Refactor gun pooling and simplify adapter option insertion

This patch refactors gun pooling to use Elixir process registry and
simplifies adapter option insertion.

Having the pool use process registry instead of a GenServer has a number of advantages:
- Simpler code: the initial implementation adds about half the lines of code it deletes
- Concurrency: unlike a GenServer, ETS-based registry can handle multiple checkout/checkin
requests at the same time
- Precise and easy idle connection clousure: current proposal for closing idle connections in
the GenServer-based pool needs to filter through all connections once a minute and compare their
last active time with closing time. With Elixir process registry this can be done
by just using `Process.send_after`/`Process.cancel_timer` in the worker process.
- Lower memory footprint: In my tests `gun-memory-leak` branch uses about 290mb on peak load (250 connections)
and 235mb on idle (5-10 connections). Registry-based pool uses 210mb on idle and 240mb on peak load
This commit is contained in:
rinpatch 2020-05-06 01:51:10 +03:00
commit 58a4f350a8
16 changed files with 402 additions and 686 deletions

View file

@ -3,40 +3,11 @@
# SPDX-License-Identifier: AGPL-3.0-only
defmodule Pleroma.Gun.Conn do
@moduledoc """
Struct for gun connection data
"""
alias Pleroma.Gun
alias Pleroma.Pool.Connections
require Logger
@type gun_state :: :up | :down
@type conn_state :: :active | :idle
@type t :: %__MODULE__{
conn: pid(),
gun_state: gun_state(),
conn_state: conn_state(),
used_by: [pid()],
last_reference: pos_integer(),
crf: float(),
retries: pos_integer()
}
defstruct conn: nil,
gun_state: :open,
conn_state: :init,
used_by: [],
last_reference: 0,
crf: 1,
retries: 0
@spec open(String.t() | URI.t(), atom(), keyword()) :: :ok | nil
def open(url, name, opts \\ [])
def open(url, name, opts) when is_binary(url), do: open(URI.parse(url), name, opts)
def open(%URI{} = uri, name, opts) do
def open(%URI{} = uri, opts) do
pool_opts = Pleroma.Config.get([:connections_pool], [])
opts =
@ -45,30 +16,10 @@ defmodule Pleroma.Gun.Conn do
|> Map.put_new(:retry, pool_opts[:retry] || 1)
|> Map.put_new(:retry_timeout, pool_opts[:retry_timeout] || 1000)
|> Map.put_new(:await_up_timeout, pool_opts[:await_up_timeout] || 5_000)
|> Map.put_new(:supervise, false)
|> maybe_add_tls_opts(uri)
key = "#{uri.scheme}:#{uri.host}:#{uri.port}"
max_connections = pool_opts[:max_connections] || 250
conn_pid =
if Connections.count(name) < max_connections do
do_open(uri, opts)
else
close_least_used_and_do_open(name, uri, opts)
end
if is_pid(conn_pid) do
conn = %Pleroma.Gun.Conn{
conn: conn_pid,
gun_state: :up,
conn_state: :active,
last_reference: :os.system_time(:second)
}
:ok = Gun.set_owner(conn_pid, Process.whereis(name))
Connections.add_conn(name, key, conn)
end
do_open(uri, opts)
end
defp maybe_add_tls_opts(opts, %URI{scheme: "http"}), do: opts
@ -81,7 +32,7 @@ defmodule Pleroma.Gun.Conn do
reuse_sessions: false,
verify_fun:
{&:ssl_verify_hostname.verify_fun/3,
[check_hostname: Pleroma.HTTP.Connection.format_host(host)]}
[check_hostname: Pleroma.HTTP.AdapterHelper.format_host(host)]}
]
tls_opts =
@ -105,7 +56,7 @@ defmodule Pleroma.Gun.Conn do
{:ok, _} <- Gun.await_up(conn, opts[:await_up_timeout]),
stream <- Gun.connect(conn, connect_opts),
{:response, :fin, 200, _} <- Gun.await(conn, stream) do
conn
{:ok, conn}
else
error ->
Logger.warn(
@ -141,7 +92,7 @@ defmodule Pleroma.Gun.Conn do
with {:ok, conn} <- Gun.open(proxy_host, proxy_port, opts),
{:ok, _} <- Gun.await_up(conn, opts[:await_up_timeout]) do
conn
{:ok, conn}
else
error ->
Logger.warn(
@ -155,11 +106,11 @@ defmodule Pleroma.Gun.Conn do
end
defp do_open(%URI{host: host, port: port} = uri, opts) do
host = Pleroma.HTTP.Connection.parse_host(host)
host = Pleroma.HTTP.AdapterHelper.parse_host(host)
with {:ok, conn} <- Gun.open(host, port, opts),
{:ok, _} <- Gun.await_up(conn, opts[:await_up_timeout]) do
conn
{:ok, conn}
else
error ->
Logger.warn(
@ -171,7 +122,7 @@ defmodule Pleroma.Gun.Conn do
end
defp destination_opts(%URI{host: host, port: port}) do
host = Pleroma.HTTP.Connection.parse_host(host)
host = Pleroma.HTTP.AdapterHelper.parse_host(host)
%{host: host, port: port}
end
@ -181,17 +132,6 @@ defmodule Pleroma.Gun.Conn do
defp add_http2_opts(opts, _, _), do: opts
defp close_least_used_and_do_open(name, uri, opts) do
with [{key, conn} | _conns] <- Connections.get_unused_conns(name),
:ok <- Gun.close(conn.conn) do
Connections.remove_conn(name, key)
do_open(uri, opts)
else
[] -> {:error, :pool_overflowed}
end
end
def compose_uri_log(%URI{scheme: scheme, host: host, path: path}) do
"#{scheme}://#{host}#{path}"
end

