还存在问题