prometheus从零开始
本次的想法是做服务监控 并告警 主要线路如下图所示
1、运行prometheus docker方式
docker run -itd \
-p 9090:9090 \
-v /opt/prometheus/prometheus.yml:/etc/prometheus/prometheus.yml \
prom/prometheus
2、prometheus.yml 初始配置文件如下:
global:
scrape_interval: 15s # By default, scrape targets every 15 seconds. 全局默认值 15秒抓取一次数据 # Attach these labels to any time series or alerts when communicating with
# external systems (federation, remote storage, Alertmanager).
external_labels:
monitor: 'codelab-monitor' # A scrape configuration containing exactly one endpoint to scrape:
# Here it's Prometheus itself.
scrape_configs:
# The job name is added as a label `job=<job_name>` to any timeseries scraped from this config.
- job_name: 'prometheus' # Override the global default and scrape targets from this job every 5 seconds.
scrape_interval: 5s static_configs:
- targets: ['localhost:9090']
3、默认 prometheus 会有自己的指标接口http://192.168.246.2:9090/metrics 内容部分截取如下
# HELP go_gc_duration_seconds A summary of the pause duration of garbage collection cycles.
# TYPE go_gc_duration_seconds summary
go_gc_duration_seconds{quantile="0"} 2.6636e-05
go_gc_duration_seconds{quantile="0.25"} 0.000123346
go_gc_duration_seconds{quantile="0.5"} 0.000159706
go_gc_duration_seconds{quantile="0.75"} 0.000190857
go_gc_duration_seconds{quantile="1"} 0.001369042
4、可以登录9090端口去看看prometheus主界面 可以执行PromQL (Prometheus Query Language) 来excute得到结果
比如这个 prometheus_target_interval_length_seconds{quantile="0.99"}
具体PromQL语法示例请参考官网https://prometheus.io/docs/prometheus/latest/querying/basics/
5、上面的数据是prometheus自己的 ,下面我们自己生产数据给它 有很多公共的exporter 可以用 比如 node_exporter 他可以暴露机器一些基本的通用指标。
也可以执行python自定义编程 取指标 让自己成为一个exporter
安装node_exporter 官网例子 但是不要使用127.0.0.1 因为我的prometheus是docker起的 和宿主机的127.0.0.1是不通的 它抓取不到数据的,请改成实际的主机地址
ps:其他exporter 可参考地址 https://prometheus.io/docs/instrumenting/exporters/
tar -xzvf node_exporter-*.*.tar.gz
cd node_exporter-*.* # Start 3 example targets in separate terminals:
./node_exporter --web.listen-address 127.0.0.1:8080
./node_exporter --web.listen-address 127.0.0.1:8081
./node_exporter --web.listen-address 127.0.0.1:8082
6、需要修改prometheus.yml增加job 抓取exproter 修改后的如下 增加了 一个job 里面有三个exporter 标签是随便配的
global:
scrape_interval: 15s # By default, scrape targets every 15 seconds. # Attach these labels to any time series or alerts when communicating with
# external systems (federation, remote storage, Alertmanager).
external_labels:
monitor: 'codelab-monitor' # A scrape configuration containing exactly one endpoint to scrape:
# Here it's Prometheus itself.
scrape_configs:
# The job name is added as a label `job=<job_name>` to any timeseries scraped from this config.
- job_name: 'prometheus' # Override the global default and scrape targets from this job every 5 seconds.
scrape_interval: 5s static_configs:
- targets: ['localhost:9090']
- job_name: 'node' # Override the global default and scrape targets from this job every 5 seconds.
