【prometheus】- 01 云原生时代的监控系统入门

cfg := flagConfig{
  notifier: notifier.Options{
   // 默认注册器注册 cpu 和 go 指标收集器
   Registerer: prometheus.DefaultRegisterer,
  },
  web: web.Options{
   Registerer: prometheus.DefaultRegisterer,
   Gatherer:   prometheus.DefaultGatherer,
  },
  promlogConfig: promlog.Config{},
 }

type flagConfig struct {
 configFile string

 localStoragePath    string
 notifier            notifier.Options
 forGracePeriod      model.Duration
 outageTolerance     model.Duration
 resendDelay         model.Duration
 web                 web.Options
 tsdb                tsdbOptions
 lookbackDelta       model.Duration
 webTimeout          model.Duration
 queryTimeout        model.Duration
 queryConcurrency    int
 queryMaxSamples     int
 RemoteFlushDeadline model.Duration

 featureList []string
 // These options are extracted from featureList
 // for ease of use.
 enablePromQLAtModifier     bool
 enablePromQLNegativeOffset bool
 enableExpandExternalLabels bool

 prometheusURL   string
 corsRegexString string

 promlogConfig promlog.Config
}

a.Flag("config.file", "Prometheus configuration file path.").
  Default("prometheus.yml").StringVar(&cfg.configFile)

a.Flag("web.listen-address", "Address to listen on for UI, API, and telemetry.").
  Default("0.0.0.0:9090").StringVar(&cfg.web.ListenAddress)

a.Flag("web.read-timeout",
  "Maximum duration before timing out read of the request, and closing idle connections.").
  Default("5m").SetValue(&cfg.webTimeout)
...

--config.file:指定prometheus主配置文件路径
--web.listen-address:指定prometheus监听地址
--storage.tsdb.path:本地存储模式数据存放目录
--storage.tsdb.retention.time:本地存储模式存放时间，默认15天
--storage.tsdb.retention.size:数据保存的最大大小，支持的单位 B, KB, MB, GB, TB, PB, EB. Ex

_, err := a.Parse(os.Args[1:])

if _, err := config.LoadFile(cfg.configFile, false, log.NewNopLogger()); err != nil {
 level.Error(logger).Log("msg", fmt.Sprintf("Error loading config (--config.file=%s)", cfg.configFile), "err", err)
 os.Exit(2)
}

// Now that the validity of the config is established, set the config
// success metrics accordingly, although the config isn't really loaded
// yet. This will happen later (including setting these metrics again),
// but if we don't do it now, the metrics will stay at zero until the
// startup procedure is complete, which might take long enough to
// trigger alerts about an invalid config.
configSuccess.Set(1)
configSuccessTime.SetToCurrentTime()

{ // Time retention settings.
    // --storage.tsdb.retention参数已被废弃，不建议使用，如果使用会打印warn警告日志
 if oldFlagRetentionDuration != 0 {
  level.Warn(logger).Log("deprecation_notice", "'storage.tsdb.retention' flag is deprecated use 'storage.tsdb.retention.time' instead.")
  cfg.tsdb.RetentionDuration = oldFlagRetentionDuration
 }

 // --storage.tsdb.retention.time赋值给cfg.tsdb.RetentionDuration
 if newFlagRetentionDuration != 0 {
  cfg.tsdb.RetentionDuration = newFlagRetentionDuration
 }

    // --storage.tsdb.retention.time和--storage.tsdb.retention.size都为0，即既没有限制存储最大时长，也没有限制存储最大空间容量场景下，默认存储15d(天)
 if cfg.tsdb.RetentionDuration == 0 && cfg.tsdb.MaxBytes == 0 {
  cfg.tsdb.RetentionDuration = defaultRetentionDuration
  level.Info(logger).Log("msg", "No time or size retention was set so using the default time retention", "duration", defaultRetentionDuration)
 }

