玩prometheus过程中遇到的一些问题

标签：name 遇到 self labels job 192.168 prometheus 9182 过程

一、pgw的无默认值监控项

1、prometheus的配置文件

global:
  scrape_interval: 15s # Set the scrape interval to every 15 seconds. Default is every 1 minute.
  evaluation_interval: 1m # Evaluate rules every 15 seconds. The default is every 1 minute.
  # scrape_timeout is set to the global default (10s).

alerting:
  alertmanagers:
    - static_configs:
        - targets:
           - localhost:9093

# Load rules once and periodically evaluate them according to the global 'evaluation_interval'.
rule_files:
   - "/opt/soft/alertmanager-0.25.0.linux-amd64/local_rules.yml"

scrape_configs:
  - job_name: "pushGateway"

    static_configs:
      - targets: ["localhost:9099"]
    #relabel_configs:
    #  - source_labels: [srcIp]
    #    separator: ':'
    #    regex: '(.*)'
    #    replacement: '${1}'
    #    target_label: instance
    
  - job_name: "windows_prometheus"
    static_configs:
      - targets: 
          - 192.168.61.153:9182
          - 192.168.62.238:9182
          - 192.168.62.157:9182
          - 192.168.62.17:9182
          - 192.168.66.175:9182
          - 192.168.62.54:9182
          - 192.168.62.94:9182
          - 192.168.62.87:9182
          - 192.168.62.52:9182
          - 192.168.62.224:9182
          - 192.168.62.222:9182
          - 192.168.62.185:9182
          - 192.168.62.71:9182
          - 192.168.62.110:9182
          - 192.168.62.88:9182
          - 192.168.62.28:9182
          - 192.168.62.175:9182
          - 192.168.62.247:9182
    relabel_configs:
      - source_labels: [__address__]
        separator: ':'
        regex: '(.*):(.*)'
        replacement: '${1}'
        target_label: src_instance

View Code

起初自己设定的配置是两个单纯的static_configs，用自带的labels进行向量匹配。

2、pushGateWay 的数据层面, pushgateway 的数据来源是推送到pgw，然后prometheus 再去拉取pgw对应的metric。所以pgw跟exporter就有个区别，pgw没有默认值。

 

 

3、那么在当上图中 short-lived jobs没有向pgw推送数据时，pgw不知道是否存在该监控向量。所以此时对应监控项用来判断应用的存活状态就多少有些麻烦。

4、贴上告警规则

groups:
- name: local_rules
  rules:

  - alert: InstanceDown
    expr: up{job="windows_prometheus"} == 0
    for: 2m
    labels:
      severity: error
    annotations:
      summary: "Instance {{ $labels.instance }} down"
      description: "{{ $labels.instance }} of job {{ $labels.job }} has been down for more than 5 minutes, current value: {{ $value }}"


  - alert: auto_wx_friend
    expr: (job_last_success_unixtime{exported_job="auto_wx_friend_from_pgw"} or (up{job="windows_prometheus"} * 0)) == 0
    for: 2m
    labels:
      severity: error
    annotations:
      summary: "auto_wx_friend down {{ $labels.src_instance }} {{  $labels.job }}"
      description: "{{ $labels.src_instance }} {{  $labels.job }}  (current value: {{ $value }})"

View Code

推送的脚本，可以直接用

from prometheus_client import CollectorRegistry, Gauge, push_to_gateway
import socket


class PushGateWayPrometheus:
    """
    pushgateway
    """

    def __init__(self):
        self.registry = CollectorRegistry()
        self.gateway = '192.168.60.203:9099'
        # label 和 value 对应
        self.label_name = ['src_instance', ]
        self.src_ip_label_value = socket.gethostbyname(socket.gethostname())
        # 无需修改
        self.job = 'auto_wx_friend_from_pgw'

    def gauge_process_alive(self, metric_name: str, describe: str) -> None:
        """
        如果对应值设置为1，则表示应用仍然存活
        :param metric_name: 
        :param describe: 
        :return: 
        """
        g = Gauge(metric_name, describe, registry=self.registry,
                  labelnames=self.label_name)
        g.labels(self.src_ip_label_value).set(1)

    def push(self, metric_name: str, describe: str) -> None:
        """
        推送对应的指标，如果有新的只需新增
        :param metric_name: 
        :param describe: 
        :return: 
        """
        self.gauge_process_alive(metric_name, describe)
        push_to_gateway(self.gateway, job=self.job, registry=self.registry)


PushGateWayPrometheus().push('job_last_success_unixtime', 'Last time a batch job successfully finished')

View Code

告警规则中，InstanceDown是exporter的，auto_wx_friend是pgw的。而在prometheus中的src_instance是为了尝试向量匹配，所以得有相同labels。

虽然简陋但也实现了效果，如果有大佬有更牛的可以赐教一番。

 

二、alertmanager 告警抑制

inhibit_rules:
  - source_match:
      severity: 'critical'
    target_match:
      severity: 'warning'
    equal: ['alertname', 'instance']

View Code

这是网上的，参考一下解释。

# 抑制器配置
inhibit_rules: # 抑制规则
  - source_match: # 源标签警报触发时抑制含有目标标签的警报，在当前警报匹配 status: 'High'
      status: 'High'  
    target_match:
      status: 'Warning' # 
    equal: ['alertname','operations', 'instance'] # 确保这个配置下的标签内容相同才会抑制，也就是说警报中必须有这三个标签值才会被抑制。

 

标签：name,遇到,self,labels,job,192.168,prometheus,9182,过程
From： https://www.cnblogs.com/bill2014/p/17227885.html