scrape_interval: 5s static_configs:
- targets: ['192.168.246.2:8080', '192.168.246.2:8081']
labels:
group: 'production' - targets: ['192.168.246.2:8082']
labels:
group: 'canary'
6、查看页面 是否ok了
7、可以看看node_exporter 暴露的指标例子 比如有如下的
node_cpu_seconds_total{cpu="0",mode="idle"} 2963.27
node_cpu_seconds_total{cpu="0",mode="iowait"} 0.38
node_cpu_seconds_total{cpu="0",mode="irq"} 0
node_cpu_seconds_total{cpu="0",mode="nice"} 0
node_cpu_seconds_total{cpu="0",mode="softirq"} 0.35
node_cpu_seconds_total{cpu="0",mode="steal"} 0
node_cpu_seconds_total{cpu="0",mode="system"} 19.19
node_cpu_seconds_total{cpu="0",mode="user"} 16.96
node_cpu_seconds_total{cpu="1",mode="idle"} 2965.47
node_cpu_seconds_total{cpu="1",mode="iowait"} 0.37
node_cpu_seconds_total{cpu="1",mode="irq"} 0
node_cpu_seconds_total{cpu="1",mode="nice"} 0.03
node_cpu_seconds_total{cpu="1",mode="softirq"} 0.28
node_cpu_seconds_total{cpu="1",mode="steal"} 0
node_cpu_seconds_total{cpu="1",mode="system"} 18.42
node_cpu_seconds_total{cpu="1",mode="user"} 17.95
8、如果我们想看 近5分钟内 每个实例的所有cpus的平均每秒CPU时间速率 可以这样写
avg by (job, instance, mode) (rate(node_cpu_seconds_total[5m]))
图例结果
9、下面设置一个rules规则,写一个文件 prometheus.rules.yml
groups:
- name: cpu-node
rules:
- record: job_instance_mode:node_cpu_seconds:avg_rate5m
expr: avg by (job, instance, mode) (rate(node_cpu_seconds_total[5m]))
10、现在发现配置文件太多了 我们重新用另一种方式启动docker 优化一下 把本地配置文件都放在 /opt/prometheus/,原来的docker可删除了。
--web.enable-lifecycle 参数支持热更新 接口是curl -X POST http://192.168.246.2:9090/-/reload
docker run -itd -p 9090:9090 -v /opt/prometheus/:/etc/prometheus/ prom/prometheus --config.file=/etc/prometheus/prometheus.yml --web.enable-lifecycle
11、查看rules
12、上面只是规则,并没有告警,我们假设 avg by (job, instance, mode) (rate(node_cpu_seconds_total[5m])) > 0.5 就触发 cpu警告 这是假设的测试
我们需要rules文件如下:
groups:
- name: example
rules:
- alert: HighCpuLatency
expr: avg by (job, instance, mode) (rate(node_cpu_seconds_total[5m])) > 0.5
for: 10s
labels:
severity: page
annotations:
summary: High request latency
写到prometheus.rules.yml文件中 ,注意groups不能复制进去 key不能重复 可以使用命令检查rules文件是否正确
[root@test prometheus]# ./prometheus-2.26.0.linux-amd64/promtool check rules prometheus.rules.yml
Checking prometheus.rules.yml
FAILED:
prometheus.rules.yml: yaml: unmarshal errors:
line 6: mapping key "groups" already defined at line 1
prometheus.rules.yml: yaml: unmarshal errors:
line 6: mapping key "groups" already defined at line 1 [root@test prometheus]# ./prometheus-2.26.0.linux-amd64/promtool check rules prometheus.rules.yml
Checking prometheus.rules.yml
SUCCESS: 2 rules found [root@test prometheus]#
13、热更新一下
curl -X POST http://192.168.246.2:9090/-/reload
这次不用重启docker了
查看页面 rules会增加一个 且alert会先有pending状态,等符合条件后就触发告警
14、下面启动alertmanager
启动之前要做两个配置,首先把 alertmanager 的IP和端口配置到prometheus.yml中
最会面增加了 alerting的配置 这样 prometheus 就连上 alertmanager
global:
scrape_interval: 15s # By default, scrape targets every 15 seconds. # Attach these labels to any time series or alerts when communicating with
# external systems (federation, remote storage, Alertmanager).
external_labels:
monitor: 'codelab-monitor' rule_files:
- 'prometheus.rules.yml'
# A scrape configuration containing exactly one endpoint to scrape:
# Here it's Prometheus itself.
scrape_configs:
# The job name is added as a label `job=<job_name>` to any timeseries scraped from this config.
- job_name: 'prometheus' # Override the global default and scrape targets from this job every 5 seconds.
scrape_interval: 5s static_configs:
- targets: ['localhost:9090']
- job_name: 'node' # Override the global default and scrape targets from this job every 5 seconds.