 // Check for overflows. This limits our max retention to 100y.
 // 如果最大存储时长过大溢出整数范围，则限制为100年
 if cfg.tsdb.RetentionDuration < 0 {
  y, err := model.ParseDuration("100y")
  if err != nil {
   panic(err)
  }
  cfg.tsdb.RetentionDuration = y
  level.Warn(logger).Log("msg", "Time retention value is too high. Limiting to: "+y.String())
 }
}

{
 if cfg.tsdb.MaxBlockDuration == 0 {
  maxBlockDuration, err := model.ParseDuration("31d")
  if err != nil {
   panic(err)
  }
  // When the time retention is set and not too big use to define the max block duration.
  if cfg.tsdb.RetentionDuration != 0 && cfg.tsdb.RetentionDuration/10 < maxBlockDuration {
   maxBlockDuration = cfg.tsdb.RetentionDuration / 10
  }
  cfg.tsdb.MaxBlockDuration = maxBlockDuration
 }
}

/**
 Prometheus的Storage组件是时序数据库，包含两个：localStorage和remoteStorage．localStorage当前版本指TSDB，
 用于对metrics的本地存储存储，remoteStorage用于metrics的远程存储，其中fanoutStorage作为localStorage和remoteStorage的读写代理服务器
  */
var (
 // 本地存储
 localStorage  = &readyStorage{}
 scraper       = &readyScrapeManager{}
 // 远程存储
 remoteStorage = remote.NewStorage(log.With(logger, "component", "remote"), prometheus.DefaultRegisterer, localStorage.StartTime, cfg.localStoragePath, time.Duration(cfg.RemoteFlushDeadline), scraper)
 // 读写代理服务器
 fanoutStorage = storage.NewFanout(logger, localStorage, remoteStorage)
)

notifierManager = notifier.NewManager(&cfg.notifier, log.With(logger, "component", "notifier"))

discoveryManagerScrape  = discovery.NewManager(ctxScrape, log.With(logger, "component", "discovery manager scrape"), discovery.Name("scrape"))

discoveryManagerNotify  = discovery.NewManager(ctxNotify, log.With(logger, "component", "discovery manager notify"), discovery.Name("notify"))

scrapeManager = scrape.NewManager(log.With(logger, "component", "scrape manager"), fanoutStorage)

// 声明 promql 的引擎配置
opts = promql.EngineOpts{
 Logger:                   log.With(logger, "component", "query engine"),
 Reg:                      prometheus.DefaultRegisterer,
 MaxSamples:               cfg.queryMaxSamples,
 Timeout:                  time.Duration(cfg.queryTimeout), // 查询超时时
 ActiveQueryTracker:       promql.NewActiveQueryTracker(cfg.localStoragePath, cfg.queryConcurrency, log.With(logger, "component", "activeQueryTracker")),
 LookbackDelta:            time.Duration(cfg.lookbackDelta),
 NoStepSubqueryIntervalFn: noStepSubqueryInterval.Get,
 EnableAtModifier:         cfg.enablePromQLAtModifier,
 EnableNegativeOffset:     cfg.enablePromQLNegativeOffset,
}
// 初始化queryEngine
queryEngine = promql.NewEngine(opts)

//ruleManager组件通过方法rules.NewManager完成初始化．其中rules.NewManager的参数涉及多个组件：存储，queryEngine和notifier，整个流程包含rule计算和发送告警
ruleManager = rules.NewManager(&rules.ManagerOptions{
 Appendable:      fanoutStorage,
 Queryable:       localStorage,
 QueryFunc:       rules.EngineQueryFunc(queryEngine, fanoutStorage),
 NotifyFunc:      sendAlerts(notifierManager, cfg.web.ExternalURL.String()),
 Context:         ctxRule,
 ExternalURL:     cfg.web.ExternalURL,
 Registerer:      prometheus.DefaultRegisterer,
 Logger:          log.With(logger, "component", "rule manager"),
 OutageTolerance: time.Duration(cfg.outageTolerance),
 ForGracePeriod:  time.Duration(cfg.forGracePeriod),
 ResendDelay:     time.Duration(cfg.resendDelay),
})

