• Stars
    star
    123
  • Rank 290,145 (Top 6 %)
  • Language
    Erlang
  • License
    Apache License 2.0
  • Created about 6 years ago
  • Updated 3 months ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

Periodically gather measurements and publish them as Telemetry events

telemetry_poller

Codecov

Allows to periodically collect measurements and dispatch them as Telemetry events.

telemetry_poller by default runs a poller to perform VM measurements:

  • [vm, memory] - contains the total memory, process memory, and all other keys in erlang:memory/0
  • [vm, total_run_queue_lengths] - returns the run queue lengths for CPU and IO schedulers. It contains the total, cpu and io measurements
  • [vm, system_counts] - returns the current process, atom and port count as per erlang:system_info/1

You can directly consume those events after adding telemetry_poller as a dependency.

Poller also provides a convenient API for running custom pollers.

Defining custom measurements

Poller also includes conveniences for performing process-based measurements as well as custom ones.

Erlang

First define the poller with the custom measurements. The first measurement is the built-in process_info measurement and the second one is given by a custom module-function-args defined by you:

telemetry_poller:start_link(
  [{measurements, [
    {process_info, [{name, my_app_worker}, {event, [my_app, worker]}, {keys, [memory, message_queue_len]}]},
    {example_app_measurements, dispatch_session_count, []}
  ]},
  {period, timer:seconds(10)}, % configure sampling period - default is timer:seconds(5)
  {name, my_app_poller}
]).

Now define the custom measurement and you are good to go:

-module(example_app_measurements).

dispatch_session_count() ->
    % emit a telemetry event when called
    telemetry:execute([example_app, session_count], #{count => example_app:session_count()}, #{}).

Elixir

You typically start the poller as a child in your supervision tree:

children = [
  {:telemetry_poller,
   # include custom measurement as an MFA tuple
   measurements: [
     {:process_info, name: :my_app_worker, event: [:my_app, :worker], keys: [:memory, :message_queue_len]},
     {ExampleApp.Measurements, :dispatch_session_count, []},
   ],
   period: :timer.seconds(10), # configure sampling period - default is :timer.seconds(5)
   name: :my_app_poller}
]

Supervisor.start_link(children, strategy: :one_for_one)

The poller above has two periodic measurements. The first is the built-in process_info measurement that will gather the memory and message queue length of a process. The second is given by a custom module-function-args defined by you, such as below:

defmodule ExampleApp.Measurements do
  def dispatch_session_count() do
    # emit a telemetry event when called
    :telemetry.execute([:example_app, :session_count], %{count: ExampleApp.session_count()}, %{})
  end
end

Documentation

See documentation for more concrete examples and usage instructions.

Copyright and License

telemetry_poller is copyright (c) 2018 Chris McCord and Erlang Solutions.

telemetry_poller source code is released under Apache License, Version 2.0.

See LICENSE and NOTICE files for more information.