Upgrading ecto - telemetry integration

Questions / Help
,
1

I couldn’t find in ecto docs or elsewhere how to upgrade my instrumentation after telemetry.

Today I have the code below.

Do you know a guide that could help me with that upgrade? Or what would be the equivalent code with telemetry?

I still want to use the functions provided by my module MyApp.Statix.

Thanks!

config :my_app, MyApp.Repo,
  loggers: [{Ecto.LogEntry, :log, []}, {MyApp.Statix.Repo, :log, []}],
  # ...

defmodule MyApp.Statix.Repo do
  import MyApp.Statix, only: [histogram: 3, increment: 3, tags: 1]
  alias Ecto.LogEntry

  @moduledoc "Provides Ecto Repo stats"

  @query_time "db.query_exec_time"
  @queue_time "db.query_queue_time"
  @decode_time "db.query_decode_time"
  @count "db.query_count"

  def log(
        %LogEntry{
          query_time: query,
          queue_time: queue,
          decode_time: decode
        } = entry
      ) do
    with {:ok, tags} when is_list(tags) <- {:db, entry} |> tags(),
         opts = [tags: tags],
         :ok <- @query_time |> histogram((query || 0) + (queue || 0), opts),
         :ok <- @queue_time |> histogram(queue || 0, opts),
         :ok <- @decode_time |> histogram(decode || 0, opts),
         :ok <- @count |> increment(1, opts) do
      entry
    end
  end
end
2 Likes
2

As co-maintainer of Spandex, and specifically spandex_ecto, I’m also keenly interested in what the story is around Ecto and Telemetry in v3.0. I just spent about an hour looking through the ecto, ecto_sql, and postgrex source code to figure out what the story was, and what I’ve found so far is:

  1. The loggers option seems to no longer be there on the Ecto.Repo config (commit that removed it)

  2. The changelog and release announcement don’t list this as a breaking change, but rather a deprecation in favor of using Telemetry instead. I haven’t confirmed yet what happens if you specify the loggers option like before. Maybe it still works?

  3. The Ecto.Repo docs recommend that adapters send [:my_app, :repo, :query] Telemetry events, containing at least the data that Ecto.LogEntry used to send to loggers, but I’m not sure where to find the actual list of events that get sent and what the metadata will be. I am guessing that it’s currently only this, which is the same thing that you used to get from the loggers call-backs. Maybe I missed something or the docs just need some clarification on where to go to find more details?

  4. Edit: Also, caller_pid was removed from the Ecto.LogEntry struct, so I’m not sure how to work around that in spandex_ecto.

3 Likes
3

To consume telemetry events, see Telemetry’s README.

For MyApp.Repo, here is how would would do it:

defmodule MyApp.Telemetry do
  def handle_event([:my_app, :repo, :query], time, metadata, config) do
    IO.inspect binding()
  end
end

and then attach this module on your Application start callback:

Telemetry.attach("my-app-handler", [:my_app, :repo, :query], MyApp.Telemetry, :handle_event,%{})

I will make sure to improve the docs.

Regarding caller_pid, it was also asked in the issues tracker here.

6 Likes
4

I am wondering, how it will work when I have multiple repos in single application? Will all be logged into the same event_name? Also how to listen on all events of repo at once? Is there any solution for listening to multiple events at once? Like using prefix instead of full name?

1 Like
5

The event name is based on the module name. So if you have MyApp.AnotherRepo, it will be [:my_app, :another_repo, :query]. You can also configure the prefix in the repo.

4 Likes