View file

@ -0,0 +1,129 @@
defmodule Pleroma.Gun.ConnectionPool do
@registry __MODULE__
def get_conn(uri, opts) do
case enforce_pool_limits() do
:ok ->
key = "#{uri.scheme}:#{uri.host}:#{uri.port}"
case Registry.lookup(@registry, key) do
# The key has already been registered, but connection is not up yet
[{worker_pid, {nil, _used_by, _crf, _last_reference}}] ->
get_gun_pid_from_worker(worker_pid)
[{worker_pid, {gun_pid, _used_by, _crf, _last_reference}}] ->
GenServer.cast(worker_pid, {:add_client, self(), false})
{:ok, gun_pid}
[] ->
# :gun.set_owner fails in :connected state for whatevever reason,
# so we open the connection in the process directly and send it's pid back
# We trust gun to handle timeouts by itself
case GenServer.start(Pleroma.Gun.ConnectionPool.Worker, [uri, key, opts, self()],
timeout: :infinity
) do
{:ok, _worker_pid} ->
receive do
{:conn_pid, pid} -> {:ok, pid}
end
{:error, {:error, {:already_registered, worker_pid}}} ->
get_gun_pid_from_worker(worker_pid)
err ->
err
end
end
:error ->
{:error, :pool_full}
end
end
@enforcer_key "enforcer"
defp enforce_pool_limits() do
max_connections = Pleroma.Config.get([:connections_pool, :max_connections])
if Registry.count(@registry) >= max_connections do
case Registry.lookup(@registry, @enforcer_key) do
[] ->
pid =
spawn(fn ->
{:ok, _pid} = Registry.register(@registry, @enforcer_key, nil)
reclaim_max =
[:connections_pool, :reclaim_multiplier]
|> Pleroma.Config.get()
|> Kernel.*(max_connections)
|> round
|> max(1)
unused_conns =
Registry.select(
@registry,
[
{{:_, :"$1", {:_, :"$2", :"$3", :"$4"}}, [{:==, :"$2", []}],
[{{:"$1", :"$3", :"$4"}}]}
]
)
case unused_conns do
[] ->
exit(:pool_full)
unused_conns ->
unused_conns
|> Enum.sort(fn {_pid1, crf1, last_reference1},
{_pid2, crf2, last_reference2} ->
crf1 <= crf2 and last_reference1 <= last_reference2
end)
|> Enum.take(reclaim_max)
|> Enum.each(fn {pid, _, _} -> GenServer.call(pid, :idle_close) end)
end
end)
wait_for_enforcer_finish(pid)
[{pid, _}] ->
wait_for_enforcer_finish(pid)
end
else
:ok
end
end
defp wait_for_enforcer_finish(pid) do
ref = Process.monitor(pid)
receive do
{:DOWN, ^ref, :process, ^pid, :pool_full} ->
:error
{:DOWN, ^ref, :process, ^pid, :normal} ->
:ok
end
end
defp get_gun_pid_from_worker(worker_pid) do
# GenServer.call will block the process for timeout length if
# the server crashes on startup (which will happen if gun fails to connect)
# so instead we use cast + monitor
ref = Process.monitor(worker_pid)
GenServer.cast(worker_pid, {:add_client, self(), true})
receive do
{:conn_pid, pid} -> {:ok, pid}
{:DOWN, ^ref, :process, ^worker_pid, reason} -> reason
end
end
def release_conn(conn_pid) do
[worker_pid] =
Registry.select(@registry, [
{{:_, :"$1", {:"$2", :_, :_, :_}}, [{:==, :"$2", conn_pid}], [:"$1"]}
])
GenServer.cast(worker_pid, {:remove_client, self()})
end
end

