Plug.Telemetry.ServerTiming (plug_telemetry_server_timing v0.2.0) View Source

This plug provide support for Server-Timing header that allows to display server-side measurements in browser developer tools and access it programatically via Performance API in JavaScript.

Usage

Just add it as a plug into your pipeline:

plug Plug.Telemetry.ServerTiming

And call install/1 with list of {event_name, measurement} in your application startup, for example for Phoenix and Ecto application:

Plug.Telemetry.ServerTiming.install([
  {[:phoenix, :endpoint, :stop], :duration},
  {[:my_app, :repo, :query], :queue_time},
  {[:my_app, :repo, :query], :query_time},
  {[:my_app, :repo, :query], :decode_time}
])

Important

You need to place this plug BEFORE Plug.Telemetry call as otherwise it will not see it's events (before_send callbacks are called in reverse order of declaration, so this one need to be added before Plug.Telemetry one.

Caveats

This will not respond with events that happened in separate processes, only events that happened in the Plug process will be recorded.

WARNING

Current specification of Server-Timing do not provide a way to specify event start time, which mean, that the data displayed in the DevTools isn't trace report (like the content of the "regular" HTTP timings) but raw dump of the data displayed as a bars. This can be a little bit confusing, but right now there is nothing I can do about it.

Link to this section Summary

Types

event()

events()

Functions

install(events)

Define which events should be available within response headers.

Link to this section Types

event()

Specs

event() ::
  {:telemetry.event_name(), atom()}
  | {:telemetry.event_name(), atom(), keyword() | map()}

events()

Specs

events() :: [event()]

Link to this section Functions

install(events)

Specs

install(events) :: :ok when events: map() | [{:telemetry.event_name(), atom()}]

Define which events should be available within response headers.