[大数据]Bug：Zabbix对Kafka topic积压数据监控

简述

《Zabbix监控Kafka topic积压数据》一文的目的是通过Zabbix自动发现实现对多个消费者组的Topic及Partition的Lag进行监控。因在实际监控中发现有问题，为给感兴趣的读者不留坑，特通过此文对监控进行优化调整。

分区自动发现

# 未优化前的计算方式：
# 自动发现配置文件
vim consumer-groups.conf
#按消费者组(Group)|Topic格式，写入自动发现配置文件
test-group|test

# 执行脚本自动发现指定消费者和topic的分区
bash consumer-groups.sh discovery
{
    "data": [
        { "{#GROUP}":"test-group", "{#TOPICP}":"test", "{#PARTITION}":"0" },
        { "{#GROUP}":"test-group", "{#TOPICP}":"test", "{#PARTITION}":"1" },
        { "{#GROUP}":"test-group", "{#TOPICP}":"test", "{#PARTITION}":"3" },
        { "{#GROUP}":"test-group", "{#TOPICP}":"test", "{#PARTITION}":"2" }
    ]
}

经过上线验证，当自动发现配置文件只有一个test-group|test是没有问题的，但当我们按需求再接入test-group|test1 （即test-group消费者组的第二个Topic）时，自动发现的结果如下：

# 未优化前的计算方式：
vim consumer-groups.conf
#按消费者组(Group)|Topic格式，写入自动发现配置文件
test-group|test
test-group|test1

# 执行脚本自动发现指定消费者和topic的分区
bash consumer-groups.sh discovery
{
    "data": [
        { "{#GROUP}":"test-group", "{#TOPICP}":"test", "{#PARTITION}":"0" },
        { "{#GROUP}":"test-group", "{#TOPICP}":"test", "{#PARTITION}":"1" },
        { "{#GROUP}":"test-group", "{#TOPICP}":"test", "{#PARTITION}":"3" },
        { "{#GROUP}":"test-group", "{#TOPICP}":"test", "{#PARTITION}":"2" }
        { "{#GROUP}":"test-group", "{#TOPICP}":"test1", "{#PARTITION}":"0" },
        { "{#GROUP}":"test-group", "{#TOPICP}":"test2", "{#PARTITION}":"1" },
        { "{#GROUP}":"test-group", "{#TOPICP}":"test3", "{#PARTITION}":"2" }
    ]
}

了解Zabbix自动发现格式的同学会发现，每个Topic的Partition会出现','，这种格式是不符合规范，这就是导致我们的监控项会出现问题，因此我们需要进一步修改脚本。

经修改后，最终效果应该如下：

# 优化后的计算方式：
vim consumer-groups.conf
#按消费者组(Group)|Topic格式，写入自动发现配置文件
test-group|test
test-group|test1

# 执行脚本自动发现指定消费者和topic的分区
bash consumer-groups.sh discovery
{
    "data": [
        { "{#GROUP}":"test-group", "{#TOPICP}":"test", "{#PARTITION}":"0" },
        { "{#GROUP}":"test-group", "{#TOPICP}":"test", "{#PARTITION}":"1" },
        { "{#GROUP}":"test-group", "{#TOPICP}":"test", "{#PARTITION}":"3" },
        { "{#GROUP}":"test-group", "{#TOPICP}":"test", "{#PARTITION}":"2" }，
        { "{#GROUP}":"test-group", "{#TOPICP}":"test1", "{#PARTITION}":"0" },
        { "{#GROUP}":"test-group", "{#TOPICP}":"test1", "{#PARTITION}":"1" },
        { "{#GROUP}":"test-group", "{#TOPICP}":"test1", "{#PARTITION}":"2" }
    ]
}

获取监控项“test-group/test/分区X”的Lag

经过自动发现后的数据，我们可以进一步获取不同分区的lag

# 优化后的计算方式：
# test-group test分区0 lag
bash consumer-groups.sh lag test-group test 0
# test-group test分区1 lag
bash consumer-groups.sh lag test-group test 1
# test-group test1分区0 lag
bash consumer-groups.sh lag test-group test1 0

