在网上查找了zabbix监控docker的方法,基本上分为两种方法:一是用国外大神写的zabbix_module_docker.so加模板来进行监控,这种方法只能看到容器id很不爽;二是自己写脚本进行监控,在网上下载了别人的脚本进行监控,发现数据总是断断续续的,隔一段时间会没数据,索性就自己来写一个简单的脚本。
我的思路是利用docker stats这个docker自带的命令来查看cpu、内存和网络,反正我要求的也不多。。下面正式开始:
1、给zabbix 免密sudo的权限
echo "zabbix ALL=(root) NOPASSWD:/bin/docker,/usr/bin/python,/usr/local/zabbix/scripts/docker_monitor.py,/usr/local/zabbix/scripts/docker_low_discovery.sh,/usr/local/zabbix/scripts/docker_processmonitor.sh">>/etc/sudoers
* 并且需要注释掉#Defaults requiretty这一行
2、在zabbix安装目录下新建一个发现docker的脚本,这个脚本是网上的没有改直接拿过来用
# cat /usr/local/zabbix/scripts/docker_low_discovery.sh
#!/bin/bash
#Fucation:docker low-level discovery
docker() {
port=($(sudo docker ps -a|grep -v "CONTAINER ID"|awk '{print $NF}'))
printf '{\n'
printf '\t"data":[\n'
for key in ${!port[@]}
do
if [[ "${#port[@]}" -gt 1 && "${key}" -ne "$((${#port[@]}-1))" ]];then
printf '\t {\n'
printf "\t\t\t\"{#CONTAINERNAME}\":\"${port[${key}]}\"},\n"
else [[ "${key}" -eq "((${#port[@]}-1))" ]]
printf '\t {\n'
printf "\t\t\t\"{#CONTAINERNAME}\":\"${port[${key}]}\"}\n"
fi
done
printf '\t ]\n'
printf '}\n'
}
case $1 in
docker)
docker
;;
*)
echo "Usage:`basename $0` {docker}"
;;
esac
2、新建第二个监控docker的脚本
#!/bin/bash
#cat /usr/local/zabbix/script/docker_statsmonitor.sh
#author wusong 703249072@qq.com
#2018-5-24
unit_change(){
tmp="$1"
num=0
if [ "$(echo ${tmp}|grep GB)" != "" ]; then
tmp=${tmp%GB*}
num=`echo "scale=2;($tmp*1024)/1" | bc`
elif [ "$(echo ${tmp}|grep kB)" != "" ]; then
tmp=${tmp%kB*}
num=`echo "scale=2;$tmp/1024" | bc | awk '{printf "%.2f", $0}'`
else
num=${tmp%MB*}
fi
echo "${num}"
}
docker_info(){
name="$1"
info_type="$2"
result=""
if [ "${info_type}" = "cpu" ];then
result=$(sudo docker stats ${name} --no-stream | grep -v "CONTAINER ID")
result=$(echo ${result} |awk '{print $3}')
echo "${result%\%*}"
elif [ "${info_type}" = "mem" ];then
result=`sudo docker stats ${name} --no-stream | grep -v "CONTAINER ID" | awk '{print $7}'`
echo "${result%\%*}"
elif [ "${info_type}" = "neti" ];then
tmp=`sudo docker stats ${name} --no-stream |grep -v "CONTAINER ID" | awk '{print $8}'`
unit_change ${tmp}
elif [ "${info_type}" = "neto" ];then
tem=`sudo docker stats ${name} --no-stream |grep -v "CONTAINER ID" |awk '{print $10}'`
unit_change ${tem}
else
echo "Error:patameter wrong"
fi
}
if [ $# -ne 2 ]; then
echo "Usage: must be have two parameter to execute 'containerName & cpu|mem|neti|neto'"
fi
case $2 in
cpu|mem|neti|neto)
docker_info $1 $2
;;
*)
echo "Usage:`basename $0` must be have two parameter to execute 'containerName & cpu|mem|neti|neto'"
;;
esac;
这个脚本作用就是查cpu、内存和网络流量,如果执行的时候报错`bc : command not found`的话,`yum install bc` 安装一下就好了。
3、修改zabbix_agentd.conf
vim zabbix_agentd.conf ,在里面加上两个参数,注意脚本路径改成你的路径
UserParameter=docker_low_discovery[*],/bin/bash /usr/local/zabbix_agent/scripts/docker_low_discovery.sh $1
UserParameter=docker_statsmonitor[*],/bin/bash /usr/local/zabbix_agent/scripts/docker_statsmonitor.sh $1 $2
4、最后在zabbix web界面导入模板,关联主机,稍等即可
模板在文后。
问题调试:
如果等了很久没有出图,或者报错,可以用zabbix_get进行调试,不过使用zabbix_get时必须开启客户端被动模式,要求暴露客户端监听端口。
使用方法:在zabbix_server机器上进入zabbix/bin目录,执行
zabbix_get -s 客户端ip -p 10050 -k "docker_low_discovery[docker]"
或
zabbix_get -s 客户端ip -p 10050 -k "docker_statsmonitor[容器name,cpu]"
模板下载地址:https://download.csdn.net/download/qq_36961530/10437083