scrape_interval: 5s static_configs:
- targets: ['192.168.246.2:8080', '192.168.246.2:8081']
labels:
group: 'production' - targets: ['192.168.246.2:8082']
labels:
group: 'canary'
alerting:
alertmanagers:
- static_configs:
- targets: ["192.168.246.2:9093"]
第二个配置 我们先测试邮件告警,写的 alertmanager 配置如下
注意修改的部分 qq如何申请授权码请百度一下
smtp_smarthost: 'smtp.qq.com:465'
smtp_from: '6171391@qq.com'
smtp_auth_username: '6171391@qq.com'
smtp_auth_password: 'qq授权码'
smtp_require_tls: false
默认不做任何过滤选择的接收人
- to: 'dfwl@163.com'
global:
# The smarthost and SMTP sender used for mail notifications.
smtp_smarthost: 'smtp.qq.com:465'
smtp_from: '6171391@qq.com'
smtp_auth_username: '6171391@qq.com'
smtp_auth_password: 'qq授权码'
smtp_require_tls: false
# The directory from which notification templates are read.
templates:
- '/etc/alertmanager/template/*.tmpl' # The root route on which each incoming alert enters.
route: group_by: ['alertname', 'cluster', 'service'] group_wait: 30s # When the first notification was sent, wait 'group_interval' to send a batch
# of new alerts that started firing for that group.
group_interval: 1m # If an alert has successfully been sent, wait 'repeat_interval' to
# resend them.
repeat_interval: 7h # A default receiver
receiver: team-X-mails # The child route trees.
routes:
# This routes performs a regular expression match on alert labels to
# catch alerts that are related to a list of services.
- match_re:
service: ^(foo1|foo2|baz)$
receiver: team-X-mails
# The service has a sub-route for critical alerts, any alerts
# that do not match, i.e. severity != critical, fall-back to the
# parent node and are sent to 'team-X-mails'
routes:
- match:
severity: critical
receiver: team-X-pager
- match:
service: files
receiver: team-Y-mails routes:
- match:
severity: critical
receiver: team-Y-pager # This route handles all alerts coming from a database service. If there's
# no team to handle it, it defaults to the DB team.
- match:
service: database
receiver: team-DB-pager
# Also group alerts by affected database.
group_by: [alertname, cluster, database]
routes:
- match:
owner: team-X
receiver: team-X-pager
continue: true
- match:
owner: team-Y
receiver: team-Y-pager # Inhibition rules allow to mute a set of alerts given that another alert is
# firing.
# We use this to mute any warning-level notifications if the same alert is
# already critical.
inhibit_rules:
- source_match:
severity: 'critical'
target_match:
severity: 'warning'
# Apply inhibition if the alertname is the same.
# CAUTION:
# If all label names listed in `equal` are missing
# from both the source and target alerts,
# the inhibition rule will apply!
equal: ['alertname', 'cluster', 'service'] receivers:
- name: 'team-X-mails'
email_configs:
- to: 'dfwl@163.com' - name: 'team-X-pager'
email_configs:
- to: 'team-X+alerts-critical@example.org'
pagerduty_configs:
- service_key: <team-X-key> - name: 'team-Y-mails'
email_configs:
- to: 'team-Y+alerts@example.org' - name: 'team-Y-pager'
pagerduty_configs:
- service_key: <team-Y-key> - name: 'team-DB-pager'
pagerduty_configs:
- service_key: <team-DB-key>
15、手动启动测试一下 生产环境可以docker或k8s等方式启动
./alertmanager --config.file=alertmanager.yml
16、alertmanager页面能同步到告警
过一会就会发邮件了
group_interval: 1m 意思 从第一次接受告警1m后还在就发
prometheus从零开始的更多相关文章
- 从零开始搭建Prometheus自动监控报警系统
从零搭建Prometheus监控报警系统 什么是Prometheus? Prometheus是由SoundCloud开发的开源监控报警系统和时序列数据库(TSDB).Prometheus使用Go语言开 ...
- 从零开始学习Prometheus监控报警系统
Prometheus简介 Prometheus是一个开源的监控报警系统,它最初由SoundCloud开发. 2016年,Prometheus被纳入了由谷歌发起的Linux基金会旗下的云原生基金会( C ...
- Prometheus监控学习记录
官方文档 Prometheus基础文档 从零开始:Prometheus 进阶之路:Prometheus —— 技巧篇 进阶之路:Prometheus —— 理解篇 prometheus的数据类型介绍 ...
