目录
该脚本用于监控VPS服务器负载,Web程序内存及CPU使用。当服务器系统负载或内存使用达到预设值,则重启该程序,或者某个php-cgi进程占用CPU过大,则直接kill掉该进程。目的在于缓解服务器资源耗尽导致意外宕机等情况。
一、使用方法:
bash
git clone git://gist.github.com/1216837.git gist-1216837
vim gist-1216837/sys-mon.sh //修改内存、CPU等预设阀值
mkdir /var/script
mv gist-1216837/sys-mon.sh /var/script设置每分钟执行一次
bash
crontab -e
* * * * * /bin/bash /var/shell/sys-mon.sh二、Shell脚本内容
建议打开下面网址查看最新版本。
bash
#! /bin/bash
#====================================================================
# sys-mon.sh
#
# Copyright (c) 2011, WangYan <webmaster@wangyan.org>
# All rights reserved.
# Distributed under the GNU General Public License, version 3.0.
#
# Monitor system mem and load, if too high, restart some service.
#
# See: http://www.wangyan.cloud/sys-mon-shell-script.html
#
# V 0.5, Date: 2011-12-08
#====================================================================
# Need to monitor the service name
# Must be in /etc/init.d folder exists
NAME_LIST="httpd nginx mysql"
# Single process to allow the maximum CPU (%)
PID_CPU_MAX="25"
# The maximum allowed memory (%)
PID_MEM_SUM_MAX="95"
# The maximum allowed system load
SYS_LOAD_MAX="6"
# Log path settings
LOG_PATH="/var/log/sys-mon.log"
# Date time format setting
DATA_TIME=$(date +"%y-%m-%d %H:%M:%S")
# Your email address
EMAIL="webmaster@example.com"
# Your website url
MY_URL="http://106.187.38.210/p.php"
#====================================================================
for NAME in $NAME_LIST
do
PID_CPU_SUM="0";PID_MEM_SUM="0"
PID_LIST=`ps aux | grep $NAME | grep -v root`
IFS_TMP="$IFS";IFS=$'\n'
for PID in $PID_LIST
do
PID_NUM=`echo $PID | awk '{print $2}'`
PID_CPU=`echo $PID | awk '{print $3}'`
PID_MEM=`echo $PID | awk '{print $4}'`
# echo "$NAME: PID_NUM($PID_NUM) PID_CPU($PID_CPU) PID_MEM($PID_MEM)"
PID_CPU_SUM=`echo "$PID_CPU_SUM + $PID_CPU" | bc`
PID_MEM_SUM=`echo "$PID_MEM_SUM + $PID_MEM" | bc`
if [ `echo "$PID_CPU >= $PID_CPU_MAX" | bc` -eq 1 ];then
if [[ "$NAME" = "php-fpm" || "$NAME" = "httpd" ]];then
sleep 5
if [ `echo "$PID_CPU >= $PID_CPU_MAX" | bc` -eq 1 ];then
echo "${DATA_TIME}: kill ${NAME}($PID_NUM) successful (CPU:$PID_CPU)" | tee -a $LOG_PATH
kill $PID_NUM
fi
else
echo "${DATA_TIME}: [WARNING!] ${NAME}($PID_NUM) cpu usage is too high! (CPU:$PID_CPU)" | tee -a $LOG_PATH
fi
fi
done
IFS="$IFS_TMP"
SYS_LOAD=`uptime | awk '{print $(NF-2)}' | sed 's/,//'`
SYS_MON="CPU:$PID_CPU_SUM MEM:$PID_MEM_SUM LOAD:$SYS_LOAD"
# echo -e "$NAME: $SYS_MON\n"
SYS_LOAD_TOO_HIGH=`awk 'BEGIN{print('$SYS_LOAD'>'$SYS_LOAD_MAX')}'`
PID_MEM_SUM_TOO_HIGH=`awk 'BEGIN{print('$PID_MEM_SUM'>'$PID_MEM_SUM_MAX')}'`
if [[ "$SYS_LOAD_TOO_HIGH" = "1" || "$PID_MEM_SUM_TOO_HIGH" = "1" ]];then
/etc/init.d/$NAME stop
sleep 5
for ((i=1;i<4;i++))
do
if [ `pgrep $NAME | wc -l` = "0" ];then
echo "$DATA_TIME: Stop $NAME successful! ($SYS_MON)" | tee -a $LOG_PATH
break
else
echo "${DATA_TIME}: [WARNING!] Stop $NAME failed[$i]! ($SYS_MON)" | tee -a $LOG_PATH
pkill $NAME && killall $NAME
fi
done
/etc/init.d/$NAME start
sleep 5
for ((ii=1;ii<4;ii++))
do
if [ `pgrep $NAME | wc -l` != "0" ];then
echo "$DATA_TIME: Start $NAME successful!" | tee -a $LOG_PATH
break
else
echo "${DATA_TIME}: [WARNING!] Start $NAME failed[$ii]! ($SYS_MON)" | tee -a $LOG_PATH
/etc/init.d/$NAME start
sleep 5
fi
done
if [ `pgrep $NAME | wc -l` != "0" ];then
echo "${DATA_TIME}: [ERROR!] Start $NAME failed! ($SYS_MON)" | mail -s "Start $NAME failed" $EMAIL
fi
fi
done
STATUS_CODE=`curl -o /dev/null -s -w %{http_code} $MY_URL`
#echo -e "STATUS CODE: $STATUS_CODE\n"
if [ "$STATUS_CODE" != "200" ];then
sleep 3
STATUS_CODE=`curl -o /dev/null -s -w %{http_code} $MY_URL`
if [ "$STATUS_CODE" != "200" ];then
echo "${DATA_TIME}: [WARNING!] Website Downtime! ($SYS_MON)" | tee -a $LOG_PATH
echo "${DATA_TIME}: [WARNING!] Website Downtime! ($SYS_MON)" | mail -s "Start $NAME failed" $EMAIL
fi
fi脚本内容不难理解,原理解释可参考 Linux 进程自动监控shell脚本
三、注意事项
- NAME_LIST 指定的监控程序必须在/etc/init.d 文件夹中存在,并且支持stop和start操作
- PID_CPU_MAX 指的是单个进程的CPU占用,只针对php-fpm或httpd。
- PID_MEM_SUM_MAX 指的是该程序所有进程实际内存占用,而并非系统总内存。
- EMAIL 只有在程序启动失败后,你才能收到邮件提醒。
四、更新历史:
2011.11.28: 去掉nginx502状态监控,完善进程cpu监控,修正数据不准确等问题。 2011.12.07: 继续修正cpu监控不正确问题,增加宕机后邮件通知功能。