第七模块 ：微服务监控告警Prometheus架构和实践

119.监控模式分类~1.mp4

logging：日志监控，Logging 的特点是，它描述一些离散的(不连续的)事件。 例如：应用通过一个滚动的文件输出 Debug 或 Error 信息，并通过日志收集系统，存储到 Elasticsearch 中; 审批明细信息通过 Kafka，存储到数据库(BigTable)中; 又或者，特定请求的元数据信息，从服务请求中剥离出来，发送给一个异常收集服务，如 NewRelic。

tracing：链路追踪 ，例如skywalking、cat、zipkin专门做分布式链路追踪

metrics：重要的数据指标、例如统计当前http的一个请求量，数据是可以度量、累加的，数据是可以聚合和累加的。通过对数据点进行聚合查看重要的指标，里面包括一些计数器、测量仪器、直方图等，还可以

在上面打标签

• Metrics - 用于记录可聚合的数据。例如，队列的当前深度可被定义为一个度量值，在元素入队或出队时被更新；HTTP 请求个数可被定义为一个计数器，新请求到来时进行累加。

普罗米修斯重点属于metrics的监控

上面重点讲下几个工具部署的成本：

capEx表示研发人员开发的成本：metrics研发人员开发成功重点，elk最低，掌握skywalking需要一定的基础

OpEx表示运维成本，elk在运维的时候需要不断扩容，运行成功最高

reaction表示当出问题的时候哪些功能能够第一时间进行告警，metries工具最高。skywalking的告警能力一般

出了问题具体分析问题哪些工具最有效：查看调用链解决问题最有效

Logging，Metrics 和 Tracing

Logging，Metrics 和 Tracing 有各自专注的部分。

• Logging - 用于记录离散的事件。例如，应用程序的调试信息或错误信息。它是我们诊断问题的依据。
• Metrics - 用于记录可聚合的数据。例如，队列的当前深度可被定义为一个度量值，在元素入队或出队时被更新；HTTP 请求个数可被定义为一个计数器，新请求到来时进行累加。
• Tracing - 用于记录请求范围内的信息。例如，一次远程方法调用的执行过程和耗时。它是我们排查系统性能问题的利器。

这三者也有相互重叠的部分，如下图所示。

通过上述信息，我们可以对已有系统进行分类。例如，Zipkin 专注于 tracing 领域；Prometheus 开始专注于 metrics，随着时间推移可能会集成更多的 tracing 功能，但不太可能深入 logging 领域； ELK，阿里云日志服务这样的系统开始专注于 logging 领域，但同时也不断地集成其他领域的特性到系统中来，正向上图中的圆心靠近。

三者关系的一篇论文：http://peter.bourgon.org/blog/2017/02/21/metrics-tracing-and-logging.html

关于三者关系的更详细信息可参考 Metrics, tracing, and logging。下面我们重点介绍下 tracing。

Tracing 的诞生

Tracing 是在90年代就已出现的技术。但真正让该领域流行起来的还是源于 Google 的一篇论文"Dapper, a Large-Scale Distributed Systems Tracing Infrastructure"，而另一篇论文"Uncertainty in Aggregate Estimates from Sampled Distributed Traces"中则包含关于采样的更详细分析。论文发表后一批优秀的 Tracing 软件孕育而生，比较流行的有：

• Dapper(Google) : 各 tracer 的基础
• StackDriver Trace (Google)
• Zipkin(twitter)
• Appdash(golang)
• 鹰眼(taobao)
• 谛听(盘古，阿里云云产品使用的Trace系统)
• 云图(蚂蚁Trace系统)
• sTrace(神马)
• X-ray(aws)

分布式追踪系统发展很快，种类繁多，但核心步骤一般有三个：代码埋点，数据存储、查询展示。

下图是一个分布式调用的例子，客户端发起请求，请求首先到达负载均衡器，接着经过认证服务，计费服务，然后请求资源，最后返回结果。

数据被采集存储后，分布式追踪系统一般会选择使用包含时间轴的时序图来呈现这个 Trace。

但在数据采集过程中，由于需要侵入用户代码，并且不同系统的 API 并不兼容，这就导致了如果您希望切换追踪系统，往往会带来较大改动。

OpenTracing

为了解决不同的分布式追踪系统 API 不兼容的问题，诞生了 OpenTracing 规范。
OpenTracing 是一个轻量级的标准化层，它位于应用程序/类库和追踪或日志分析程序之间。

+-------------+  +---------+  +----------+  +------------+
| Application |  | Library |  |   OSS    |  |  RPC/IPC   |
|    Code     |  |  Code   |  | Services |  | Frameworks |
+-------------+  +---------+  +----------+  +------------+
       |              |             |             |
       |              |             |             |
       v              v             v             v
  +------------------------------------------------------+
  |                     OpenTracing                      |
  +------------------------------------------------------+
     |                |                |               |
     |                |                |               |
     v                v                v               v
+-----------+  +-------------+  +-------------+  +-----------+
|  Tracing  |  |   Logging   |  |   Metrics   |  |  Tracing  |
| System A  |  | Framework B |  | Framework C |  | System D  |
+-----------+  +-------------+  +-------------+  +-----------+

OpenTracing 的优势

• OpenTracing 已进入 CNCF，正在为全球的分布式追踪，提供统一的概念和数据标准。
• OpenTracing 通过提供平台无关、厂商无关的 API，使得开发人员能够方便的添加（或更换）追踪系统的实现。

OpenTracing 数据模型

OpenTracing 中的 Trace（调用链）通过归属于此调用链的 Span 来隐性的定义。
特别说明，一条 Trace（调用链）可以被认为是一个由多个 Span 组成的有向无环图（DAG图），Span 与 Span 的关系被命名为 References。

例如：下面的示例 Trace 就是由8个 Span 组成：

单个 Trace 中，span 间的因果关系

        [Span A]  ←←←(the root span)
            |
     +------+------+
     |             |
 [Span B]      [Span C] ←←←(Span C 是 Span A 的孩子节点, ChildOf)
     |             |
 [Span D]      +---+-------+
               |           |
           [Span E]    [Span F] >>> [Span G] >>> [Span H]
                                       ↑
                                       ↑
                                       ↑
                         (Span G 在 Span F 后被调用, FollowsFrom)

有些时候，使用下面这种，基于时间轴的时序图可以更好的展现 Trace（调用链）：

单个 Trace 中，span 间的时间关系

––|–––––––|–––––––|–––––––|–––––––|–––––––|–––––––|–––––––|–> time

 [Span A···················································]
   [Span B··············································]
      [Span D··········································]
    [Span C········································]
         [Span E·······]        [Span F··] [Span G··] [Span H··]

每个 Span 包含以下的状态:（译者注：由于这些状态会反映在 OpenTracing API 中，所以会保留部分英文说明）

• An operation name，操作名称
• A start timestamp，起始时间
• A finish timestamp，结束时间
• Span Tag，一组键值对构成的 Span 标签集合。键值对中，键必须为 string，值可以是字符串，布尔，或者数字类型。
• Span Log，一组 span 的日志集合。
  每次 log 操作包含一个键值对，以及一个时间戳。

键值对中，键必须为 string，值可以是任意类型。
但是需要注意，不是所有的支持 OpenTracing 的 Tracer，都需要支持所有的值类型。

• SpanContext，Span 上下文对象 (下面会详细说明)
• References(Span间关系)，相关的零个或者多个 Span（Span 间通过 SpanContext 建立这种关系）

每一个 SpanContext 包含以下状态：

• 任何一个 OpenTracing 的实现，都需要将当前调用链的状态（例如：trace 和 span 的 id），依赖一个独特的 Span 去跨进程边界传输
• Baggage Items，Trace 的随行数据，是一个键值对集合，它存在于 trace 中，也需要跨进程边界传输

更多关于 OpenTracing 数据模型的知识，请参考 OpenTracing语义标准。

OpenTracing 实现

这篇文档列出了所有 OpenTracing 实现。在这些实现中，比较流行的为 Jaeger 和 Zipkin。

metrs主要用于监控告警，出了问题之后再通过链路追踪或者elk来查看发现问题解决问题

metries可以对系统层、应用层、业务层进行监控

121.Prometheus 简介~1.mp4

时间数据库

在t0产生数据v0，在t1产生数据v1，在t2产生数据v2，将这些数据点弄起来就是一个时间序列

infuxdb和普罗米修斯都是时间序列数据库

122.Prometheus 架构设计~1.mp4

123.Prometheus 基本概念~1.mp4

第一种计数器：例如统计的http数目、下单数目等

测量仪器：例如当前同时在线用户数目。磁盘使用率

直方图：响应时间在某个区间内的分布情况

汇总：90%的响应时间

target：可以是操作系统、机器、应用、服务等需要暴露metrries端点，每隔15秒中通过/metries抓取数据

应用直接采用：应用直接采集，直接在应用程序中埋点，直接使用普罗米修斯采集

第二种间接采集使用，使用exporter采集，执行的exporter如下所示，redis、apache、操作系统等

web ui和grafan通过promql就可以来查询对于的数据

124.【实验】Prometheus 起步查询实验（上）~1.mp4

第一步

将http-simulatror导入到eclipse中

springboot集成prometheus

Maven pom.xml引入依赖

<dependency>
<groupId>io.prometheus</groupId>
<artifactId>simpleclient_spring_boot</artifactId>
</dependency>

