ezmetrics

Simple, lightweight and fast metrics aggregation for Rails.

Installation

gem 'ezmetrics'

Available metrics

Type	Aggregate functions
duration	avg, max, percentile
views	avg, max, percentile
db	avg, max, percentile
queries	avg, max, percentile
requests	all, 2xx, 3xx, 4xx, 5xx

Usage

Getting started

This tool captures and aggregates Rails application metrics such as

• duration
• views
• db
• queries
• status

and stores them for the timeframe you specified, 60 seconds by default.

You can change the timeframe according to your needs and save the metrics by calling log method:

# Store the metrics for 60 seconds (default behaviour)
EZmetrics.new.log(
  duration: 100.5,
  views:    40.7,
  db:       59.8,
  queries:  4,
  status:   200
)

# Store the metrics for 10 minutes
EZmetrics.new(10.minutes).log(
  duration: 100.5,
  views:    40.7,
  db:       59.8,
  queries:  4,
  status:   200
)

---

For displaying metrics you need to call show method:

# Aggregate and show metrics for last 60 seconds (default behaviour)
EZmetrics.new.show

# Aggregate and show metrics for last 10 minutes
EZmetrics.new(10.minutes).show

You can combine these timeframes, for example - store for 10 minutes, display for 5 minutes.

Capture metrics

Just add an initializer to your application:

# config/initializers/ezmetrics.rb

ActiveSupport::Notifications.subscribe("sql.active_record") do |*args|
  event = ActiveSupport::Notifications::Event.new(*args)
  unless event.payload[:name] == "SCHEMA"
    Thread.current[:queries] ||= 0
    Thread.current[:queries] += 1
  end
end

ActiveSupport::Notifications.subscribe("process_action.action_controller") do |*args|
  event = ActiveSupport::Notifications::Event.new(*args)
  EZmetrics.new.log(
    duration: event.duration.to_f,
    views:    event.payload[:view_runtime].to_f,
    db:       event.payload[:db_runtime].to_f,
    status:   event.payload[:status].to_i || 500,
    queries:  Thread.current[:queries].to_i,
  )
end

Display metrics

As simple as:

EZmetrics.new.show

This will return a hash with the following structure:

{
  duration: {
    avg: 5569,
    max: 9675
  },
  views: {
    avg: 12,
    max: 240
  },
  db: {
    avg: 155,
    max: 4382
  },
  queries: {
    avg: 26,
    max: 76
  },
  requests: {
    all: 2000,
    grouped: {
      "2xx" => 1900,
      "3xx" => 15,
      "4xx" => 80,
      "5xx" => 5
    }
  }
}

---

If you prefer a single level object - you can change the default output structure by calling .flatten before .show

EZmetrics.new(1.hour).flatten.show(db: :avg, duration: [:avg, :max])

{
  db_avg:       182,
  duration_avg: 205,
  duration_max: 5171
}

---

Same for partitioned aggregation

EZmetrics.new(1.hour).partition_by(:minute).flatten.show(db: :avg, duration: [:avg, :max])

[
  {
    timestamp:    1575242880,
    db_avg:       387,
    duration_avg: 477,
    duration_max: 8566
  },
  {
    timestamp:    1575242940,
    db_avg:       123,
    duration_avg: 234,
    duration_max: 3675
  }
]

Aggregation

The aggregation can be easily configured by specifying aggregation options as in the following examples:

1. Single

EZmetrics.new.show(duration: :max)

{
  duration: {
    max: 9675
  }
}

---

2. Multiple

EZmetrics.new.show(queries: [:max, :avg])

{
  queries: {
    max: 76,
    avg: 26
  }
}

---

3. Requests

EZmetrics.new.show(requests: true)

{
  requests: {
    all: 2000,
    grouped: {
      "2xx" => 1900,
      "3xx" => 15,
      "4xx" => 80,
      "5xx" => 5
    }
  }
}

---

4. Combined

EZmetrics.new.show(views: :avg, :db: [:avg, :max], requests: true)

{
  views: {
    avg: 12
  },
  db: {
    avg: 155,
    max: 4382
  },
  requests: {
    all: 2000,
    grouped: {
      "2xx" => 1900,
      "3xx" => 15,
      "4xx" => 80,
      "5xx" => 5
    }
  }
}

---

5. Percentile

This feature is available since version 2.0.0.

By default percentile aggregation is turned off because it requires to store each value of all metrics.

To enable this feature - you need to set store_each_value: true when saving the metrics:

EZmetrics.new.log(
  duration:         100.5,
  views:            40.7,
  db:               59.8,
  queries:          4,
  status:           200,
  store_each_value: true
)

The aggregation syntax has the following format metrics_type: :percentile_{number} where number is any integer in the 1..99 range.

EZmetrics.new.show(db: [:avg, :percentile_90, :percentile_95], duration: :percentile_99)

{
  db: {
    avg: 155,
    percentile_90: 205,
    percentile_95: 215
  },
  duration: {
    percentile_99: 236
  }
}

6. Percentile distribution

If you want to visualize percentile distribution (from 1% to 99%):

EZmetrics.new.show(duration: :percentile_distribution)

{
  duration: {
    percentile_distribution: {
      1: 12,
      2: 15,
      3: 19,
      #...
      97: 6540,
      98: 6682,
      99: 6730
    }
  }
}

Partitioning

If you want to visualize your metrics by using a line chart, you will need to use partitioning.

To aggregate metrics, partitioned by a unit of time you need to call .partition_by({time_unit}) before calling .show

# Aggregate metrics for last hour, partition by minute
EZmetrics.new(1.hour).partition_by(:minute).show(duration: [:avg, :max], db: :avg)

This will return an array of objects with the following structure:

[
  {
    timestamp: # UNIX timestamp
    data:      # a hash with aggregated metrics
  }
]

like in the example below:

[
  {
    timestamp: 1575242880,
    data: {
      duration: {
        avg: 477,
        max: 8566
      },
      db: {
        avg: 387
      }
    }
  },
  {
    timestamp: 1575242940,
    data: {
      duration: {
        avg: 234,
        max: 3675
      },
      db: {
        avg: 123
      }
    }
  }
]

Available time units for partitioning: second, minute, hour, day. Default: minute.

Performance

The aggregation speed relies on the performance of Redis (data storage) and Oj (json serialization/parsing).

You can check the aggregation time (in seconds) by running:

# 1. Simple
EZmetrics::Benchmark.new.measure_aggregation

# 2. Partitioned
EZmetrics::Benchmark.new.measure_aggregation(:minute)

# 3. Percentile
EZmetrics::Benchmark.new(true).measure_aggregation

# 4.  Percentile (partitioned)
EZmetrics::Benchmark.new(true).measure_aggregation(:minute)

Interval	Simple aggregation	Partitioned	Percentile	Percentile (partitioned)
1 minute	0.0	0.0	0.0	0.0
1 hour	0.02	0.02	0.14	0.16
12 hours	0.22	0.25	2.11	1.97
24 hours	0.61	0.78	5.85	5.85
48 hours	1.42	1.75	14.1	13.9

The benchmarks above were run on a 2017 Macbook Pro 2.9 GHz Intel Core i7 with 16 GB of RAM

License

ezmetrics is released under the MIT License.