负载、CPU、内存、磁盘、Http服务、Tcp服务监控脚本

原创

ilex 2010-01-05 18:32:02 博主文章分类：LINUX ©著作权

©著作权归作者所有：来自51CTO博客作者ilex的原创作品，请联系作者获取转载授权，否则将追究法律责任

#/bin/sh
#负载、CPU、内存、磁盘、Http服务、Tcp服务监控脚本
#实现发邮件、短信
#基本配置
IP=192.168.2.248
export LANG=en_US.UTF-8

#通知邮件地址列表配置
email_list=(\
liwh@entel.com.cn \
)

#短信通知手机号
sms_list=(\
13800138000 \
)

#需要监控的Http服务器列表
http_list=(\
127.0.0.1:80/secure/Dashboard.jspa \
#127.0.0.1:80/ \
127.0.0.1:80/index.jsp \
)

#需要监控的Tcp服务器列表配置
tcp_list=(\
127.0.0.1:22 \
)

#基本配置结束

#时间格式化
date=$(date -d "today" +"%Y-%m-%d_%H:%M:%S")

#发送邮件函数
#调用方法：send_email subject tempfile

send_email()
{

email_len=${#email_list[*]}
ei=0
while [ $ei -lt $email_len ]
do
[ -e $2    ] &&    mail -s "$1 $date" ${email_list[$ei]} < $2    || echo "TempFile not found:$2" | mail -s "File not found $date" ${email_list[$ei]}

let ei++

done

}
#Test send email
#send_email "Test email" /tmp/a

#飞信发短信
#调用方法 send_sms tempFile
send_sms()
{

     tomobile=`echo    "${sms_list[*]}" |sed 's/ /,/g'`
#     echo $tomobile
     if [ "${#tomobile}" != 0    ];then

  [ -e $1 ] && LD_LIBRARY_PATH=/usr/local/fx/lib    /usr/local/fx/fetion --hide    --config=/usr/local/fx/fx.conf --index=1 --to="$tomobile"    --msg-type=1 --file-utf8="$1"    >/dev/null 2>&1 || LD_LIBRARY_PATH=/usr/local/fx/lib    /usr/local/fx/fetion --hide    --config=/usr/local/fx/fx.conf --index=1 --to="$tomobile"    --msg-type=1 --msg-utf8="Error no this file $1" >/dev/null 2>&1

     fi

}
#Test send_sms
#send_sms /tmp/a.log

#监控系统负载
#调用 monitor_load
monitor_load() {
NOTIFY=$1
#警告阀值
NOTIFY=${NOTIFY:-"6.0"}
#邮件标题
SUBJECT="Alert $IP $(hostname) Load Average"
OS="$(uname)"
TRUE="1"
if [ "$OS" == "FreeBSD" ]; then
                TEMPFILE="$(mktemp /tmp/$(basename $0).tmp.XXX)"
FTEXT='load averages:'
elif [ "$OS" == "Linux" ]; then
                TEMPFILE="$(mktemp)"
FTEXT='load average:'
fi

F5M="$(uptime | awk -F "$FTEXT" '{ print $2 }' | cut -d, -f1)"
F10M="$(uptime | awk -F "$FTEXT" '{ print $2 }' | cut -d, -f2)"
F15M="$(uptime | awk -F "$FTEXT" '{ print $2 }' | cut -d, -f3)"

# mail message
echo "Alert From $IP $(hostname) " >> $TEMPFILE
echo "Load average Crossed allowed limit $NOTIFY." >> $TEMPFILE
echo "Hostname: $(hostname)" >> $TEMPFILE
echo "Local Date & Time : $date" >> $TEMPFILE
echo "5 min load :$F5M"    >> $TEMPFILE
echo "10 min load :$F10M"    >> $TEMPFILE
echo "15 min load :$F15M"    >> $TEMPFILE

# Look if it crossed limit
# compare it with last 15 min load average
RESULT=0
RESULT=$(echo "$F15M > $NOTIFY" | bc)


# if so send an email
if [ "$RESULT" == "$TRUE" ]; then
             send_email "$SUBJECT"    $TEMPFILE
             send_sms    $TEMPFILE
fi

# remove file
rm -f $TEMPFILE
}

#Test monitor_load
#monitor_load 3.0

#监控磁盘空间
#调用方法 monitor_disk

monitor_disk(){

MAXDISK=$1
#警告阀值
MAXDISK=${MAXDISK:-"80"}

SUBJECT="Alert $IP $(hostname) Disk use surpasses the limitation value "

#date=$(date -d "today" +"%Y-%m-%d_%H:%M:%S")