//
scraper.Set(scrapeManager)

// Depends on cfg.web.ScrapeManager so needs to be after cfg.web.ScrapeManager = scrapeManager.
webHandler := web.New(log.With(logger, "component", "web"), &cfg.web)

reloaders := []reloader{
 {
  name:     "remote_storage",
  // 存储配置
  reloader: remoteStorage.ApplyConfig,
 }, {
  name:     "web_handler",
  // web配置
  reloader: webHandler.ApplyConfig,
 }, {
  name: "query_engine",
  reloader: func(cfg *config.Config) error {
   if cfg.GlobalConfig.QueryLogFile == "" {
    queryEngine.SetQueryLogger(nil)
    return nil
   }

   l, err := logging.NewJSONFileLogger(cfg.GlobalConfig.QueryLogFile)
   if err != nil {
    return err
   }
   queryEngine.SetQueryLogger(l)
   return nil
  },
 }, {
  // The Scrape and notifier managers need to reload before the Discovery manager as
  // they need to read the most updated config when receiving the new targets list.
  // scrape 和 notifier manager 要在 discovery manager 之前重新加载，因为它们要在获取新的监控目标之前最新配置
  name:     "scrape",
   // scrapeManger配置
  reloader: scrapeManager.ApplyConfig,
 }, {
  name: "scrape_sd",
  //从配置文件中提取Section:scrape_configs
  reloader: func(cfg *config.Config) error {
   c := make(map[string]discovery.Configs)
   for _, v := range cfg.ScrapeConfigs {
    c[v.JobName] = v.ServiceDiscoveryConfigs
   }
   return discoveryManagerScrape.ApplyConfig(c)
  },
 }, {
  name:     "notify",
  // notifier配置
  reloader: notifierManager.ApplyConfig,
 }, {
  name: "notify_sd",
  //从配置文件中提取Section:alerting
  reloader: func(cfg *config.Config) error {
   c := make(map[string]discovery.Configs)
   for k, v := range cfg.AlertingConfig.AlertmanagerConfigs.ToMap() {
    c[k] = v.ServiceDiscoveryConfigs
   }
   return discoveryManagerNotify.ApplyConfig(c)
  },
 }, {
  name: "rules",
  //从配置文件中提取Section:rule_files
  reloader: func(cfg *config.Config) error {
   // Get all rule files matching the configuration paths.
   var files []string
   for _, pat := range cfg.RuleFiles {
    fs, err := filepath.Glob(pat)
    if err != nil {
     // The only error can be a bad pattern.
     return errors.Wrapf(err, "error retrieving rule files for %s", pat)
    }
    files = append(files, fs...)
   }
   return ruleManager.Update(
    time.Duration(cfg.GlobalConfig.EvaluationInterval),
    files,
    cfg.GlobalConfig.ExternalLabels,
   )
  },
 },
}

var g run.Group

//加入goroutine（或者称为 actor）
g.Add(...)
g.Add(...)

//调用Run方法后会启动所有goroutine（或者称为 actor）
if err := g.Run(); err != nil {
 level.Error(logger).Log("err", err)
 os.Exit(1)
}

// 创建监听退出chan
term := make(chan os.Signal, 1)
// pkill信号syscall.SIGTERM
// ctrl+c信号os.Interrupt
// 首先我们创建一个os.Signal channel，然后使用signal.Notify注册要接收的信号。
signal.Notify(term, os.Interrupt, syscall.SIGTERM)
cancel := make(chan struct{})

