5. 梳理离线批次spark作业提交脚本

下面脚本用于检测运行的job个数

#!/bin/bash -x

#basic config

EMAIL=XXX@qq.com

JOB_NAME="SparkALS"

CDHUSER=admin

CDHPASSWORD=admin

CDHURL=http://ip:7180

JSON=/tmp/yarnApplications_${JOB_NAME}.json

LOG=/tmp/applications.log

#check job is or not running?

# 搜索 cm api 查看 api版本 cdh5.15是v19

## 通过cdh的接口获得 运行作业的数据, 包括作业名, 运行状态, appid, 资源等信息

curl -u $CDHUSER:$CDHPASSWORD $CDHURL/api/v16/clusters/RZCluster/services/yarn/yarnApplications > $JSON

## grep 过滤出job

cat $JSON | grep -A 4 $JOB_NAME > $LOG

## 找到 运行状态的 job 个数

RUNNINGNUM=`cat $LOG | grep "state" | grep "RUNNING"  | wc -l`

echo "The running $JOB_NAME job num is $RUNNINGNUM"

## 如果运行的job 个数 大于0, 就发送邮件

if [ $RUNNINGNUM -gt 0  ]

then

        echo -e "`date "+%Y-%m-%d %H:%M:%S"` : The current running $JOB_NAME job num is $RUNNINGNUM." | mail \

        -r "From: alertAdmin <${EMAIL}>" \

        -s "Warn: Skip the new $JOB_NAME spark job." ${EMAIL}

        exit 0

fi

#spark parameters

CLASS_NAME=org.apache.spark.examples.SparkALS

JAR_PATH=/opt/cloudera/parcels/SPARK2/lib/spark2/examples/jars/spark-examples_2.11-2.1.0.cloudera1.jar

#submit spark job

spark2-submit \

--master yarn \

--class $CLASS_NAME \

$JAR_PATH

exit