Subscribe Free

Join 2670+ others. No spamming.
I promise!

We are currently under high development. Follow us at github.

Looking for Python Tutorials?
Check these awesome tutorials



hahuang65 / beaker


Statistics and Metrics library for Elixir



Build Status Inline docs

Beaker is a tool that can be used to keep track of metrics for your Elixir project. It aims to provide an easy way to register statistics as well as an easy way to visualize them.

Note: Beaker metrics are currently ephemeral and are scoped to the app (or more specifically Beaker) being stopped. Metrics are not persisted across restarts.

For more information, see the online documentation.


To include Beaker in your application, add it to your mix.exs file:

defp deps do
    {:beaker, ">= 1.2.0"}

And add it to the list of applications started with yours:

def application do
  [applications: [:beaker]]

Integration with Phoenix

Beaker provides a way to visualize your metrics through Phoenix.

It'll end up looking something like:


The source code for this frontend can be found at Beaker-Frontend! The API Documentation which the frontend consumes can be found at API Documentation

1) Add beaker and phoenix to the dependencies in your Mixfile:

defp deps do
    {:phoenix, ">= 1.1"},
    {:phoenix_html, ">= 2.3"}

2) Add beaker and phoenix_html to the started applications in your Mixfile:

def application do
  applications: [:phoenix, :phoenix_html, :beaker]

3) Forward requests to Beaker.Web in your router:

forward "/beaker", Beaker.Web

This will add a page at /beaker with all your metrics visualized on the page. Gauges and Counters will display a box with their name and value. Time Series will display a chart with the last 120 minutes worth of aggregated data.

If you'd like to track your Phoenix performance, you can add that to your Phoenix Endpoint:

defmodule MyApp.Endpoint do
  use Phoenix.Endpoint, otp_app: :my_app

  plug Beaker.Integrations.Phoenix

  plug ...

This will keep track ALL requests (including requests for static assets) passing through Phoenix, and keep track of their response time and keep a counter of how many requests were made. NOTE: It is EXTREMELY important that this is plugged before anything else in order to get the most accurate response timings.

Currently, these are not in-depth to split apart requests going to specific controllers, but provides a good overview to see how your overall Phoenix performance is. Hopefully in-depth tracking of requests can be implemented in a future release. Also, it'll be nice to be able to provide an option to ignore serving of static assets.

Integration with Ecto

Beaker provides a simple way to integrate with Ecto to track the performance of your queries.

To use it, in your Repo, just:

defmodule MyApp.Repo do
  use Ecto.Repo, otp_app: :my_app
  use Beaker.Integrations.Ecto

Doing so will keep track of your queries times, queue times, and query counts for all your Ecto queries automatically. These will show up in Beaker's web interface.

Currently, these are not in-depth to split apart different types of queries, but provides a good overview to see how your overall Ecto performance is. Hopefully in-depth tracking of queries can be implemented in a future release.


Beaker provides a variety of different metric types:


The gauge is a simple gauge. It's a metric where a value can be set and retrieved.

It is commonly used for metrics that return a single value. Examples are:

  • Average response time
  • Uptime (Availability)
  • Latency / Ping

You can set and retrieve the value of a gauge like so:

iex> Beaker.Gauge.set("foo", 50)

iex> Beaker.Gauge.get("foo")

You can also set a minimum and maximum value for the gauge which can be used by the frontend for displaying your data.

iex> Beaker.Gauge.set("foo-min-max", 50, 1, 100)

iex> Beaker.Gauge.get("foo-min-max")
%{max: 100, min: 1, name: "foo-min-max", value: 50}

Sometimes you'll want to time something, and set that duration to a gauge. We provide a convenience for that.

iex> Beaker.Gauge.time("foo", fn -> 2 + 2 end)
# Or if you prefer `do` syntax:
Beaker.Gauge.time("foo") do
  2 + 2

You can get all of your gauges in the form of a map if you need:

iex> Beaker.Gauge.all
%{"foo" => 10, "bar" => 45}

You can remove all your gauges and start from scratch:

iex> Beaker.Gauge.clear

Or just clear out a single gauge:

iex> Beaker.Gauge.clear("foo")


The counter is a signed bi-directional integer counter. It can keep track of integers and increment and decrement them.