g.Add(
 func() error {
  // Don't forget to release the reloadReady channel so that waiting blocks can exit normally.
  select {
  case <-term: //监听到系统ctrl+c或kill等程序退出信号
   level.Warn(logger).Log("msg", "Received SIGTERM, exiting gracefully...")
   reloadReady.Close()
  case <-webHandler.Quit()://监听到web服务停止
   level.Warn(logger).Log("msg", "Received termination request via web service, exiting gracefully...")
  case <-cancel://其它需要退出程序信号
   reloadReady.Close()
  }
  return nil
 },
 func(err error) {
  close(cancel)
 },
)

g.Add(
 func() error {
  err := discoveryManagerScrape.Run()
  level.Info(logger).Log("msg", "Scrape discovery manager stopped")
  return err
 },
 func(err error) {
  level.Info(logger).Log("msg", "Stopping scrape discovery manager...")
  cancelScrape()
 },
)

g.Add(
 func() error {
  err := discoveryManagerNotify.Run()
  level.Info(logger).Log("msg", "Notify discovery manager stopped")
  return err
 },
 func(err error) {
  level.Info(logger).Log("msg", "Stopping notify discovery manager...")
  cancelNotify()
 },
)

g.Add(
 func() error {
  // When the scrape manager receives a new targets list
  // it needs to read a valid config for each job.
  // It depends on the config being in sync with the discovery manager so
  // we wait until the config is fully loaded.
  // 当所有配置都准备好
  // scrape manager 获取到新的抓取目标列表时，它需要读取每个 job 的合法的配置。
  // 这依赖于正在被 discovery manager 同步的配置文件，所以要等到配置加载完成。
  <-reloadReady.C
  level.Info(logger).Log("cus_msg", "--->ScrapeManager reloadReady")
  // 启动scrapeManager
  //ScrapeManager组件的启动函数
  err := scrapeManager.Run(discoveryManagerScrape.SyncCh())
  level.Info(logger).Log("msg", "Scrape manager stopped")
  return err
 },
 func(err error) {
  // 失败处理
  // Scrape manager needs to be stopped before closing the local TSDB
  // so that it doesn't try to write samples to a closed storage.
  level.Info(logger).Log("msg", "Stopping scrape manager...")
  scrapeManager.Stop()
 },
)

// Make sure that sighup handler is registered with a redirect to the channel before the potentially
// long and synchronous tsdb init.
// tsdb 初始化时间可能很长，确保 sighup 处理函数在这之前注册完成。
hup := make(chan os.Signal, 1)
signal.Notify(hup, syscall.SIGHUP)
cancel := make(chan struct{})
g.Add(
 func() error {
  <-reloadReady.C
  for {
   select {
   case <-hup:
    if err := reloadConfig(cfg.configFile, cfg.enableExpandExternalLabels, logger, noStepSubqueryInterval, reloaders...); err != nil {
     level.Error(logger).Log("msg", "Error reloading config", "err", err)
    }
   case rc := <-webHandler.Reload():
    if err := reloadConfig(cfg.configFile, cfg.enableExpandExternalLabels, logger, noStepSubqueryInterval, reloaders...); err != nil {
     level.Error(logger).Log("msg", "Error reloading config", "err", err)
     rc <- err
    } else {
     rc <- nil
    }
   case <-cancel:
    return nil
   }
  }
 },
 func(err error) {
  // Wait for any in-progress reloads to complete to avoid
  // reloading things after they have been shutdown.
  cancel <- struct{}{}
 },
)

cancel := make(chan struct{})
g.Add(
 func() error {
  select {
  case <-dbOpen:
  // In case a shutdown is initiated before the dbOpen is released
  case <-cancel:
   reloadReady.Close()
   return nil
  }