2 启动类引入注解

import io.prometheus.client.spring.boot.EnablePrometheusEndpoint;
import io.prometheus.client.spring.boot.EnableSpringBootMetricsCollector;
import org.springframework.boot.SpringApplication;
import org.springframework.boot.autoconfigure.SpringBootApplication;

@SpringBootApplication
@EnablePrometheusEndpoint
@EnableSpringBootMetricsCollector
public class Application {

public static void main(String[] args) {
SpringApplication.run(Application.class, args);
}
3 Controller类写需要监控的指标，比如Counter

import io.prometheus.client.Counter;
import org.springframework.web.bind.annotation.RequestMapping;
import org.springframework.web.bind.annotation.RestController;

import java.util.Random;

@RestController
public class SampleController {

    private static Random random = new Random();

    private static final Counter requestTotal = Counter.build()
        .name("my_sample_counter")
        .labelNames("status")
        .help("A simple Counter to illustrate custom Counters in Spring Boot and Prometheus").register();

    @RequestMapping("/endpoint")
    public void endpoint() {
        if (random.nextInt() > ) {
            requestTotal.labels("success").inc();
        } else {
            requestTotal.labels("error").inc();
        }
    }
}

4 设置springboot应用的服务名和端口，在application.properties

spring.application.name=mydemo
server.port=

5 配置prometheus.yml

global:
  scrape_interval:     15s # By default, scrape targets every  seconds.
  evaluation_interval: 15s # By default, scrape targets every  seconds.
  # scrape_timeout is set to the global default (10s).

  # Attach these labels to any time series or alerts when communicating with
  # external systems (federation, remote storage, Alertmanager).
  external_labels:
      monitor: 'codelab-monitor'

# Load rules once and periodically evaluate them according to the global 'evaluation_interval'.
rule_files:
  # - "first.rules"
  # - "second.rules"

# 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  seconds.
    scrape_interval: 5s

    # metrics_path defaults to '/metrics'
    # scheme defaults to 'http'.

    static_configs:
      - targets: ['localhost:9090']

  - job_name: 'mydemo'

    # Override the global default and scrape targets from this job every  seconds.
    scrape_interval: 5s

    metrics_path: '/prometheus'
    # scheme defaults to 'http'.

    static_configs:
      - targets: ['10.94.20.52:8888']

最关键的配置就是targets: [‘10.94.20.52:8888’]，就是springboot应用的ip和端口

注：在application.xml里设置属性：spring.metrics.servo.enabled=false，去掉重复的metrics，不然在prometheus的控制台的targets页签里，会一直显示此endpoint为down状态。

pom.xml

<?xml version="1.0" encoding="UTF-8"?>
<project xmlns="http://maven.apache.org/POM/4.0.0" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
    xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd">
    <modelVersion>4.0.</modelVersion>

    <groupId>io.spring2go.promdemo</groupId>
    <artifactId>http-simulator</artifactId>
    <version>0.0.-SNAPSHOT</version>
    <packaging>jar</packaging>

    <name>http-simulator</name>
    <description>Demo project for Spring Boot</description>

    <parent>
        <groupId>org.springframework.boot</groupId>
        <artifactId>spring-boot-starter-parent</artifactId>
        <version>1.5..RELEASE</version>
        <relativePath/> <!-- lookup parent from repository -->
    </parent>

    <properties>
        <project.build.sourceEncoding>UTF-</project.build.sourceEncoding>
        <project.reporting.outputEncoding>UTF-</project.reporting.outputEncoding>
        <java.version>1.8</java.version>
    </properties>

    <dependencies>
        <dependency>
            <groupId>org.springframework.boot</groupId>
            <artifactId>spring-boot-starter-web</artifactId>
        </dependency>

        <!-- The prometheus client -->
        <dependency>
            <groupId>io.prometheus</groupId>
            <artifactId>simpleclient_spring_boot</artifactId>
            <version>0.5.</version>
        </dependency>

        <dependency>
            <groupId>org.springframework.boot</groupId>
            <artifactId>spring-boot-starter-test</artifactId>
            <scope>test</scope>
        </dependency>
    </dependencies>

    <build>
        <plugins>
            <plugin>
                <groupId>org.springframework.boot</groupId>
                <artifactId>spring-boot-maven-plugin</artifactId>
            </plugin>
        </plugins>
    </build>

</project>

启动类：

package io.spring2go.promdemo.httpsimulator;

import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.boot.CommandLineRunner;
import org.springframework.boot.SpringApplication;
import org.springframework.boot.autoconfigure.SpringBootApplication;
import org.springframework.context.ApplicationListener;
import org.springframework.context.annotation.Bean;
import org.springframework.context.event.ContextClosedEvent;
import org.springframework.core.task.SimpleAsyncTaskExecutor;
import org.springframework.core.task.TaskExecutor;
import org.springframework.stereotype.Controller;
import org.springframework.web.bind.annotation.PathVariable;
import org.springframework.web.bind.annotation.RequestMapping;
import org.springframework.web.bind.annotation.RequestMethod;
import org.springframework.web.bind.annotation.ResponseBody;

import io.prometheus.client.spring.boot.EnablePrometheusEndpoint;

@Controller
@SpringBootApplication
@EnablePrometheusEndpoint
public class HttpSimulatorApplication implements ApplicationListener<ContextClosedEvent> {

    @Autowired
    private SimulatorOpts opts;

    private ActivitySimulator simulator;

    public static void main(String[] args) {

        SpringApplication.run(HttpSimulatorApplication.class, args);
    }

    @RequestMapping(value = "/opts")
    public @ResponseBody String getOps() {
        return opts.toString();
    }

    @RequestMapping(value = "/spike/{mode}", method = RequestMethod.POST)
    public @ResponseBody String setSpikeMode(@PathVariable("mode") String mode) {
        boolean result = simulator.setSpikeMode(mode);
        if (result) {
            return "ok";
        } else {
            return "wrong spike mode " + mode;
        }
    }

    @RequestMapping(value = "error_rate/{error_rate}", method = RequestMethod.POST)
    public @ResponseBody String setErrorRate(@PathVariable("error_rate") int errorRate) {
        simulator.setErrorRate(errorRate);
        return "ok";
    }

    @Bean
    public TaskExecutor taskExecutor() {
        return new SimpleAsyncTaskExecutor();
    }

    @Bean
    public CommandLineRunner schedulingRunner(TaskExecutor executor) {
        return new CommandLineRunner() {
            public void run(String... args) throws Exception {
                simulator = new ActivitySimulator(opts);
                executor.execute(simulator);
                System.out.println("Simulator thread started...");
            }
        };
    }

    @Override
    public void onApplicationEvent(ContextClosedEvent event) {
        simulator.shutdown();
        System.out.println("Simulator shutdown...");
    }

}

application.properties

management.security.enabled=false

opts.endpoints=/login, /login, /login, /login, /login, /login, /login, /users, /users, /users, /users/{id}, /register, /register, /logout, /logout, /logout, /logout
opts.request_rate=
opts.request_rate_uncertainty=
opts.latency_min=
opts.latency_p50=
opts.latency_p90=
opts.latency_p99=
opts.latency_max=
opts.latency_uncertainty=

opts.error_rate=
opts.spike_start_chance=
opts.spike_end_chance=

最关键的核心类

ActivitySimulator

package io.spring2go.promdemo.httpsimulator;

import java.util.Random;

import io.prometheus.client.Counter;
import io.prometheus.client.Histogram;

public class ActivitySimulator implements Runnable {

    private SimulatorOpts opts;

    private Random rand = new Random();

    private boolean spikeMode = false;

    private volatile boolean shutdown = false;

    private final Counter httpRequestsTotal = Counter.build()
            .name("http_requests_total")
            .help("Total number of http requests by response status code")
            .labelNames("endpoint", "status")
            .register();

    private final Histogram httpRequestDurationMs = Histogram.build()
            .name("http_request_duration_milliseconds")
            .help("Http request latency histogram")
            .exponentialBuckets(, , )
            .labelNames("endpoint", "status")
            .register();

    public ActivitySimulator(SimulatorOpts opts) {
        this.opts = opts;
        System.out.println(opts);
    }