OS="$(uname)"
TRUE="1"
if [ "$OS" == "FreeBSD" ]; then
                TEMPFILE="$(mktemp /tmp/$(basename $0).tmp.XXX)"
elif [ "$OS" == "Linux" ]; then
                TEMPFILE="$(mktemp)"
fi

DISK=`df -h |sed 's/ /1/'|awk {'print $5'}|sed 's/%//g'|sed '1,2d'|sort -n |tail -1`

RESULT=0
RESULT=$(echo "$DISK > $MAXDISK" | bc)

if [ "$RESULT" == "$TRUE" ]; then
             echo "Alert From $IP $(hostname) " >> $TEMPFILE
             echo "DISK Crossed allowed limit $MAXDISK%." >> $TEMPFILE
             echo " " >> $TEMPFILE
             echo    "`df -h`" >>$TEMPFILE
             send_email "$SUBJECT"    $TEMPFILE
             send_sms    $TEMPFILE
fi

rm -f $TEMPFILE

}

#Test monitor_disk
#monitor_disk 40

#CPU监控

monitor_cpu(){

MAXCPU=$1
#警告阀值
MAXCPU=${MAXCPU:-"70"}

SUBJECT="Alert $IP $(hostname) CPU use surpasses the limitation value "

#date=$(date -d "today" +"%Y-%m-%d_%H:%M:%S")

OS="$(uname)"
TRUE="1"
if [ "$OS" == "FreeBSD" ]; then
                TEMPFILE="$(mktemp /tmp/$(basename $0).tmp.XXX)"
elif [ "$OS" == "Linux" ]; then
                TEMPFILE="$(mktemp)"
fi

CPU=`vmstat|sed 1,2d|awk '{print $13}'`

RESULT=0
RESULT=$(echo "$CPU > $MAXCPU" | bc)

if [ "$RESULT" == "$TRUE" ]; then
             echo "Alert From $IP $(hostname) " >> $TEMPFILE
             echo "CPU Crossed allowed limit $MAXCPU%." >> $TEMPFILE
             echo " " >> $TEMPFILE
             ps aux|sed 1d |awk '{if    ($3 >= 0) {print $0}}' >>$TEMPFILE
             send_email "$SUBJECT"    $TEMPFILE
             send_sms $TEMPFILE
fi

rm -f $TEMPFILE

}

#Test monitor_cpu
#monitor_cpu 30

#内存监控
monitor_mem(){

MAXMEM=$1
#剩余内存阀值
MAXMEM=${MAXMEM:-"200"}

SUBJECT="Alert $IP $(hostname) MEM check info "

OS="$(uname)"
TRUE="1"
if [ "$OS" == "FreeBSD" ]; then
                TEMPFILE="$(mktemp /tmp/$(basename $0).tmp.XXX)"
elif [ "$OS" == "Linux" ]; then
                TEMPFILE="$(mktemp)"
fi

MEMFREE=`free -m|grep '^M'|awk '{print $4}'`

RESULT=0
RESULT=$(echo "$MEMFREE <= $MAXMEM" | bc)

if [ "$RESULT" == "$TRUE" ]; then
#             echo "The memory surplus space is smaller than    $MAXMEM M." >> $TEMPFILE
            echo "Alert From $IP $(hostname) " >> $TEMPFILE
            echo "Memory remaining space is less than    $MAXMEM M." >> $TEMPFILE
            echo "Memory surplus space : $MEMFREE M." >> $TEMPFILE
            echo " " >> $TEMPFILE
            free -m|grep '^[M,S]'|awk 'BEGIN{print "Item\tTotal\tUSED\tFREE\tFree%"}{z=100*$5}{print $1"\t"$2"\t"$3"\t"$4"\t"$4/$2*100}' >>$TEMPFILE
            send_email "$SUBJECT"    $TEMPFILE
            send_sms $TEMPFILE
fi

rm -f $TEMPFILE

}

#Test monitor_mem
#monitor_mem 30

#HTTP服务器监控

monitor_http(){

SUBJECT="Alert $IP $(hostname) Http check Info "

#date=$(date -d "today" +"%Y-%m-%d_%H:%M:%S")

http_len=${#http_list[*]}

hi=0
while    [ $hi -lt $http_len ]
do
     OS="$(uname)"
     if [ "$OS" == "FreeBSD" ]; then
                TEMPFILE="$(mktemp /tmp/$(basename $0).tmp.XXX)"
     elif [ "$OS" == "Linux" ]; then
                TEMPFILE="$(mktemp)"
     fi

     server_ip=$(echo ${http_list[$hi]} | awk -F ':' '{print $1}')
     server_port=$(echo ${http_list[$hi]} | awk -F ':' '{print $2}' | awk -F '/' '{print $1}')
     server_url=$(echo ${http_list[$hi]}    | awk -F '/' '{$1="";print $0}'| sed 's/ /\//g')

