From 46236d1d873473d95b11cd7bfdcaa284ea55a9ad Mon Sep 17 00:00:00 2001 From: rinpatch Date: Wed, 2 Sep 2020 12:42:25 +0300 Subject: [PATCH 1/3] html.ex: optimize external url extraction By using a :not() selector and only extracting attributes from the first match. --- lib/pleroma/html.ex | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/lib/pleroma/html.ex b/lib/pleroma/html.ex index dc1b9b840..20b02f091 100644 --- a/lib/pleroma/html.ex +++ b/lib/pleroma/html.ex @@ -109,8 +109,9 @@ defmodule Pleroma.HTML do result = content |> Floki.parse_fragment!() - |> Floki.filter_out("a.mention,a.hashtag,a.attachment,a[rel~=\"tag\"]") - |> Floki.attribute("a", "href") + |> Floki.find("a:not(.mention,.hashtag,.attachment,[rel~=\"tag\"])") + |> Enum.take(1) + |> Floki.attribute("href") |> Enum.at(0) {:commit, {:ok, result}} From 19691389b92e617f1edad7d4e3168fe917d0a675 Mon Sep 17 00:00:00 2001 From: rinpatch Date: Wed, 2 Sep 2020 14:21:28 +0300 Subject: [PATCH 2/3] Rich media: Add failure tracking --- CHANGELOG.md | 3 ++ config/config.exs | 1 + config/description.exs | 7 +++ docs/configuration/cheatsheet.md | 1 + lib/pleroma/web/rich_media/parser.ex | 56 ++++++++++++--------- test/web/rich_media/aws_signed_url_test.exs | 2 +- 6 files changed, 46 insertions(+), 24 deletions(-) diff --git a/CHANGELOG.md b/CHANGELOG.md index 07bc6d77c..4424c060f 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -5,6 +5,9 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/). ## unreleased-patch - ??? +### Added +- Rich media failure tracking (along with `:failure_backoff` option) + ### Fixed - Mastodon API: Search parameter `following` now correctly returns the followings rather than the followers diff --git a/config/config.exs b/config/config.exs index ea8869d48..ed37b93c0 100644 --- a/config/config.exs +++ b/config/config.exs @@ -412,6 +412,7 @@ config :pleroma, :rich_media, Pleroma.Web.RichMedia.Parsers.TwitterCard, Pleroma.Web.RichMedia.Parsers.OEmbed ], + failure_backoff: 60_000, ttl_setters: [Pleroma.Web.RichMedia.Parser.TTL.AwsSignedUrl] config :pleroma, :media_proxy, diff --git a/config/description.exs b/config/description.exs index 29a657333..5e08ba109 100644 --- a/config/description.exs +++ b/config/description.exs @@ -2385,6 +2385,13 @@ config :pleroma, :config_description, [ suggestions: [ Pleroma.Web.RichMedia.Parser.TTL.AwsSignedUrl ] + }, + %{ + key: :failure_backoff, + type: :integer, + description: + "Amount of milliseconds after request failure, during which the request will not be retried.", + suggestions: [60_000] } ] }, diff --git a/docs/configuration/cheatsheet.md b/docs/configuration/cheatsheet.md index 2f440adf4..a9a650fab 100644 --- a/docs/configuration/cheatsheet.md +++ b/docs/configuration/cheatsheet.md @@ -361,6 +361,7 @@ config :pleroma, Pleroma.Web.MediaProxy.Invalidation.Http, * `ignore_hosts`: list of hosts which will be ignored by the metadata parser. For example `["accounts.google.com", "xss.website"]`, defaults to `[]`. * `ignore_tld`: list TLDs (top-level domains) which will ignore for parse metadata. default is ["local", "localdomain", "lan"]. * `parsers`: list of Rich Media parsers. +* `failure_backoff`: Amount of milliseconds after request failure, during which the request will not be retried. ## HTTP server diff --git a/lib/pleroma/web/rich_media/parser.ex b/lib/pleroma/web/rich_media/parser.ex index e9aa2dd03..e98c743ca 100644 --- a/lib/pleroma/web/rich_media/parser.ex +++ b/lib/pleroma/web/rich_media/parser.ex @@ -17,14 +17,25 @@ defmodule Pleroma.Web.RichMedia.Parser do else @spec parse(String.t()) :: {:ok, map()} | {:error, any()} def parse(url) do - Cachex.fetch!(:rich_media_cache, url, fn _ -> - with {:ok, data} <- parse_url(url) do - {:commit, {:ok, data}} - else - error -> {:ignore, error} - end - end) - |> set_ttl_based_on_image(url) + with {:ok, data} <- get_cached_or_parse(url), + {:ok, _} <- set_ttl_based_on_image(data, url) do + {:ok, data} + else + error -> + Logger.error(fn -> "Rich media error: #{inspect(error)}" end) + end + end + + defp get_cached_or_parse(url) do + case Cachex.fetch!