通过命令可以看到，我们的参数通过消费者组、Topic、Partition来获取最终的lag值，如果不加消费者区分，那么无法区分不同消费者组和不同Topic相应的lag结果：

# 未优化前的计算方式：
# 获取分区0 lag
bash consumer-groups.sh lag 0
# 获取分区1 lag
bash consumer-groups.sh lag 1
# 获取分区2 lag
bash consumer-groups.sh lag 2
# 获取分区3 lag
bash consumer-groups.sh lag 3

最终优化后脚本

# 自动发现配置文件
vim consumer-groups.conf
#按消费者组(Group)|Topic格式，写入自动发现配置文件
test-group|test
test-group|test1

# 自动发现、lag计算脚本
vim consumer-groups.sh
#!/bin/bash
##comment: 根据消费者组监控topic lag，进行监控告警
#配置文件说明
#消费者组|Topic
#test-group|test

#获取topic 信息
cal_topic() {
    if [ $# -ne 2 ]; then
        echo "parameter num error, 读取topic信息失败"
        exit 1
    else
        /usr/local/kafka/bin/./kafka-consumer-groups.sh --bootstrap-server 192.168.3.55:9092 --describe --group $1 |grep -w $2|grep -v none 
    fi
}
#topic+分区自动发现
topic_discovery() {
    printf "{\n"
    printf "\t\"data\": [\n"
    m=0
    num=`cat /etc/zabbix/monitor_scripts/consumer-groups.conf|wc -l`
    for line in `cat /etc/zabbix/monitor_scripts/consumer-groups.conf`
    do  
        m=`expr $m + 1`
        group=`echo ${line} | awk -F'|' '{print $1}'`
        topic=`echo ${line} | awk -F'|' '{print $2}'`
        cal_topic $group $topic > /tmp/consumer-group-tmp
        count=`cat /tmp/consumer-group-tmp|wc -l`
        n=0
        while read line
        do
             n=`expr  $n + 1`
             #判断最后一行
             if [ $n -eq $count ] && [ $m -eq $num ]; then
                 topicp=`echo $line | awk '{print $1}'`
                 partition=`echo $line  | awk '{print $2}'`
                 printf "\t\t{ \"{#GROUP}\":\"${group}\", \"{#TOPICP}\":\"${topicp}\", \"{#PARTITION}\":\"${partition}\" }\n"
             else
                 topicp=`echo $line | awk '{print $1}'`
                 partition=`echo $line  | awk '{print $2}'`
                 printf "\t\t{ \"{#GROUP}\":\"${group}\", \"{#TOPICP}\":\"${topicp}\", \"{#PARTITION}\":\"${partition}\" },\n"
             fi
        done < /tmp/consumer-group-tmp
    done
    printf "\t]\n"
    printf "}\n"
}


if [ $1 == "discovery" ]; then
    topic_discovery
elif [ $1 == "lag" ];then
    cal_topic $2 $3 > /tmp/consumer-group
    cat /tmp/consumer-group |awk -v t=$3 -v p=$4 '{if($1==t && $2==p ){print $5}}'
else
    echo "Usage: /data/scripts/consumer-group.sh discovery | lag"
fi

# 手动运行
## 自动发现
bash consumer-groups.sh discovery
## test-group test分区0 lag
bash consumer-groups.sh lag test-group test 0

接入Zabbix

1.Zabbix配置文件

vim userparameter_kafka.conf
UserParameter=topic_discovery,bash /data/scripts/consumer-groups.sh discovery
UserParameter=topic_log[*],bash /data/scripts/consumer-groups.sh lag "$1" "$2" "$3"

2.Zabbix自动发现

3.监控项配置

4.告警信息

告警主机：Kafka_192.168.3.55
主机IP：192.168.3.55
主机组：Kafka
告警时间：2022.03.21 00:23:10
告警等级：Average
告警信息：test-group/test/分区1:数据积压100
告警项目：topic_lag[test-group,test,1]
问题详情：
test-group/test/1: 62