     if curl -m 10 -I -f    http://${http_list[$hi]} > /dev/null 2>&1
     then
         #status:        0,http down        1,http ok        2,http down but ping ok
         status=1
                     echo "Server:${server_ip}, Port:${server_port}, URL:${server_url} is OK!" >> $TEMPFILE
     else
             if curl -m 30 -I -f http://${http_list[$hi]} > /dev/null 2>&1
             then
                     status=1
                     echo "Server:${server_ip}, Port:${server_port}, URL:${server_url} is OK !" >> $TEMPFILE
             else
                     if ping -c 1 $server_ip > /dev/null 2>&1
                     then
                             status=2
                echo "Alert From $IP $(hostname) " >> $TEMPFILE
                             echo "Server:${server_ip}, Port:${server_port}, URL:${server_url} is ERROR! Ping is OK!" >> $TEMPFILE
                     else
                             status=0
                echo "Alert From $IP $(hostname) " >> $TEMPFILE
                             echo "Server:${server_ip}, Port:${server_port}, URL:${server_url} is ERROR! Ping is ERROR!" >>    $TEMPFILE
                     fi
        curl -m 10 -I http://${http_list[$hi]} >>    $TEMPFILE    2>/dev/null
             fi
     fi

     if [ $status != 1 ] ;then
  send_email "$SUBJECT"    $TEMPFILE
  send_sms     $TEMPFILE
     fi
     rm -f $TEMPFILE

        let hi++

done
}

#Test Http
#monitor_http

#TCP服务器监控

monitor_tcp(){

SUBJECT="Alert $IP $(hostname) TCP check Info "

tcp_len=${#tcp_list[*]}
ti=0
while    [ $ti -lt $tcp_len ]
do

     OS="$(uname)"
     if [ "$OS" == "FreeBSD" ]; then
                TEMPFILE="$(mktemp /tmp/$(basename $0).tmp.XXX)"
     elif [ "$OS" == "Linux" ]; then
                TEMPFILE="$(mktemp)"
     fi

     server_ip=$(echo ${tcp_list[$ti]} | awk -F ':' '{print $1}')
     server_port=$(echo ${tcp_list[$ti]} | awk -F ':' '{print $2}')
     if nc -vv -z -w 3 $server_ip $server_port > /dev/null 2>&1
     then
             #status:        0,http down        1,http ok        2,http down but ping ok
             status=1
             echo "Server:${server_ip}, Port:${server_port} is Ok !" >> $TEMPFILE
     else
             if nc -vv -z -w 10 $server_ip $server_port > /dev/null 2>&1
             then
                     status=1
                     echo "Server:${server_ip}, Port:${server_port} is Ok !"    >> $TEMPFILE
             else
                     if ping -c 1 $server_ip > /dev/null 2>&1
                     then
                             status=2
                echo "Alert From $IP $(hostname) " >> $TEMPFILE
                             echo "Server:${server_ip}, Port:${server_port}is ERROR,Ping is Ok!" >> $TEMPFILE
                     else
                             status=0
                echo "Alert From $IP $(hostname) " >> $TEMPFILE
                             echo "Server:${server_ip}, Port:${server_port}is ERROR,Ping is ERROR !" >> $TEMPFILE
                     fi
             fi
     fi

     if [ $status != 1 ] ;then
                send_email "$SUBJECT"    $TEMPFILE
  send_sms    $TEMPFILE
     fi
     rm -f $TEMPFILE

     let ti++
done

}

#Test monitor_tcp
#monitor_tcp

#调用函数

monitor_load 3.0
monitor_disk 40
monitor_cpu 30
monitor_mem 20
monitor_http
monitor_tcp

飞信安装到 /usr/local/fx 目录

配置帐户文件 /usr/local/fx/fx.conf (index mobile password)

下地址：http://www.it-adv.net/fetion/downng/fetion20091117-linux.tar.gz

命令：

    以下为 windows 下举例:
    fetion --mobile=13711123456 --pwd=mypwd --to=137xxxxxxxx --msg-gb=测试
    fetion --sid=6630321 --pwd=mypwd --to=137xxxxxxxx --msg-gb=测试
    fetion --config=sample.conf --index=1 --to=137xxxxxxxx --msg-gb=测试

    linux下，请使用如下命令：
    LD_LIBRARY_PATH=. ./fetion --mobile=13711123456 --pwd=mypwd --to=137xxxxxxxx --msg-utf8=测试
    LD_LIBRARY_PATH=. ./fetion --sid=6630321 --pwd=mypwd --to=137xxxxxxxx --msg-utf8=测试
    LD_LIBRARY_PATH=. ./fetion --config=sample.conf --index=1 --to=137xxxxxxxx --msg-utf8=测试