(:rich_media_cache, url, fn _ -> {:commit, parse_url(url)} end) do + {:ok, _data} = res -> + res + + {:error, _} = e -> + ttl = Pleroma.Config.get([:rich_media, :failure_backoff], 60_000) + Cachex.expire(:rich_media_cache, url, ttl) + e + end end end @@ -50,22 +61,21 @@ defmodule Pleroma.Web.RichMedia.Parser do config :pleroma, :rich_media, ttl_setters: [MyModule] """ - @spec set_ttl_based_on_image({:ok, map()} | {:error, any()}, String.t()) :: - {:ok, map()} | {:error, any()} - def set_ttl_based_on_image({:ok, data}, url) do - with {:ok, nil} <- Cachex.ttl(:rich_media_cache, url), - {:ok, ttl} when is_number(ttl) <- get_ttl_from_image(data, url) do - Cachex.expire_at(:rich_media_cache, url, ttl * 1000) - {:ok, data} - else - _ -> - {:ok, data} - end - end + @spec set_ttl_based_on_image(map(), String.t()) :: + {:ok, Integer.t() | :noop} | {:error, :no_key} + def set_ttl_based_on_image(data, url) do + case get_ttl_from_image(data, url) do + {:ok, ttl} when is_number(ttl) -> + ttl = ttl * 1000 - def set_ttl_based_on_image({:error, _} = error, _) do - Logger.error("parsing error: #{inspect(error)}") - error + case Cachex.expire_at(:rich_media_cache, url, ttl) do + {:ok, true} -> {:ok, ttl} + {:ok, false} -> {:error, :no_key} + end + + _ -> + {:ok, :noop} + end end defp get_ttl_from_image(data, url) do diff --git a/test/web/rich_media/aws_signed_url_test.exs b/test/web/rich_media/aws_signed_url_test.exs index a21f3c935..1ceae1a31 100644 --- a/test/web/rich_media/aws_signed_url_test.exs +++ b/test/web/rich_media/aws_signed_url_test.exs @@ -55,7 +55,7 @@ defmodule Pleroma.Web.RichMedia.TTL.AwsSignedUrlTest do Cachex.put(:rich_media_cache, url, metadata) - Pleroma.Web.RichMedia.Parser.set_ttl_based_on_image({:ok, metadata}, url) + Pleroma.Web.RichMedia.Parser.set_ttl_based_on_image(metadata, url) {:ok, cache_ttl} = Cachex.ttl(:rich_media_cache, url) From d48fc90978eee46c8fba00a80082d14fc31a0eec Mon Sep 17 00:00:00 2001 From: rinpatch Date: Wed, 2 Sep 2020 13:44:08 +0300 Subject: [PATCH 3/3] StatusView: Start fetching rich media cards as soon as possible --- CHANGELOG.md | 2 ++ .../web/mastodon_api/views/status_view.ex | 16 ++++++++++++++++ 2 files changed, 18 insertions(+) diff --git a/CHANGELOG.md b/CHANGELOG.md index 4424c060f..54685cfb8 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -10,6 +10,8 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/). ### Fixed - Mastodon API: Search parameter `following` now correctly returns the followings rather than the followers +- Mastodon API: Timelines hanging for (`number of posts with links * rich media timeout`) in the worst case. + Reduced to just rich media timeout. ## [2.1.0] - 2020-08-28 diff --git a/lib/pleroma/web/mastodon_api/views/status_view.ex b/lib/pleroma/web/mastodon_api/views/status_view.ex index 01b8bb6bb..3fe1967be 100644 --- a/lib/pleroma/web/mastodon_api/views/status_view.ex +++ b/lib/pleroma/web/mastodon_api/views/status_view.ex @@ -23,6 +23,17 @@ defmodule Pleroma.Web.MastodonAPI.StatusView do import Pleroma.Web.ActivityPub.Visibility, only: [get_visibility: 1, visible_for_user?: 2] + # This is a naive way to do this, just spawning a process per activity + # to fetch the preview. However it should be fine considering + # pagination is restricted to 40 activities at a time + defp fetch_rich_media_for_activities(activities) do + Enum.each(activities, fn activity -> + spawn(fn -> + Pleroma.Web.RichMedia.Helpers.fetch_data_for_activity(activity) + end) + end) + end + # TODO: Add cached version. defp get_replied_to_activities([]), do: %{} @@ -80,6 +91,11 @@ defmodule Pleroma.Web.MastodonAPI.StatusView do # To do: check AdminAPIControllerTest on the reasons behind nil activities in the list activities = Enum.filter(opts.activities, & &1) + + # Start fetching rich media before doing anything else, so that later calls to get the cards + # only block for timeout in the worst case, as opposed to + # length(activities_with_links) * timeout + fetch_rich_media_for_activities(activities) replied_to_activities = get_replied_to_activities(activities) parent_activities =