    public void shutdown() {
        this.shutdown = true;
    }

    public void updateOpts(SimulatorOpts opts) {
        this.opts = opts;
    }

    public boolean setSpikeMode(String mode) {
        boolean result = true;
        switch (mode) {
        case "on":
            opts.setSpikeMode(SpikeMode.ON);
            System.out.println("Spike mode is set to " + mode);
            break;
        case "off":
            opts.setSpikeMode(SpikeMode.OFF);
            System.out.println("Spike mode is set to " + mode);
            break;
        case "random":
            opts.setSpikeMode(SpikeMode.RANDOM);
            System.out.println("Spike mode is set to " + mode);
            break;
        default:
            result = false;
            System.out.println("Can't recognize spike mode " + mode);
        }
        return result;
    }

    public void setErrorRate(int rate) {
        if (rate > ) {
            rate = ;
        }
        if (rate < ) {
            rate = ;
        }
        opts.setErrorRate(rate);
        System.out.println("Error rate is set to " + rate);
    }

    public SimulatorOpts getOpts() {
        return this.opts;
    }

    public void simulateActivity() {
        int requestRate = this.opts.getRequestRate();
        if (this.giveSpikeMode()) {
            requestRate *= ( + this.rand.nextInt());
        }

        int nbRequests = this.giveWithUncertainty(requestRate, this.opts.getRequestRateUncertainty());
        for (int i = ; i < nbRequests; i++) {
            String statusCode = this.giveStatusCode();
            String endpoint = this.giveEndpoint();
            this.httpRequestsTotal.labels(endpoint, statusCode).inc();
            int latency = this.giveLatency(statusCode);
            if (this.spikeMode) {
                latency *= ;
            }
            this.httpRequestDurationMs.labels(endpoint, statusCode).observe(latency);
        }
    }

    public boolean giveSpikeMode() {
        switch (this.opts.getSpikeMode()) {
        case ON:
            this.spikeMode = true;
            break;
        case OFF:
            this.spikeMode = false;
            break;
        case RANDOM:
            int n = rand.nextInt();
            if (!this.spikeMode && n < this.opts.getSpikeStartChance()) {
                this.spikeMode = true;
            } else if (this.spikeMode && n < this.opts.getSpikeEndChance()) {
                this.spikeMode = false;
            }
            break;
        }

        return this.spikeMode;
    }

    public int giveWithUncertainty(int n, int u) {
        int delta = this.rand.nextInt(n * u / ) - (n * u / );
        return n + delta;
    }

    public String giveStatusCode() {
        if (this.rand.nextInt() < this.opts.getErrorRate()) {
            return "";
        } else {
            return "";
        }
    }

    public String giveEndpoint() {
        int n = this.rand.nextInt(this.opts.getEndopints().length);
        return this.opts.getEndopints()[n];
    }

    public int giveLatency(String statusCode) {
        if (!"".equals(statusCode)) {
            return  + this.rand.nextInt();
        }

        int p = this.rand.nextInt();

        if (p < ) {
            return this.giveWithUncertainty(this.opts.getLatencyMin() + this.rand.nextInt(this.opts.getLatencyP50() - this.opts.getLatencyMin()), this.opts.getLatencyUncertainty());
        }
        if (p < ) {
            return this.giveWithUncertainty(this.opts.getLatencyP50() + this.rand.nextInt(this.opts.getLatencyP90() - this.opts.getLatencyP50()), this.opts.getLatencyUncertainty());
        }
        if (p < ) {
            return this.giveWithUncertainty(this.opts.getLatencyP90() + this.rand.nextInt(this.opts.getLatencyP99() - this.opts.getLatencyP90()), this.opts.getLatencyUncertainty());
        }

        return this.giveWithUncertainty(this.opts.getLatencyP99() + this.rand.nextInt(this.opts.getLatencyMax() - this.opts.getLatencyP99()), this.opts.getLatencyUncertainty());
    }

    @Override
    public void run() {
        while(!shutdown) {
            System.out.println("Simulator is running...");
            this.simulateActivity();
            try {
                Thread.sleep();
            } catch (InterruptedException e) {
                // TODO Auto-generated catch block
                e.printStackTrace();
            }
        }
    }
}

SimulatorOpts

package io.spring2go.promdemo.httpsimulator;

import java.util.Arrays;

import org.springframework.beans.factory.annotation.Value;
import org.springframework.context.annotation.Configuration;

import com.fasterxml.jackson.annotation.JsonAutoDetect;

@Configuration
@JsonAutoDetect(fieldVisibility = JsonAutoDetect.Visibility.ANY)
public class SimulatorOpts {

    // Endpoints, Weighted map of endpoints to simulate
    @Value("${opts.endpoints}")
    private String[] endopints;

    // RequestRate, requests per second
    @Value("${opts.request_rate}")
    private int requestRate;

    // RequestRateUncertainty, Percentage of uncertainty when generating requests (+/-)
    @Value("${opts.request_rate_uncertainty}")
    private int requestRateUncertainty;

    // LatencyMin in milliseconds
    @Value("${opts.latency_min}")
    private int latencyMin;

    // LatencyP50 in milliseconds
    @Value("${opts.latency_p50}")
    private int latencyP50;

    // LatencyP90 in milliseconds
    @Value("${opts.latency_p90}")
    private int latencyP90;

    // LatencyP99 in milliseconds
    @Value("${opts.latency_p99}")
    private int latencyP99;

    // LatencyMax in milliseconds
    @Value("${opts.latency_max}")
    private int latencyMax;

    // LatencyUncertainty, Percentage of uncertainty when generating latency (+/-)
    @Value("${opts.latency_uncertainty}")
    private int latencyUncertainty;

    // ErrorRate, Percentage of chance of requests causing 500
    @Value("${opts.error_rate}")
    private int errorRate;

    // SpikeStartChance, Percentage of chance of entering spike mode
    @Value("${opts.spike_start_chance}")
    private int spikeStartChance;

    // SpikeStartChance, Percentage of chance of exiting spike mode
    @Value("${opts.spike_end_chance}")
    private int spikeEndChance;

    // SpikeModeStatus ON/OFF/RANDOM
    private SpikeMode spikeMode = SpikeMode.OFF;

    public String[] getEndopints() {
        return endopints;
    }

    public void setEndopints(String[] endopints) {
        this.endopints = endopints;
    }

    public int getRequestRate() {
        return requestRate;
    }

    public void setRequestRate(int requestRate) {
        this.requestRate = requestRate;
    }

    public int getRequestRateUncertainty() {
        return requestRateUncertainty;
    }

    public void setRequestRateUncertainty(int requestRateUncertainty) {
        this.requestRateUncertainty = requestRateUncertainty;
    }

    public int getLatencyMin() {
        return latencyMin;
    }

    public void setLatencyMin(int latencyMin) {
        this.latencyMin = latencyMin;
    }

    public int getLatencyP50() {
        return latencyP50;
    }

    public void setLatencyP50(int latencyP50) {
        this.latencyP50 = latencyP50;
    }

    public int getLatencyP90() {
        return latencyP90;
    }

    public void setLatencyP90(int latencyP90) {
        this.latencyP90 = latencyP90;
    }

    public int getLatencyP99() {
        return latencyP99;
    }

    public void setLatencyP99(int latencyP99) {
        this.latencyP99 = latencyP99;
    }

    public int getLatencyMax() {
        return latencyMax;
    }

    public void setLatencyMax(int latencyMax) {
        this.latencyMax = latencyMax;
    }

    public int getLatencyUncertainty() {
        return latencyUncertainty;
    }

    public void setLatencyUncertainty(int latencyUncertainty) {
        this.latencyUncertainty = latencyUncertainty;
    }

    public int getErrorRate() {
        return errorRate;
    }

    public void setErrorRate(int errorRate) {
        this.errorRate = errorRate;
    }

    public int getSpikeStartChance() {
        return spikeStartChance;
    }

    public void setSpikeStartChance(int spikeStartChance) {
        this.spikeStartChance = spikeStartChance;
    }

    public int getSpikeEndChance() {
        return spikeEndChance;
    }

    public void setSpikeEndChance(int spikeEndChance) {
        this.spikeEndChance = spikeEndChance;
    }

    public SpikeMode getSpikeMode() {
        return spikeMode;
    }

    public void setSpikeMode(SpikeMode spikeMode) {
        this.spikeMode = spikeMode;
    }

    @Override
    public String toString() {
        return "SimulatorOpts [endopints=" + Arrays.toString(endopints) + ", requestRate=" + requestRate
                + ", requestRateUncertainty=" + requestRateUncertainty + ", latencyMin=" + latencyMin + ", latencyP50="
                + latencyP50 + ", latencyP90=" + latencyP90 + ", latencyP99=" + latencyP99 + ", latencyMax="
                + latencyMax + ", latencyUncertainty=" + latencyUncertainty + ", errorRate=" + errorRate
                + ", spikeStartChance=" + spikeStartChance + ", spikeEndChance=" + spikeEndChance + ", spikeMode="
                + spikeMode + "]";
    }