View file

@ -0,0 +1,95 @@
defmodule Pleroma.Gun.ConnectionPool.Worker do
alias Pleroma.Gun
use GenServer
@registry Pleroma.Gun.ConnectionPool
@impl true
def init([uri, key, opts, client_pid]) do
time = :os.system_time(:second)
# Register before opening connection to prevent race conditions
with {:ok, _owner} <- Registry.register(@registry, key, {nil, [client_pid], 1, time}),
{:ok, conn_pid} <- Gun.Conn.open(uri, opts),
Process.link(conn_pid) do
{_, _} =
Registry.update_value(@registry, key, fn {_, used_by, crf, last_reference} ->
{conn_pid, used_by, crf, last_reference}
end)
send(client_pid, {:conn_pid, conn_pid})
{:ok, %{key: key, timer: nil}, :hibernate}
else
err -> {:stop, err}
end
end
@impl true
def handle_cast({:add_client, client_pid, send_pid_back}, %{key: key} = state) do
time = :os.system_time(:second)
{{conn_pid, _, _, _}, _} =
Registry.update_value(@registry, key, fn {conn_pid, used_by, crf, last_reference} ->
{conn_pid, [client_pid | used_by], crf(time - last_reference, crf), time}
end)
if send_pid_back, do: send(client_pid, {:conn_pid, conn_pid})
state =
if state.timer != nil do
Process.cancel_timer(state[:timer])
%{state | timer: nil}
else
state
end
{:noreply, state, :hibernate}
end
@impl true
def handle_cast({:remove_client, client_pid}, %{key: key} = state) do
{{_conn_pid, used_by, _crf, _last_reference}, _} =
Registry.update_value(@registry, key, fn {conn_pid, used_by, crf, last_reference} ->
{conn_pid, List.delete(used_by, client_pid), crf, last_reference}
end)
timer =
if used_by == [] do
max_idle = Pleroma.Config.get([:connections_pool, :max_idle_time], 30_000)
Process.send_after(self(), :idle_close, max_idle)
else
nil
end
{:noreply, %{state | timer: timer}, :hibernate}
end
@impl true
def handle_info(:idle_close, state) do
# Gun monitors the owner process, and will close the connection automatically
# when it's terminated
{:stop, :normal, state}
end
# Gracefully shutdown if the connection got closed without any streams left
@impl true
def handle_info({:gun_down, _pid, _protocol, _reason, []}, state) do
{:stop, :normal, state}
end
# Otherwise, shutdown with an error
@impl true
def handle_info({:gun_down, _pid, _protocol, _reason, _killed_streams} = down_message, state) do
{:stop, {:error, down_message}, state}
end
@impl true
def handle_call(:idle_close, _, %{key: key} = state) do
Registry.unregister(@registry, key)
{:stop, :normal, state}
end
# LRFU policy: https://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.55.1478
defp crf(time_delta, prev_crf) do
1 + :math.pow(0.5, time_delta / 100) * prev_crf
end
end