安装alertmanager

到https://prometheus.io/download/下载alertmanager

解压后编辑alertmanager.yml，只实现报警功能，修改如下

global:resolve_timeout: 5m                 smtp_smarthost: 'smtp.163.com:25'smtp_from: '***@163.com'smtp_auth_username: '***@163.com'smtp_auth_password: '******'                  #授权密码smtp_require_tls: false
route:group_by: ['alertname']group_wait: 10sgroup_interval: 10srepeat_interval: 1m                                  #重复间隔，这里设置为1m，生产环境设置为20m-30m左右receiver: 'mail'
receivers:
- name: 'mail'email_configs:- to: '@@@@@@163.com'

启动

nohup ./alertmanager --config.file=/root/alertmanager-0.17.0.linux-amd64/alertmanager.yml &

修改prometheus配置如下（和alertmanager同主机）：

alerting:alertmanagers:- static_configs:- targets:- 127.0.0.1:9093    #本机启动Alertmanager因此使用127.0.0.1，也可部署在其他主机
rule_files:- "rules/*.yml"          #设置报警规则文件

添加node普遍报警法规则内容如下：

groups:
- name: general.rulesrules:# Alert for any ×××tance that is unreachable for >5 minutes.- alert: InstanceDownexpr: up == 0for: 1mlabels:severity: errorannotations:summary: "Instance {{ $labels.×××tance }} down"description: "{{ $labels.×××tance }} of job {{ $labels.job }} has been down for more than 5 minutes."

查看prometheus中，Targets状态，此时node  http://192.168.199.221:9100/metrics是up状态

将221的node_exporter停掉，再次观察

查看Alerts状态

稍等片刻收到报警

添加内存的报警规则内容如下：

groups:
- name: mem.rulesrules:# Alert for any ×××tance that is unreachable for >5 minutes.- alert: NodeMemoryUsageexpr: (node_memory_MemTotal_bytes - (node_memory_MemFree_bytes+node_memory_Buffers_bytes+node_memory_Cached_bytes )) / node_memory_MemTotal_bytes * 100 > 5for: 1mlabels:severity: errorannotations:summary: "Instance {{ $labels.×××tance }} down"description: "{{ $labels.×××tance }} of job {{ $labels.job }} has been down for more than 5 minutes."

说明：由于是测试使用，将使用率调整为超过5%就报警

重载prometheus。查看prometheus  ui中alert状态

查看规则是否生效

过片刻，收到报警

添加cpu报警规则文件内容如下：

groups:
- name: cpu.rulesrules:# Alert for any ×××tance that is unreachable for >5 minutes.- alert: NodeCpuUsageexpr: 100-irate(node_cpu_seconds_total{job="node",mode="idle"}[5m])*100 > 1for: 1mlabels:severity: errorannotations:summary: "{{ $labels.×××tance }} cpu useage load too high"description: "{{ $labels.×××tance }} of job {{ $labels.job }} has been too hgih for more than 1 minutes."

这里设置阈值为1%，只是测试使用。重载prometheus，过片刻收到报警