}

SpikeMode

package io.spring2go.promdemo.httpsimulator;

public enum SpikeMode {

    OFF, ON, RANDOM

}

我们将应用运行起来

模拟一个简单的HTTP微服务，生成Prometheus Metrics，可以Spring Boot方式运行

Metrics

运行时访问端点：

http://SERVICE_URL:8080/prometheus

包括：

• http_requests_total：请求计数器，endpoint和status为label
• http_request_duration_milliseconds：请求延迟分布(histogram)

运行时options

Spike Mode

在Spike模式下，请求数会乘以一个因子(5~15)，延迟加倍

Spike模式可以是on, off或者random, 改变方式：

# ON
curl -X POST http://SERVICE_URL:8080/spike/on

# OFF
curl -X POST http://SERVICE_URL:8080/spike/off

# RANDOM
curl -X POST http://SERVICE_URL:8080/spike/random

Error rate

缺省错误率1%，可以调整(0~100)，方法：

# Setting error to 50%
curl -X POST http://SERVICE_URL:8080/error_rate/50

其它参数

配置在application.properties中

opts.endpoints=/login, /login, /login, /login, /login, /login, /login, /users, /users, /users, /users/{id}, /register, /register, /logout, /logout, /logout, /logout
opts.request_rate=1000
opts.request_rate_uncertainty=70
opts.latency_min=10
opts.latency_p50=25
opts.latency_p90=150
opts.latency_p99=750
opts.latency_max=10000
opts.latency_uncertainty=70

opts.error_rate=1
opts.spike_start_chance=5
opts.spike_end_chance=30

运行时校验端点：

http://SERVICE_URL:8080/opts

参考

https://github.com/PierreVincent/prom-http-simulator

我们在浏览器输入http://localhost:8080/prometheus，我们可以查看收集到了metrie信息

# HELP http_requests_total Total number of http requests by response status code
# TYPE http_requests_total counter
http_requests_total{endpoint="/login",status="",} 188.0
http_requests_total{endpoint="/register",status="",} 55.0
http_requests_total{endpoint="/login",status="",} 18863.0
http_requests_total{endpoint="/register",status="",} 5425.0
http_requests_total{endpoint="/users/{id}",status="",} 26.0
http_requests_total{endpoint="/users/{id}",status="",} 2663.0
http_requests_total{endpoint="/logout",status="",} 10722.0
http_requests_total{endpoint="/users",status="",} 8034.0
http_requests_total{endpoint="/users",status="",} 94.0
http_requests_total{endpoint="/logout",status="",} 93.0
# HELP http_request_duration_milliseconds Http request latency histogram
# TYPE http_request_duration_milliseconds histogram
http_request_duration_milliseconds_bucket{endpoint="/login",status="",le="25.0",} 85.0
http_request_duration_milliseconds_bucket{endpoint="/login",status="",le="50.0",} 174.0
http_request_duration_milliseconds_bucket{endpoint="/login",status="",le="100.0",} 188.0
http_request_duration_milliseconds_bucket{endpoint="/login",status="",le="200.0",} 188.0
http_request_duration_milliseconds_bucket{endpoint="/login",status="",le="400.0",} 188.0
http_request_duration_milliseconds_bucket{endpoint="/login",status="",le="800.0",} 188.0
http_request_duration_milliseconds_bucket{endpoint="/login",status="",le="1600.0",} 188.0
http_request_duration_milliseconds_bucket{endpoint="/login",status="",le="+Inf",} 188.0
http_request_duration_milliseconds_count{endpoint="/login",status="",} 188.0
http_request_duration_milliseconds_sum{endpoint="/login",status="",} 5499.0
http_request_duration_milliseconds_bucket{endpoint="/register",status="",le="25.0",} 27.0
http_request_duration_milliseconds_bucket{endpoint="/register",status="",le="50.0",} 50.0
http_request_duration_milliseconds_bucket{endpoint="/register",status="",le="100.0",} 55.0
http_request_duration_milliseconds_bucket{endpoint="/register",status="",le="200.0",} 55.0
http_request_duration_milliseconds_bucket{endpoint="/register",status="",le="400.0",} 55.0
http_request_duration_milliseconds_bucket{endpoint="/register",status="",le="800.0",} 55.0
http_request_duration_milliseconds_bucket{endpoint="/register",status="",le="1600.0",} 55.0
http_request_duration_milliseconds_bucket{endpoint="/register",status="",le="+Inf",} 55.0
http_request_duration_milliseconds_count{endpoint="/register",status="",} 55.0
http_request_duration_milliseconds_sum{endpoint="/register",status="",} 1542.0
http_request_duration_milliseconds_bucket{endpoint="/login",status="",le="25.0",} 8479.0
http_request_duration_milliseconds_bucket{endpoint="/login",status="",le="50.0",} 11739.0
http_request_duration_milliseconds_bucket{endpoint="/login",status="",le="100.0",} 14454.0
http_request_duration_milliseconds_bucket{endpoint="/login",status="",le="200.0",} 17046.0
http_request_duration_milliseconds_bucket{endpoint="/login",status="",le="400.0",} 17882.0
http_request_duration_milliseconds_bucket{endpoint="/login",status="",le="800.0",} 18482.0
http_request_duration_milliseconds_bucket{endpoint="/login",status="",le="1600.0",} 18705.0
http_request_duration_milliseconds_bucket{endpoint="/login",status="",le="+Inf",} 18863.0
http_request_duration_milliseconds_count{endpoint="/login",status="",} 18863.0
http_request_duration_milliseconds_sum{endpoint="/login",status="",} 2552014.0
http_request_duration_milliseconds_bucket{endpoint="/register",status="",le="25.0",} 2388.0
http_request_duration_milliseconds_bucket{endpoint="/register",status="",le="50.0",} 3367.0
http_request_duration_milliseconds_bucket{endpoint="/register",status="",le="100.0",} 4117.0
http_request_duration_milliseconds_bucket{endpoint="/register",status="",le="200.0",} 4889.0
http_request_duration_milliseconds_bucket{endpoint="/register",status="",le="400.0",} 5136.0
http_request_duration_milliseconds_bucket{endpoint="/register",status="",le="800.0",} 5310.0
http_request_duration_milliseconds_bucket{endpoint="/register",status="",le="1600.0",} 5379.0
http_request_duration_milliseconds_bucket{endpoint="/register",status="",le="+Inf",} 5425.0
http_request_duration_milliseconds_count{endpoint="/register",status="",} 5425.0
http_request_duration_milliseconds_sum{endpoint="/register",status="",} 739394.0
http_request_duration_milliseconds_bucket{endpoint="/users/{id}",status="",le="25.0",} 14.0
http_request_duration_milliseconds_bucket{endpoint="/users/{id}",status="",le="50.0",} 25.0
http_request_duration_milliseconds_bucket{endpoint="/users/{id}",status="",le="100.0",} 26.0
http_request_duration_milliseconds_bucket{endpoint="/users/{id}",status="",le="200.0",} 26.0
http_request_duration_milliseconds_bucket{endpoint="/users/{id}",status="",le="400.0",} 26.0
http_request_duration_milliseconds_bucket{endpoint="/users/{id}",status="",le="800.0",} 26.0
http_request_duration_milliseconds_bucket{endpoint="/users/{id}",status="",le="1600.0",} 26.0
http_request_duration_milliseconds_bucket{endpoint="/users/{id}",status="",le="+Inf",} 26.0
http_request_duration_milliseconds_count{endpoint="/users/{id}",status="",} 26.0
http_request_duration_milliseconds_sum{endpoint="/users/{id}",status="",} 752.0
http_request_duration_milliseconds_bucket{endpoint="/users/{id}",status="",le="25.0",} 1220.0
http_request_duration_milliseconds_bucket{endpoint="/users/{id}",status="",le="50.0",} 1657.0
http_request_duration_milliseconds_bucket{endpoint="/users/{id}",status="",le="100.0",} 2030.0
http_request_duration_milliseconds_bucket{endpoint="/users/{id}",status="",le="200.0",} 2383.0
http_request_duration_milliseconds_bucket{endpoint="/users/{id}",status="",le="400.0",} 2508.0
http_request_duration_milliseconds_bucket{endpoint="/users/{id}",status="",le="800.0",} 2608.0
http_request_duration_milliseconds_bucket{endpoint="/users/{id}",status="",le="1600.0",} 2637.0
http_request_duration_milliseconds_bucket{endpoint="/users/{id}",status="",le="+Inf",} 2663.0
http_request_duration_milliseconds_count{endpoint="/users/{id}",status="",} 2663.0
http_request_duration_milliseconds_sum{endpoint="/users/{id}",status="",} 402375.0
http_request_duration_milliseconds_bucket{endpoint="/logout",status="",le="25.0",} 4790.0
http_request_duration_milliseconds_bucket{endpoint="/logout",status="",le="50.0",} 6634.0
http_request_duration_milliseconds_bucket{endpoint="/logout",status="",le="100.0",} 8155.0
http_request_duration_milliseconds_bucket{endpoint="/logout",status="",le="200.0",} 9609.0
http_request_duration_milliseconds_bucket{endpoint="/logout",status="",le="400.0",} 10113.0
http_request_duration_milliseconds_bucket{endpoint="/logout",status="",le="800.0",} 10493.0
http_request_duration_milliseconds_bucket{endpoint="/logout",status="",le="1600.0",} 10622.0
http_request_duration_milliseconds_bucket{endpoint="/logout",status="",le="+Inf",} 10722.0
http_request_duration_milliseconds_count{endpoint="/logout",status="",} 10722.0
http_request_duration_milliseconds_sum{endpoint="/logout",status="",} 1502959.0
http_request_duration_milliseconds_bucket{endpoint="/users",status="",le="25.0",} 3622.0
http_request_duration_milliseconds_bucket{endpoint="/users",status="",le="50.0",} 4967.0
http_request_duration_milliseconds_bucket{endpoint="/users",status="",le="100.0",} 6117.0
http_request_duration_milliseconds_bucket{endpoint="/users",status="",le="200.0",} 7254.0
http_request_duration_milliseconds_bucket{endpoint="/users",status="",le="400.0",} 7624.0
http_request_duration_milliseconds_bucket{endpoint="/users",status="",le="800.0",} 7866.0
http_request_duration_milliseconds_bucket{endpoint="/users",status="",le="1600.0",} 7966.0
http_request_duration_milliseconds_bucket{endpoint="/users",status="",le="+Inf",} 8034.0
http_request_duration_milliseconds_count{endpoint="/users",status="",} 8034.0
http_request_duration_milliseconds_sum{endpoint="/users",status="",} 1100809.0
http_request_duration_milliseconds_bucket{endpoint="/users",status="",le="25.0",} 41.0
http_request_duration_milliseconds_bucket{endpoint="/users",status="",le="50.0",} 88.0
http_request_duration_milliseconds_bucket{endpoint="/users",status="",le="100.0",} 94.0
http_request_duration_milliseconds_bucket{endpoint="/users",status="",le="200.0",} 94.0
http_request_duration_milliseconds_bucket{endpoint="/users",status="",le="400.0",} 94.0
http_request_duration_milliseconds_bucket{endpoint="/users",status="",le="800.0",} 94.0
http_request_duration_milliseconds_bucket{endpoint="/users",status="",le="1600.0",} 94.0
http_request_duration_milliseconds_bucket{endpoint="/users",status="",le="+Inf",} 94.0
http_request_duration_milliseconds_count{endpoint="/users",status="",} 94.0
http_request_duration_milliseconds_sum{endpoint="/users",status="",} 2685.0
http_request_duration_milliseconds_bucket{endpoint="/logout",status="",le="25.0",} 41.0
http_request_duration_milliseconds_bucket{endpoint="/logout",status="",le="50.0",} 85.0
http_request_duration_milliseconds_bucket{endpoint="/logout",status="",le="100.0",} 93.0
http_request_duration_milliseconds_bucket{endpoint="/logout",status="",le="200.0",} 93.0
http_request_duration_milliseconds_bucket{endpoint="/logout",status="",le="400.0",} 93.0
http_request_duration_milliseconds_bucket{endpoint="/logout",status="",le="800.0",} 93.0
http_request_duration_milliseconds_bucket{endpoint="/logout",status="",le="1600.0",} 93.0
http_request_duration_milliseconds_bucket{endpoint="/logout",status="",le="+Inf",} 93.0
http_request_duration_milliseconds_count{endpoint="/logout",status="",} 93.0
http_request_duration_milliseconds_sum{endpoint="/logout",status="",} 2683.0

