TL;DR

  • Github：yabeda-rb/yabeda
  • 使用Yabeda.configure定义指标
  • 记得记得使用Yabeda.configure!执行指标定义！（在rails中是自动的）
  • 使用Yabeda.group_name.metric_name.increment设置指数
跟踪

跟踪github示例代码

# declare.rb
require 'yabeda'
require 'trace_tree'

binding.trace_tree(htmp: 'configure') do
  Yabeda.configure do
    group :your_app do
      counter   :bells_rang_count, comment: "Total number of bells being rang", tags: %i[bell_size]
      gauge     :whistles_active,  comment: "Number of whistles ready to whistle"
      histogram :whistle_runtime do
        comment "How long whistles are being active"
        unit :ms
        buckets [0.1, 10, 100]
      end
    end
  end
end

binding.trace_tree(htmp: 'configure_bang') do
  Yabeda.configure!
end

得：

![[yabedaconfigtrace.html]]

![[yabedaconfigbang_trace.html]]

使用configure定义指标

这里configure的逻辑比较简单，只是将配置收集起来，等待configure!调用时再执行

# yabeda-0.11.0/lib/yabeda/dsl/class_methods.rb
def configure(&block)
  Yabeda.configurators.push([@group, block])
  class_exec(&block) if Yabeda.configured?
  @group = nil
end

configure!执行时，会将block中的配置对应到lib/yabeda/dsl/class_methods.rb中的DSL方法来运行

那些counter、gauge、histogramDSL方法定义如下

# yabeda-0.11.0/lib/yabeda/dsl/class_methods.rb
def counter(*args, **kwargs, &block)
  metric = MetricBuilder.new(Counter).build(args, kwargs, @group, &block)
  register_metric(metric)
end

# Register a gauge
def gauge(*args, **kwargs, &block)
  metric = MetricBuilder.new(Gauge).build(args, kwargs, @group, &block)
  register_metric(metric)
end

# Register a histogram
def histogram(*args, **kwargs, &block)
  metric = MetricBuilder.new(Histogram).build(args, kwargs, @group, &block)
  register_metric(metric)
end

生成Counter、Gauge、Histogram对象后，还会在::Yabeda上定义类方法以便用户获取它们

例如上述的bells_rang_count，可以用Yabeda.your_app_bells_rang_count取得，也可以用Yabeda.your_app.bells_rang_count

实现如下

# yabeda-0.11.0/lib/yabeda/dsl/class_methods.rb
def register_metric(metric)
  name = [metric.group, metric.name].compact.join("_")
  ::Yabeda.define_singleton_method(name) { metric }
  ::Yabeda.metrics[name] = metric
  register_group_for(metric) if metric.group
  ::Yabeda.adapters.each_value { |adapter| adapter.register!(metric) } if ::Yabeda.configured?
  metric
end

def register_group_for(metric)
  group = ::Yabeda.groups[metric.group]

  if group.nil?
    group = Group.new(metric.group)
    ::Yabeda.groups[metric.group] = group
  end

  ::Yabeda.define_singleton_method(metric.group) { group } unless ::Yabeda.respond_to?(metric.group)

  group.register_metric(metric)
end

Histogram类

调用Yabeda.xxxx_histogram.measure记录分布值

可以直接给它时长：Yabeda.xx_histogram({k: v}, 0.03)；也可以让它帮你计时：Yabeda.xx_histogram({k: v}){ ... }，默认计算秒数

# yabeda-0.11.0/lib/yabeda/histogram.rb
module Yabeda
  class Histogram < Metric
    option :buckets

    def measure(tags, value = nil)
      # 异或：不准同时提供value和block
      if value.nil? ^ block_given?
        raise ArgumentError, "You must provide either numeric value or block for Yabeda::Histogram#measure!"
      end

      if block_given?
        starting = Process.clock_gettime(Process::CLOCK_MONOTONIC)
        yield
        value = (Process.clock_gettime(Process::CLOCK_MONOTONIC) - starting)
      end

      # 记录到values这个hash中
      all_tags = ::Yabeda::Tags.build(tags, group)
      values[all_tags] = value

      # 如果有其它监控系统的适配器，则将标签和值传给它们
      ::Yabeda.adapters.each do |_, adapter|
        adapter.perform_histogram_measure!(self, all_tags, value)
      end
      value
    end
  end
end

非事件产生的指标

按照其README所说，对于一些非事件产生的指标，可以使用collect来定义

Yabeda.configure do
  # This block will be executed periodically few times in a minute
  # (by timer or external request depending on adapter you're using)
  # Keep it fast and simple!
  collect do
    your_app.whistles_active.set({}, Whistle.where(state: :active).count)
  end
end

当监控系统例如yabeda-prometheus拉取数据时，会调用collect!方法，计算上述指标

# yabeda-0.11.0/lib/yabeda.rb
def collect!
  collectors.each do |collector|
    if config.debug?
      yabeda.collect_duration.measure({ location: collector.source_location.join(":") }, &collector)
    else
      collector.call
    end
  end
end