- 基于prometheus监控k8s集群
本文建立在你已经会安装prometheus服务的基础之上,如果你还不会安装,请参考:prometheus多维度监控容器 如果你还没有安装库k8s集群,情参考: 从零开始搭建基于calico的kuben ...
- kubernetes 1.15.1 高可用部署 -- 从零开始
这是一本书!!! 一本写我在容器生态圈的所学!!! 重点先知: 1. centos 7.6安装优化 2. k8s 1.15.1 高可用部署 3. 网络插件calico 4. dashboard 插件 ...
- 你必须知道的容器监控 (3) Prometheus
本篇已加入<.NET Core on K8S学习实践系列文章索引>,可以点击查看更多容器化技术相关系列文章.上一篇介绍了Google开发的容器监控工具cAdvisor,但是其提供的操作界面 ...
- 从零搭建Prometheus监控报警系统
什么是Prometheus? Prometheus是由SoundCloud开发的开源监控报警系统和时序列数据库(TSDB).Prometheus使用Go语言开发,是Google BorgMon监控系统 ...
- 3W字干货深入分析基于Micrometer和Prometheus实现度量和监控的方案
前提 最近线上的项目使用了spring-actuator做度量统计收集,使用Prometheus进行数据收集,Grafana进行数据展示,用于监控生成环境机器的性能指标和业务数据指标.一般,我们叫这样 ...
- Prometheus入门教程(二):Prometheus + Grafana实现可视化、告警
文章首发于[陈树义]公众号,点击跳转到原文:https://mp.weixin.qq.com/s/56S290p4j9KROB5uGRcGkQ Prometheus UI 提供了快速验证 PromQL ...
随机推荐
- 常见的六种容错机制:Fail-Over、Fail-Fast、Fail-Back、Fail-Safe,Forking 和 Broadcast
目录 1.Fail-Over:故障转移 2.Fail-Fast:快速失败 3.Fail-Back:失效自动恢复 4.Fail-Safe:失效安全 5.Forking:并行调用多个服务 6.Broadc ...
- S3C2440—10.代码重定位
文章目录 一.启动方式 1.1 NAND FLASH 启动 1.2 NOR FLASH 启动 二. 段的概念 2.1 重定位数据段 2.2 加载地址的引出 三.链接脚本 3.1 链接脚本的引入 3.2 ...
- 并发编程 Process 互斥锁
进程理论 程序与进程的区别 ''' 程序不是存在硬盘上的代码,相对来说是静态的 进程表示程序在执行的过程,是动态的 ''' 进程的调度 先来先服务调度算法 '''对长作业有利,对短作业无益''' 短作 ...
- Redis应用场景及缓存问题
1.应用场景 (1) 缓存 缓存机制几乎在所有的大型网站都有使用,合理地使用缓存不仅可以加快数据的访问速度,而且能够有效地降低后端数据源的压力.Redis 提供了键值过期时间设置,并且也提供了灵活 ...
- DLL-创建DLL
动态链接库的创建 fangyukuan 2010.6.20 编写DLL时的函数与一般的函数方法基本一样.但要对库中的函数进行必要的声明,以说明哪些函数是可以导出的,哪些函数是不可以导出的. 把DLL中 ...
- .Net Core 集成 Redis
首先安装RedisServer 安装教程可参照 http://www.redis.cn/download.html 或者 https://www.runoob.com/redis/redis-inst ...
- WPF 中TextBox 增加输入检测,错误提示
先来总结下实现错误提示功能的几个要点 1:binding 的ValidationRules 2 :Validation.ErrorTemplate 首先我们在界面添加一个TextBox, Text绑定 ...
- 关于腾讯云redis 无法外网访问的解决方案
问题简介: 今天购买了一台腾讯云的redis:如图 可是我没有找到 腾讯云提供的外网地址,我该怎么连接呢?百度了一大堆 全部是 在腾讯云服务器上搭建的Redis实例的解决办法.完全不匹配. 开始解决: ...
- mysql基础操作(三):数据约束
首先创建一个数据库 create database homework default character set utf8; use homework; 1.1 默认值约束(default) -- 数 ...
- 初识cookie
package day01.cookies; import java.io.IOException; import java.net.URLDecoder; import java.net.URLEn ...