我们来看具体的代码

metric有Counter、Gauge、Histogram和Summary四种类型

this.httpRequestsTotal.labels(endpoint, statusCode).inc();这里是创建一个计数器，统计http请求的信息

对不同的http请求的端点，改端点下不同的响应都会进行记录

http_requests_total{endpoint="/login",status="500",} 188.0
http_requests_total{endpoint="/register",status="500",} 55.0
http_requests_total{endpoint="/login",status="200",} 18863.0

= Histogram.build()
.name("http_request_duration_milliseconds")
.help("Http request latency histogram")
.exponentialBuckets(25, 2, 7)
.labelNames("endpoint", "status")
.register();

这里是创建一个直方图，用来统计延迟数据分布

http_request_duration_milliseconds_bucket{endpoint="/login",status="500",le="25.0",} 85.0
http_request_duration_milliseconds_bucket{endpoint="/login",status="500",le="50.0",} 174.0
http_request_duration_milliseconds_bucket{endpoint="/login",status="500",le="100.0",} 188.0
http_request_duration_milliseconds_bucket{endpoint="/login",status="500",le="200.0",} 188.0
http_request_duration_milliseconds_bucket{endpoint="/login",status="500",le="400.0",} 188.0
http_request_duration_milliseconds_bucket{endpoint="/login",status="500",le="800.0",} 188.0
http_request_duration_milliseconds_bucket{endpoint="/login",status="500",le="1600.0",} 188.0

125.【实验】Prometheus起步查询实验（中）~1.mp4

首先安装普罗米修斯

接下来我们要让普罗米修斯抓取我们上面的springboot http-simulation的数据http://localhost:8080/，需要修改prometheus.ym

# my global config
global:
  scrape_interval:     5s # Set the scrape interval to every  seconds. Default is every  minute.
  evaluation_interval: 5s # Evaluate rules every  seconds. The default is every  minute.
  # scrape_timeout is set to the global default (10s).

# Alertmanager configuration
alerting:
  alertmanagers:
  - static_configs:
    - targets:
      # - alertmanager:

# Load rules once and periodically evaluate them according to the global 'evaluation_interval'.
rule_files:
  # - "first_rules.yml"
  # - "second_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'

    # metrics_path defaults to '/metrics'
    # scheme defaults to 'http'.

    static_configs:
    - targets: ['localhost:9090']
  - job_name: 'http-simulation'

    metrics_path: /prometheus
    # scheme defaults to 'http'.

    static_configs:
    - targets: ['localhost:8080']    

创建一个新的job，job的名称是'http-simulation'，业务metrics的路径是/prometheus，对于的tartgets是localhost:8080'

下面我们将普罗米修斯运行起来

我们使用git bash窗口将普罗米修斯运行起来。这里不要使用windows的cmd窗口

启动成功之后，浏览器输入http://localhost:9090/graph

点击查看target可以看到当前监控了哪些target

可以看到当前监控了哪些实例

我们要查看http-simulation的请求数目，如何做的统计了:http_requests_total{job="http-simulation"}

对于统计我们可以点击上方的graph，查看一个图形的一个统计，查看的时间是可以手动修改的

校验http-simulator在1状态

up{job="http-simulator"}
查询http请求数

http_requests_total{job="http-simulator"}
查询成功login请求数

http_requests_total{job="http-simulator", status="", endpoint="/login"}
查询成功请求数，以endpoint区分

http_requests_total{job="http-simulator", status=""}
查询总成功请求数

sum(http_requests_total{job="http-simulator", status=""})
查询成功请求率，以endpoint区分

rate(http_requests_total{job="http-simulator", status=""}[5m])
查询总成功请求率

sum(rate(http_requests_total{job="http-simulator", status=""}[5m]))

http_requests_total{job="http-simulator", status="200", endpoint="/login"}
126.【实验】Prometheus起步查询实验（下）~1.mp4

. 延迟分布(Latency distribution)查询
查询http-simulator延迟分布

http_request_duration_milliseconds_bucket{job="http-simulator"}
查询成功login延迟分布

http_request_duration_milliseconds_bucket{job="http-simulator", status="", endpoint="/login"}
不超过200ms延迟的成功login请求占比

sum(http_request_duration_milliseconds_bucket{job="http-simulator", status="", endpoint="/login", le="200.0"}) / sum(http_request_duration_milliseconds_count{job="http-simulator", status="", endpoint="/login"})
成功login请求延迟的99百分位

histogram_quantile(0.99, rate(http_request_duration_milliseconds_bucket{job="http-simulator", status="", endpoint="/log

