需求:通过hadoop查询指定目录下创建日期为180天之前的目录,将其删除,通过crontab定期执行

编辑shell脚本文件,其中调用hadoop命令,本地sh执行没有问题,但是通过crontab调用时未执行,原因可能是环境变量未设置导致

先查询本地环境变量

echo $PATH

然后将环境变量内容写入shell脚本中

hadoop fs -ls查询指定目录下的所有目录名称

[hdp-test@hadoopclient01v ~]$ hadoop fs -ls /home/hdp-xxx/test1/data/
Found 15 items
drwxr-xr-x   - work hdp-test          0 2020-02-02 15:01 /home/hdp-test/test1/data/axx
drwxr-xr-x   - work hdp-test          0 2020-02-02 14:46 /home/hdp-test/test1/data/bxx
drwxr-xr-x   - work hdp-test          0 2020-02-02 14:46 /home/hdp-test/test1/data/cxx
drwxr-xr-x   - work hdp-test          0 2020-02-02 14:46 /home/hdp-test/test1/data/dxx
drwxr-xr-x   - work hdp-test          0 2020-02-02 17:26 /home/hdp-test/test1/data/exx
drwxr-xr-x   - work hdp-test          0 2020-02-02 17:26 /home/hdp-test/test1/data/flow
drwxr-xr-x   - work hdp-test          0 2020-02-02 14:46 /home/hdp-test/test1/data/ftp
drwxr-xr-x   - work hdp-test          0 2020-02-02 14:46 /home/hdp-test/test1/data/ldap
drwxr-xr-x   - work hdp-test          0 2020-02-02 14:47 /home/hdp-test/test1/data/login
drwxr-xr-x   - work hdp-test          0 2020-02-02 14:47 /home/hdp-test/test1/data/mail
drwxr-xr-x   - work hdp-test          0 2020-02-02 14:47 /home/hdp-test/test1/data/tcpxx
drwxr-xr-x   - work hdp-test          0 2020-02-02 13:49 /home/hdp-test/test1/data/udpxx

因为文件名称是第8个字段,所以通过 | awk '{printf $8"\n"}' 可以只得到目录名称

判断目录创建日期是否为180天以前,是的话就删除该目录(hadoop fs -rm -r)

脚本内容如下:

#!/bin/sh

PATH=/bin:/home/hdp-test/.local/bin:/root/.local/bin:/root/perl5/bin:/sbin:/usr/bin:/usr/bin/hadoop/software/distcp:/usr/bin/hadoop/software/hadoop//bin:/usr/bin/hadoop/software/hbase//bin:/usr/bin/hadoop/software/hive//bin:/usr/bin/hadoop/software/java//bin:/usr/bin/hadoop/software/spark//bin:/usr/sbin

export $PATH

dirs=$(hadoop fs -ls /home/hdp-test/test1/data | awk '{printf $8"\n"}')

for dir in $dirs
do
    old_version=$(hadoop fs -ls $dir | awk 'BEGIN{ days_ago=strftime("%F", systime()-180*24*3600) }{ if($6<days_ago){printf "%s\n", $8} }')
    for version in $old_version
    do
        #printf $version'\n'
        hadoop fs -rm -r $version
    done
done

 

posted on 2020-02-02 17:43  longfei2021  阅读(602)  评论(0编辑  收藏  举报