  //加载解析prometheus.yml配置文件，并调用各个组件ApplyConfig()方法将配置传入
  if err := reloadConfig(cfg.configFile, cfg.enableExpandExternalLabels, logger, noStepSubqueryInterval, reloaders...); err != nil {
   return errors.Wrapf(err, "error loading config from %q", cfg.configFile)
  }
  //配置加载完毕，执行reloadReady.Close()关闭reloadReady.C通道，这样  <-reloadReady.C 阻塞地方可以继续向下执行
  reloadReady.Close()
  webHandler.Ready()
  level.Info(logger).Log("msg", "Server is ready to receive web requests.")
  <-cancel
  return nil
 },
 func(err error) {
  close(cancel)
 },
)

g.Add(
 func() error {
  <-reloadReady.C
  ruleManager.Run()
  return nil
 },
 func(err error) {
  ruleManager.Stop()
 },
)

opts := cfg.tsdb.ToTSDBOptions()
cancel := make(chan struct{})
g.Add(
 func() error {
  level.Info(logger).Log("msg", "Starting TSDB ...")
  if cfg.tsdb.WALSegmentSize != 0 {
   if cfg.tsdb.WALSegmentSize < 10*1024*1024 || cfg.tsdb.WALSegmentSize > 256*1024*1024 {
    return errors.New("flag 'storage.tsdb.wal-segment-size' must be set between 10MB and 256MB")
   }
  }
  if cfg.tsdb.MaxBlockChunkSegmentSize != 0 {
   if cfg.tsdb.MaxBlockChunkSegmentSize < 1024*1024 {
    return errors.New("flag 'storage.tsdb.max-block-chunk-segment-size' must be set over 1MB")
   }
  }
  db, err := openDBWithMetrics(
   cfg.localStoragePath,
   logger,
   prometheus.DefaultRegisterer,
   &opts,
  )

  if err != nil {
   return errors.Wrapf(err, "opening storage failed")
  }

  switch fsType := prom_runtime.Statfs(cfg.localStoragePath); fsType {
  case "NFS_SUPER_MAGIC":
   level.Warn(logger).Log("fs_type", fsType, "msg", "This filesystem is not supported and may lead to data corruption and data loss. Please carefully read https://prometheus.io/docs/prometheus/latest/storage/ to learn more about supported filesystems.")
  default:
   level.Info(logger).Log("fs_type", fsType)
  }

  level.Info(logger).Log("msg", "TSDB started")
  level.Debug(logger).Log("msg", "TSDB options",
   "MinBlockDuration", cfg.tsdb.MinBlockDuration,
   "MaxBlockDuration", cfg.tsdb.MaxBlockDuration,
   "MaxBytes", cfg.tsdb.MaxBytes,
   "NoLockfile", cfg.tsdb.NoLockfile,
   "RetentionDuration", cfg.tsdb.RetentionDuration,
   "WALSegmentSize", cfg.tsdb.WALSegmentSize,
   "AllowOverlappingBlocks", cfg.tsdb.AllowOverlappingBlocks,
   "WALCompression", cfg.tsdb.WALCompression,
  )

  startTimeMargin := int64(2 * time.Duration(cfg.tsdb.MinBlockDuration).Seconds() * 1000)
  localStorage.Set(db, startTimeMargin)
  time.Sleep(time.Duration(10)*time.Second)
  close(dbOpen)
  <-cancel
  return nil
 },
 func(err error) {
  if err := fanoutStorage.Close(); err != nil {
   level.Error(logger).Log("msg", "Error stopping storage", "err", err)
  }
  close(cancel)
 },
)

g.Add(
 func() error {
  if err := webHandler.Run(ctxWeb, listener, *webConfig); err != nil {
   return errors.Wrapf(err, "error starting web server")
  }
  return nil
 },
 func(err error) {
  cancelWeb()
 },
)

g.Add(
 func() error {
  // When the notifier manager receives a new targets list
  // it needs to read a valid config for each job.
  // It depends on the config being in sync with the discovery manager
  // so we wait until the config is fully loaded.
  <-reloadReady.C
  notifierManager.Run(discoveryManagerNotify.SyncCh())
  level.Info(logger).Log("msg", "Notifier manager stopped")
  return nil
 },
 func(err error) {
  notifierManager.Stop()
 },
)