127.【实验】Prometheus + Grafana 展示实验（上）~1.mp4

安装grafana，下载之后解压完成之后就可以了

运行grafana之前，保证普罗米修斯以及我们要监控的应用已经正常启动起来了

我们使用git bash窗口运行

密码是admin/admin

登录成功之后，我们要给grafana设置普罗米修斯的数据源

直接点击add data source

其他参数默认不做修改

添加Proemethes数据源

Name -> prom-datasource
Type -> Prometheus
HTTP URL -> http://localhost:9090
其它缺省即可

http://localhost:9090是普罗米修斯运行的程序端口

Save & Test确保连接成功

3. 创建一个Dashboard
点击**+图标创建一个Dashbaord，点击保存**图标保存Dashboard，使用缺省Folder，给Dashboard起名为prom-demo。

4. 展示请求率

点击Add panel图标，点击Graph图标添加一个Graph，

点击Graph上的Panel Title->Edit进行编辑

修改Title：General -> Title = Request Rate

设置Metrics

sum(rate(http_requests_total{job="http-simulator"}[5m]))

调整Lagend

• 以表格展示As Table
• 显示Min/Max/Avg/Current/Total
• 根据需要调整Axis

注意保存Dahsboard。

设置完成metries之后点击下右边的那个小三角

128.【实验】Prometheus + Grafana 展示实验（下）~1.mp4

5. 展示实时错误率

点击Add panel图标，点击Singlestat图标添加一个Singlestat，

点击Graph上的Panel Title->Edit进行编辑

修改Title：General -> Title = Live Error Rate

设置Metrics

sum(rate(http_requests_total{job="http-simulator", status="500"}[5m])) / sum(rate(http_requests_total{job="http-simulator"}[5m]))

调整显示单位unit：Options->Unit，设置为None->percent(0.0-1.0)

调整显示值(目前为平均)为当前值(now)：Options->Value->Stat，设置为Current

添加阀值和颜色：Options->Coloring，选中Value，将Threshold设置为0.01,0.05，表示

• 绿色：0-1%
• 橙色：1-5%
• 红色：>5%

添加测量仪效果：Options->Gauge，选中Show，并将Max设为1

添加错误率演变曲线：选中Spark lines -> Show

注意保存Dahsboard。

sum(rate(http_requests_total{job="http-simulation",status="500"}[5m])) /sum(rate(http_requests_total{job="http-simulation"}[5m]))

调整显示单位unit：Options->Unit，设置为None->percent(0.0-1.0)

调整显示值(目前为平均)为当前值(now)：Options->Value->Stat，设置为Current

设置测量仪效果

6. 展示Top requested端点
点击Add panel图标，点击Table图标添加一个Table，

设置Metrics

sum(rate(http_requests_total{job="http-simulator"}[5m])) by (endpoint)
减少表中数据项，在Metrics下，选中Instant只显示当前值

隐藏Time列，在Column Sytle下，Apply to columns named为Time，将Type->Type设置为Hidden

将Value列重命名，添加一个Column Style，Apply to columns named为Value，将Column Header设置为Requests/s

点击表中的Requests/s header，让其中数据根据端点活跃度进行排序。

注意调整Widget位置并保存Dahsboard。

sum(rate(http_requests_total{job="http-simulation"}[5m])) by (endpoint)

勾选instance只看当前存在的请求类型

点击value可以按照值大小进行排序，让当前请求最大的值在前面

这样可以实时的查看当前的一个请求的统计。访问最频繁的五个端点信息等

129.【实验】Prometheus + Alertmanager 告警实验（上）~1.mp4

注意启用--web.enable-lifecycle，让Prometheus支持通过web端点动态更新配置

接下来我们做这样的一个功能，当我们之前运行的http-simulation这个应用挂了，我们发出一个告警

. HttpSimulatorDown告警
在Prometheus目录下:

添加simulator_alert_rules.yml告警配置文件

groups:
- name: simulator-alert-rule
  rules:
  - alert: HttpSimulatorDown
    expr: sum(up{job="http-simulation"}) ==
    for: 1m
    labels:
      severity: critical

一分钟内统计

 sum(up{job="http-simulation"})的值都是0，说明1分钟内实例都没有启动，发出告警
修改prometheus.yml，引用simulator_alert_rules.yml文件

# my global config
global:
  scrape_interval:     5s # Set the scrape interval to every  seconds. Default is every  minute.
  evaluation_interval: 5s # Evaluate rules every  seconds. The default is every  minute.
  # scrape_timeout is set to the global default (10s).

# Alertmanager configuration
alerting:
  alertmanagers:
  - static_configs:
    - targets:
      # - alertmanager:

# Load rules once and periodically evaluate them according to the global 'evaluation_interval'.
rule_files:
  - "simulator_alert_rules.yml"
  # - "second_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'

    # metrics_path defaults to '/metrics'
    # scheme defaults to 'http'.

    static_configs:
    - targets: ['localhost:9090']
  - job_name: 'http-simulation'
    metrics_path: /prometheus
    static_configs:
    - targets: ['localhost:8080']    

这样说明配置已经成功，我们将http-simulation应用停止，一分钟之后会触发报警

状态为firing表示告警已经触发了

3. ErrorRateHigh告警

假设已经执行上面的步骤2，则重新运行Prometheus HTTP Metrics Simulator

在simulator_alert_rules.yml文件中增加告警配置

- alert: ErrorRateHigh
    expr: sum(rate(http_requests_total{job="http-simulator", status=""}[5m])) / sum(rate(http_requests_total{job="http-simulator"}[5m])) > 0.02
    for: 1m
    labels:
      severity: major
    annotations:
      summary: "High Error Rate detected"
      description: "Error Rate is above 2% (current value is: {{ $value }}"

整个文件的内容如下

groups:
- name: simulator-alert-rule
  rules:
  - alert: HttpSimulatorDown
    expr: sum(up{job="http-simulation"}) ==
    for: 1m
    labels:
      severity: critical
  - alert: ErrorRateHigh
    expr: sum(rate(http_requests_total{job="http-simulator", status=""}[5m])) / sum(rate(http_requests_total{job="http-simulator"}[5m])) > 0.02
    for: 1m
    labels:
      severity: major
    annotations:
      summary: "High Error Rate detected"
      description: "Error Rate is above 2% (current value is: {{ $value }}"      

130.【实验】Prometheus + Alertmanager 告警实验（下）~1.mp4

上面我们已经设置了告警，接下来当产生告警的时候，能够发送邮件

首先要下载下载Alertmanager 0.15.2 for Windows，并解压到本地目录。

配置好邮箱地址之后，要启动

启动Alertmanager

./alertmanager.exe

在Prometheus目录下，修改prometheus.yml配置Alertmanager地址，默认是9093端口

# Alertmanager configuration
alerting:
alertmanagers:
- static_configs:
- targets:
- localhost:9093

通过Prometheus->Status的Configuration和Rules确认配置和告警设置生效

通过Alertmanager UI界面和设置的邮箱，校验ErrorRateHigh告警触发

Alertmanager UI访问地址：

http://localhost:9093

131.【实验】Java 应用埋点和监控实验~1.mp4

我们可以通过http api给队列添加job，worker进行干活

实验四、Java应用埋点和监控实验
实验步骤
1. Review和运行埋点样例代码
将instrumentation-example导入Eclipse IDE
Review代码理解模拟任务系统原理和埋点方式
以Spring Boot方式运行埋点案例
通过http://localhost:8080/prometheus查看metrics

pom.xml

<?xml version="1.0" encoding="UTF-8"?>
<project xmlns="http://maven.apache.org/POM/4.0.0" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
    xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd">
    <modelVersion>4.0.</modelVersion>

    <groupId>io.spring2go.promdemo</groupId>
    <artifactId>instrumentation-example</artifactId>
    <version>0.0.-SNAPSHOT</version>
    <packaging>jar</packaging>

    <name>instrumentation-example</name>
    <description>Demo project for Spring Boot</description>

    <parent>
        <groupId>org.springframework.boot</groupId>
        <artifactId>spring-boot-starter-parent</artifactId>
        <version>1.5..RELEASE</version>
        <relativePath/> <!-- lookup parent from repository -->
    </parent>

    <properties>
        <project.build.sourceEncoding>UTF-</project.build.sourceEncoding>
        <project.reporting.outputEncoding>UTF-</project.reporting.outputEncoding>
        <java.version>1.8</java.version>
    </properties>

    <dependencies>
        <dependency>
            <groupId>org.springframework.boot</groupId>
            <artifactId>spring-boot-starter-web</artifactId>
        </dependency>

        <!-- The prometheus client -->
        <dependency>
            <groupId>io.prometheus</groupId>
            <artifactId>simpleclient_spring_boot</artifactId>
            <version>0.5.</version>
        </dependency>

        <dependency>
            <groupId>org.springframework.boot</groupId>
            <artifactId>spring-boot-starter-test</artifactId>
            <scope>test</scope>
        </dependency>
    </dependencies>

