Apache 2 Memory Footprint Monitor script in Linux

The Web server encountered a strange problem, in the running of many apache2 process, there will be a process, slowly occupied full memory, and then the machine is like dead,
Write an automatic check script, check and kill the problem of the apache2 process, the issue is mitigated, but not resolved, because slowly fill up the memory, will make Linux release the hard disk cache
and cyclical performance degradation. Limiting memory with RLIMITMEM parameters does not work, and bw_mod limiting traffic does not work

Today modified the script, before killing the problem process, record this process to connect the external IP and port, &http://www.aliyun.com/zixun/aggregation/37954.html ">NBSP;" and modify the Apache log format to increase the process number field.
Check the Access.log log of the faulty process number and find no specific URL access.

The offending process is connected to an external IP that is 2 IP, and these 2 IP agent fields declare that they are from Sogou and are Sogou robots.
With tcpdump to these 2 IP grab bag (tcpdump host 220.181.94.213-s0-w/home/sogou.cap), with Wireshark look grab bag content, traffic is very small, also nothing special.

Monitoring record script:

#!/bin/sh
While [1]
Do
Ps-u www-data-u www-data-u root-u root U |grep apache2 read user PID CPU mem Other
Do
if [${mem%\.*} GT 15]
Then
ip=$ (netstat-npa |grep $pid |tr ': ' |awk ' {printf $} ')
echo ' Date +%y-%m-%d:%h:%m:%s ' kill PID: $pid CPU: $CPU mem: $MEM IP: $ip >>/var/log/apache2/autokill.log
Kill-9 $pid
Fi
Done
Sleep 30
Done