It is commonly used for metrics that keep track of some cumulative value. Examples are:

  • Total number of downloads
  • Number of queued jobs
  • Quotas

You can set and retrieve the value of a counter like so:

iex> Beaker.Counter.set("foo", 10)

iex> Beaker.Counter.get("foo")

You can also use a counter more traditionally, via incrementing and decrementing:

iex> Beaker.Counter.incr("foo")

iex> Beaker.Counter.get("foo")

iex> Beaker.Counter.decr("foo")

iex> Beaker.Counter.get("foo")

If incrementing and decrementing by 1 is not a big enough step for you:

iex> Beaker.Counter.incr_by("foo", 5)

iex> Beaker.Counter.get("foo")

iex> Beaker.Counter.decr_by("foo", 10)

iex> Beaker.Counter.get("foo")

You can get all of your counters in the form of a map if you need:

iex> Beaker.Counter.all
%{"foo" => 10, "bar" => 45}

You can remove all your counters and start from scratch:

iex> Beaker.Counter.clear

Or just clear out a single counter:

iex> Beaker.Counter.clear("foo")

Time Series

The time series is basically a series of values with a time (epoch) attached to each value at the time the value was recorded.

It is commonly used to keep track of the change in value of some metric across a period of time. Examples are:

  • Response time across a period of time
  • Error rates across a period of time
  • Download count across a period of time

To sample (record a value) a time series:

iex> Beaker.TimeSeries.sample("foo", 50)
iex> Beaker.TimeSeries.sample("foo", 66)
iex> Beaker.TimeSeries.sample("foo", 30)
iex> Beaker.TimeSeries.sample("bar", 10)
iex> Beaker.TimeSeries.sample("bar", 50)

Sometimes you'll want to time something, and sample that duration to a time series. We provide a convenience for that.

iex> Beaker.TimeSeries.time("baz", fn -> 2 + 2 end)
# Or if you prefer `do` block syntax:
Beaker.TimeSeries.time("baz") do
  2 + 2

Anytime a time series is retrieved, it will be in the format of a list of pairs. Each pair consists of a timestamp in epoch and the value sampled, i.e. {timestamp, value}. The list will be guaranteed to be in reverse chronological order, that is, the latest sample will be the first in the list.

To get the time series that have been recorded for a key:

iex> Beaker.TimeSeries.get("foo")
[{1434738115306786, 30}, {1434738112851607, 66}, {1434738107132294, 50}]

To retrieve all time series:

iex> Beaker.TimeSeries.all
%{"bar" => [{1434738203344586, 50}, {1434738201507329, 10}],
  "foo" => [{1434738115306786, 32}, {1434738112851607, 87}, {1434738107132294, 50}]}

And to clear all time series:

iex> Beaker.TimeSeries.clear

Or clear a single time series:

iex> Beaker.TimeSeries.clear("foo")

Time Series Aggregation

Time series will be aggregated once every 60 seconds, for the last full minute (e.g. if aggregation is run at 01:22:32, it will run for the minute of 01:21 to 01:22).

For each minute, aggregation will calculate the minimum, maximum, and average values as well as the number of values for that minute and store it in Beaker.TimeSeries.Aggregated.

Note: Aggregation is not destructive. Raw data will remain in Beaker.TimeSeries, and the calculated values from aggregation will be stored in Beaker.TimeSeries.Aggregated.

To inspect and use aggregated data, the Beaker.TimeSeries.Aggregated.get/1 and Beaker.TimeSeries.Aggregated.all/0 functions are available and function exactly as their Beaker.TimeSeries counterparts. However, please be aware that although aggregated data is also returned as a list of {time, value} pairs, value is actually a tuple that looks like {average, minimum, maximum, count} for it's paired minute.

iex> Beaker.TimeSeries.Aggregated.get("foo")
[{1434738060000000, {48.666666666666664, 30, 66, 3}}] # 48.666 is the average, 30 is the minimum, 66 is the maximum, and 3 is the count of entries for the minute of 1434738060000000 in epoch time

To check the last time aggregation was successfully run:

iex> Beaker.TimeSeries.Aggregator.last_aggregated_at
1442989860000000 # Epoch time

Performance: Currently, the aggregation algorithm is not very optimized. I did some basic performance testing on a 2014 MacBook Pro with 16GB of RAM and a 3 Ghz i7 processor.

600 data points within a minute (~10 samples per second) will take each aggregation roughly between 3 to 6 milliseconds.

600 data points within a minute for 10 time series (6000 data points total) will take each aggregation roughly between 5 and 15 milliseconds.

I hope to improve this algorithm to be faster in the next few releases.

Important Links


DOC=1 mix test will create a markdown file representing the request and response for the API.