    <build>
        <plugins>
            <plugin>
                <groupId>org.springframework.boot</groupId>
                <artifactId>spring-boot-maven-plugin</artifactId>
            </plugin>
        </plugins>
    </build>

</project>

InstrumentApplication

package io.spring2go.promdemo.instrument;

import org.springframework.boot.CommandLineRunner;
import org.springframework.boot.SpringApplication;
import org.springframework.boot.autoconfigure.SpringBootApplication;
import org.springframework.context.annotation.Bean;
import org.springframework.core.task.SimpleAsyncTaskExecutor;
import org.springframework.core.task.TaskExecutor;
import org.springframework.stereotype.Controller;
import org.springframework.web.bind.annotation.RequestMapping;
import org.springframework.web.bind.annotation.RequestMethod;
import org.springframework.web.bind.annotation.ResponseBody;

import io.prometheus.client.spring.boot.EnablePrometheusEndpoint;

@Controller
@SpringBootApplication
@EnablePrometheusEndpoint
public class InstrumentApplication {

    private JobQueue queue = new JobQueue();

    private WorkerManager workerManager;

    public static void main(String[] args) {

        SpringApplication.run(InstrumentApplication.class, args);
    }

    @RequestMapping(value = "/hello-world")
    public @ResponseBody String sayHello() {
        return "hello, world";
    }

    @RequestMapping(value = "/jobs", method = RequestMethod.POST)
    public @ResponseBody String jobs() {
        queue.push(new Job());
        return "ok";
    }

    @Bean
    public TaskExecutor taskExecutor() {
        return new SimpleAsyncTaskExecutor();
    }

    @Bean
    public CommandLineRunner schedulingRunner(TaskExecutor executor) {
        return new CommandLineRunner() {
            public void run(String... args) throws Exception {
                // 10 jobs per worker
                workerManager = new WorkerManager(queue, , , );
                executor.execute(workerManager);
                System.out.println("WorkerManager thread started...");
            }
        };
    }

}

Job

package io.spring2go.promdemo.instrument;

import java.util.Random;
import java.util.UUID;

public class Job {

    private String id;

    private Random rand = new Random();

    public Job() {
        this.id = UUID.randomUUID().toString();
    }

    public void run() {
        try {
            // Run the job (5 - 15 seconds)
            Thread.sleep(( + rand.nextInt()) * );
        } catch (InterruptedException e) {
            // TODO Auto-generated catch block
            e.printStackTrace();
        }
    }

    public String getId() {
        return id;
    }

    public void setId(String id) {
        this.id = id;
    }

}

JobQueue

package io.spring2go.promdemo.instrument;

import java.util.Queue;
import java.util.concurrent.LinkedBlockingQueue;

import io.prometheus.client.Gauge;

public class JobQueue {

    private final Gauge jobQueueSize = Gauge.build()
            .name("job_queue_size")
            .help("Current number of jobs waiting in queue")
            .register();

    private Queue<Job> queue = new LinkedBlockingQueue<Job>();

    public int size() {
        return queue.size();
    }

    public void push(Job job) {
        queue.offer(job);
        jobQueueSize.inc();
    }

    public Job pull() {
        Job job = queue.poll();
        if (job != null) {
            jobQueueSize.dec();
        }
        return job;
    }

}

Worker

package io.spring2go.promdemo.instrument;

import java.util.UUID;

import io.prometheus.client.Histogram;

public class Worker extends Thread {

    private static final Histogram jobsCompletionDurationSeconds  = Histogram.build()
            .name("jobs_completion_duration_seconds")
            .help("Histogram of job completion time")
            .linearBuckets(, , )
            .register();

    private String id;

    private JobQueue queue;

    private volatile boolean shutdown;

    public Worker(JobQueue queue) {
        this.queue = queue;
        this.id = UUID.randomUUID().toString();
    }

    @Override
    public void run() {
        System.out.println(String.format("[Worker %s] Starting", this.id));
        while(!shutdown) {
            this.pullJobAndRun();
        }
        System.out.println(String.format("[Worker %s] Stopped", this.id));
    }

    public void shutdown() {
        this.shutdown = true;
        System.out.println(String.format("[Worker %s] Shutting down", this.id));
    }

    public void pullJobAndRun() {
        Job job = this.queue.pull();
        if (job != null) {
            long jobStart = System.currentTimeMillis();
            System.out.println(String.format("[Worker %s] Starting job: %s", this.id, job.getId()));
            job.run();
            System.out.println(String.format("[Worker %s] Finished job: %s", this.id, job.getId()));
            int duration = (int)((System.currentTimeMillis() - jobStart) / );
            jobsCompletionDurationSeconds.observe(duration);
        } else {
            System.out.println(String.format("[Worker %s] Queue is empty. Backing off 5 seconds", this.id));
            try {
                Thread.sleep( * );
            } catch (InterruptedException e) {
                // TODO Auto-generated catch block
                e.printStackTrace();
            }
        }
    }

}

WorkerManager

package io.spring2go.promdemo.instrument;

import java.util.LinkedList;
import java.util.Queue;

public class WorkerManager extends Thread {

    private Queue<Worker> workers = new LinkedList<Worker>();

    private JobQueue queue;

    private int minWorkers;
    private int maxWorkers;

    private int jobsWorkerRatio;

    public WorkerManager(JobQueue queue, int minWorkers, int maxWorkers, int jobsWorkerRatio) {
        this.queue = queue;
        this.minWorkers = minWorkers;
        this.maxWorkers = maxWorkers;
        this.jobsWorkerRatio = jobsWorkerRatio;

        // Initialize workerpool
        for (int i = ; i < minWorkers; i++) {
            this.addWorker();
        }
    }

    public void addWorker() {
        Worker worker = new Worker(queue);
        this.workers.offer(worker);
        worker.start();
    }

    public void shutdownWorker() {
        if (this.workers.size() > ) {
            Worker worker = this.workers.poll();
            worker.shutdown();
        }
    }

    public void run() {
        this.scaleWorkers();
    }

    public void scaleWorkers() {
        while(true) {
            int queueSize = this.queue.size();
            int workerCount = this.workers.size();

            if ((workerCount + ) * jobsWorkerRatio < queueSize && workerCount < this.maxWorkers) {
                System.out.println("[WorkerManager] Too much work, starting extra worker.");
                this.addWorker();
            }

            if ((workerCount - ) * jobsWorkerRatio > queueSize && workerCount > this.minWorkers) {
                System.out.println("[WorkerManager] Too much workers, shutting down 1 worker");
                this.shutdownWorker();
            }

            try {
                Thread.sleep( * );
            } catch (InterruptedException e) {
                // TODO Auto-generated catch block
                e.printStackTrace();
            }
        }
    }

}

application.properties

management.security.enabled=false

我们要监听两个指标：第一job队列的大小，二是队列中每个任务执行的时间
job的大小我们在job中添加监听，从队列中post job的时候队列大小增加1，取出减少1

2. 配置和运行Promethus
添加针对instrumentation-example的监控job

- job_name: 'instrumentation-example'
metrics_path: /prometheus
static_configs:
- targets: ['localhost:8080']
运行Prometheus

./prometheus.exe
通过Prometheus->Status的configuration和targets校验配置正确

3. 生成测试数据和查询Metrics
查询instrumentation-example在UP1状态

up{job="instrumentation-example"}
运行queueUpJobs.sh产生100个job

./queueUpJobs.sh
查询JobQueueSize变化曲线(调整时间范围到5m)：

job_queue_size{job="instrumentation-example"}
查询90分位Job执行延迟分布：

histogram_quantile(0.90, rate(jobs_completion_duration_seconds_bucket{job="instrumentation-example"}[5m]))

132.【实验】NodeExporter 系统监控实验~1.mp4

1. 下载和运行wmi-exporter
下载wmi_exporter-amd64，并解压到本地目录

校验metrics端点

http://localhost:9182/metrics

2. 配置和运行Promethus
在Prometheus安装目录下

在prometheus.yml 中添加针对wmi-exporter的监控job

# my global config
global:
  scrape_interval:     5s # Set the scrape interval to every  seconds. Default is every  minute.
  evaluation_interval: 5s # Evaluate rules every  seconds. The default is every  minute.
  # scrape_timeout is set to the global default (10s).

# Alertmanager configuration
alerting:
  alertmanagers:
  - static_configs:
    - targets:
      # - alertmanager:

# Load rules once and periodically evaluate them according to the global 'evaluation_interval'.
rule_files:
  - "simulator_alert_rules.yml"
  # - "second_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: 'wmi-exporter'

    # metrics_path defaults to '/metrics'
    # scheme defaults to 'http'.

    static_configs:
    - targets: ['localhost:9182']
 

3. Grafana Dashboard for wmi-exporter
在Grafana安装目录下启动Grafana服务器

./bin/grafana-server.exe
登录Grafana UI(admin/admin)

http://localhost:3000
通过Grafana的**+**图标导入(Import) wmi-exporter dashboard：

grafana id = 2129
注意选中prometheus数据源
查看Windows Node dashboard。

4. 参考

Grafana Dashboard仓库

https://grafana.com/dashboards

然后输入2129，会自动导入变成下面的信息

这里数据源要选择普罗米修斯数据源

133.【实验】Spring Boot Actuator 监控实验~1.mp4

实验步骤
1. 运行Spring Boot + Actuator
将actuatordemo应用导入Eclipse IDE

Review actuatordemo代码

以Spring Boot方式运行actuatordemo

校验metrics端点

http://localhost:8080/prometheus
2. 配置和运行Promethus
在Prometheus安装目录下

在prometheus.yml 中添加针对wmi-exporter的监控job

- job_name: 'actuator-demo'
metrics_path: '/prometheus'
static_configs:
- targets: ['localhost:8080']
运行Prometheus

./prometheus.exe
访问Prometheus UI

http://localhost:9090
通过Prometheus->Status的configuration和targets校验配置正确

通过Prometheus->Graph查询actuator-demo在UP=1状态

up{job="actuatordemo"}
3. Grafana Dashboard for JVM (Micrometer)
在Grafana安装目录下启动Grafana服务器

./bin/grafana-server.exe
登录Grafana UI(admin/admin)

http://localhost:3000
通过Grafana的**+**图标导入(Import) JVM (Micrometer) dashboard：

grafana id = 4701
注意选中prometheus数据源
查看JVM (Micormeter) dashboard。

4. 参考
Grafana Dashboard仓库

https://grafana.com/dashboards
Micrometer Prometheus支持

https://micrometer.io/docs/registry/prometheus
Micrometer Springboot 1.5支持

https://micrometer.io/docs/ref/spring/1.5

133.【实验】Spring Boot Actuator 监控实验~1.mp4

整个程序的代码如下

首先我们来看下代码

<?xml version="1.0" encoding="UTF-8"?>
<project xmlns="http://maven.apache.org/POM/4.0.0" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
    xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd">
    <modelVersion>4.0.</modelVersion>

    <groupId>io.spring2go.promdemo</groupId>
    <artifactId>actuatordemo</artifactId>
    <version>0.0.-SNAPSHOT</version>
    <packaging>jar</packaging>

    <name>actuatordemo</name>
    <description>Demo project for Spring Boot</description>

    <parent>
        <groupId>org.springframework.boot</groupId>
        <artifactId>spring-boot-starter-parent</artifactId>
        <version>1.5..RELEASE</version>
        <relativePath/> <!-- lookup parent from repository -->
    </parent>

    <properties>
        <project.build.sourceEncoding>UTF-</project.build.sourceEncoding>
        <project.reporting.outputEncoding>UTF-</project.reporting.outputEncoding>
        <java.version>1.8</java.version>
    </properties>

    <dependencies>
        <dependency>
            <groupId>org.springframework.boot</groupId>
            <artifactId>spring-boot-starter-actuator</artifactId>
        </dependency>

        <dependency>
            <groupId>org.springframework.boot</groupId>
            <artifactId>spring-boot-starter-web</artifactId>
        </dependency>

        <dependency>
          <groupId>io.micrometer</groupId>
          <artifactId>micrometer-spring-legacy</artifactId>
          <version>1.0.</version>
        </dependency>

        <dependency>
          <groupId>io.micrometer</groupId>
          <artifactId>micrometer-registry-prometheus</artifactId>
          <version>1.0.</version>
        </dependency>        

        <dependency>
            <groupId>io.github.mweirauch</groupId>
            <artifactId>micrometer-jvm-extras</artifactId>
            <version>0.1.</version>
        </dependency>

        <dependency>
            <groupId>org.springframework.boot</groupId>
            <artifactId>spring-boot-starter-test</artifactId>
            <scope>test</scope>
        </dependency>
    </dependencies>

    <build>
        <plugins>
            <plugin>
                <groupId>org.springframework.boot</groupId>
                <artifactId>spring-boot-maven-plugin</artifactId>
            </plugin>
        </plugins>
    </build>

</project>

在自己本地电脑上建立一个Prometheus和Grafana仪表盘，用来可视化监控Spring Boot应用产生的所有metrics。

Spring Boot使用Micrometer，一个应用metrics组件，将actuator metrics整合到外部监控系统中。
它支持很多种监控系统，比如Netflix Atalas, AWS Cloudwatch, Datadog, InfluxData, SignalFx, Graphite, Wavefront和Prometheus等。
为了整合Prometheus，你需要增加micrometer-registry-prometheus依赖：

在SpringBoot2.X中，spring-boot-starter-actuator引入了io.micrometer，对1.X中的metrics进行了重构，主要特点是支持tag/label，配合支持tag/label的监控系统，使得我们可以更加方便地对metrics进行多维度的统计查询及监控
spring-boot-starter-actuator，主要是提供了Prometheus端点，不用重复造*。

ActuatordemoApplication

package io.spring2go.promdemo.actuatordemo;

import org.springframework.boot.SpringApplication;
import org.springframework.boot.autoconfigure.SpringBootApplication;
import org.springframework.context.annotation.Bean;
import org.springframework.stereotype.Controller;
import org.springframework.web.bind.annotation.RequestMapping;
import org.springframework.web.bind.annotation.ResponseBody;

import io.micrometer.core.instrument.MeterRegistry;
import io.micrometer.spring.autoconfigure.MeterRegistryCustomizer;

@SpringBootApplication
@Controller
public class ActuatordemoApplication {

    public static void main(String[] args) {
        SpringApplication.run(ActuatordemoApplication.class, args);
    }

    @RequestMapping(value = "/hello-world")
    public @ResponseBody String sayHello() {
        return "hello, world";
    }

    @Bean
    MeterRegistryCustomizer<MeterRegistry> metricsCommonTags() {
      return registry -> registry.config().commonTags("application", "actuator-demo");
    }
}

上面要设置注册功能

commonTags必须叫做application

application.properties

endpoints.sensitive=false

访问actouor端点的时候不需要用户名和密码

接下来我们启动应用

如果要收到埋点，使用下面的micrometer提供的方法在springboot应用中进行埋点就可以了

# my global config
global:
scrape_interval: 5s # Set the scrape interval to every 15 seconds. Default is every 1 minute.
evaluation_interval: 5s # Evaluate rules every 15 seconds. The default is every 1 minute.
# scrape_timeout is set to the global default (10s).

# Alertmanager configuration
alerting:
alertmanagers:
- static_configs:
- targets:
# - alertmanager:9093

# Load rules once and periodically evaluate them according to the global 'evaluation_interval'.
rule_files:
#- "simulator_alert_rules.yml"
# - "second_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: 'actuator-demo'
metrics_path: /prometheus
# metrics_path defaults to '/metrics'
# scheme defaults to 'http'.

static_configs:
- targets: ['localhost:8080']

这里代码  - job_name: 'actuator-demo'要和代码  return registry -> registry.config().commonTags("application", "actuator-demo");中的对应起来
 这样整个监控就启动起来了

jvm dashboard对于的id为4701

添加4701，数据源一定要选择普罗米修斯的数据源

134.Prometheus 监控最佳实践~1.mp4

例如http的请求数目，http请求的平均演示，http500的错误数目。cpu、内存、磁盘的使用情况等

online为请求响应的系统：前台的web系统，请求、错误、db、缓存、请求延迟等进行跟踪

offline serving system：任务队列系统 队里工作线程的利用率、任务线程执行的情况

Batch Jobs：批处理系统，普罗米修斯到网关上拉去数据

普罗米修斯的高可用

普罗米修斯默认只支持15天的数据，超过整个范围的数据需要单独做处理

135.主流开源时序数据库比较~1.mp4

137.微服务监控体系总结~1.mp4

系统层监控：底层硬件，操作系统的监控，普罗米修斯做系统操作系统的监控

应用层的监控：消息队列redis、mysql spingboot 普罗米修斯进行埋点，cat skywalking做应用层的监控 elk可以统计系统的日志做监控

业务层：主要桶业务进行统计，例如转账等业务指标，普罗米修斯就可以搞定

端用户体验：客户访问网页的时间，听云等

微服务架构的监控体系

左边是对一个微服务架构做监控

主要的监控类别有下面的三种：log、Trace、Metrics

log：微服务产生log之后使用logstash进行日志的收集、然后使用kafka队列做缓冲，保证消息不被丢失，然后存储到es数据库中

trace：追踪分布式链路调用的场景

metris：在微服务里面进行埋点，抓取metris数据，然后做告警等操作

普罗米修斯也可以对kafka、cat等中间件进行监控，然后进行告警操作