summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorlain <lain@soykaf.club>2020-05-20 10:35:36 +0000
committerlain <lain@soykaf.club>2020-05-20 10:35:36 +0000
commit6dd1575c64ccc2ffe82155f70935b2aa58d22684 (patch)
tree35c506f7f402592ab0489996106350b7d529084b
parentc3cfe87c60c1edf046cc4656e7a7704db65a9c85 (diff)
parent376147fb828a75b5000262a376cee173bfc98551 (diff)
Merge branch 'issue/1509' into 'develop'
[#1509] purge media from cache after delete See merge request pleroma/pleroma!2539
-rw-r--r--config/config.exs4
-rw-r--r--docs/configuration/cheatsheet.md34
-rwxr-xr-xinstallation/nginx-cache-purge.sh.example40
-rw-r--r--lib/pleroma/object.ex34
-rw-r--r--lib/pleroma/web/media_proxy/invalidation.ex26
-rw-r--r--lib/pleroma/web/media_proxy/invalidations/http.ex40
-rw-r--r--lib/pleroma/web/media_proxy/invalidations/script.ex41
-rw-r--r--lib/pleroma/workers/attachments_cleanup_worker.ex49
-rw-r--r--test/web/media_proxy/invalidations/http_test.exs35
-rw-r--r--test/web/media_proxy/invalidations/script_test.exs20
10 files changed, 293 insertions, 30 deletions
diff --git a/config/config.exs b/config/config.exs
index c51884f3a..8833b3e27 100644
--- a/config/config.exs
+++ b/config/config.exs
@@ -376,6 +376,10 @@ config :pleroma, :rich_media,
config :pleroma, :media_proxy,
enabled: false,
+ invalidation: [
+ enabled: false,
+ provider: Pleroma.Web.MediaProxy.Invalidation.Script
+ ],
proxy_opts: [
redirect_on_failure: false,
max_body_length: 25 * 1_048_576,
diff --git a/docs/configuration/cheatsheet.md b/docs/configuration/cheatsheet.md
index ce3bf3af7..9af8ee95a 100644
--- a/docs/configuration/cheatsheet.md
+++ b/docs/configuration/cheatsheet.md
@@ -249,6 +249,40 @@ This section describe PWA manifest instance-specific values. Currently this opti
* `base_url`: The base URL to access a user-uploaded file. Useful when you want to proxy the media files via another host/CDN fronts.
* `proxy_opts`: All options defined in `Pleroma.ReverseProxy` documentation, defaults to `[max_body_length: (25*1_048_576)]`.
* `whitelist`: List of domains to bypass the mediaproxy
+* `invalidation`: options for remove media from cache after delete object:
+ * `enabled`: Enables purge cache
+ * `provider`: Which one of the [purge cache strategy](#purge-cache-strategy) to use.
+
+### Purge cache strategy
+
+#### Pleroma.Web.MediaProxy.Invalidation.Script
+
+This strategy allow perform external bash script to purge cache.
+Urls of attachments pass to script as arguments.
+
+* `script_path`: path to external script.
+
+Example:
+```elixir
+config :pleroma, Pleroma.Web.MediaProxy.Invalidation.Script,
+ script_path: "./installation/nginx-cache-purge.example"
+```
+
+#### Pleroma.Web.MediaProxy.Invalidation.Http
+
+This strategy allow perform custom http request to purge cache.
+
+* `method`: http method. default is `purge`
+* `headers`: http headers. default is empty
+* `options`: request options. default is empty
+
+Example:
+```elixir
+config :pleroma, Pleroma.Web.MediaProxy.Invalidation.Http,
+ method: :purge,
+ headers: [],
+ options: []
+```
## Link previews
diff --git a/installation/nginx-cache-purge.sh.example b/installation/nginx-cache-purge.sh.example
new file mode 100755
index 000000000..b2915321c
--- /dev/null
+++ b/installation/nginx-cache-purge.sh.example
@@ -0,0 +1,40 @@
+#!/bin/sh
+
+# A simple shell script to delete a media from the Nginx cache.
+
+SCRIPTNAME=${0##*/}
+
+# NGINX cache directory
+CACHE_DIRECTORY="/tmp/pleroma-media-cache"
+
+## Return the files where the items are cached.
+## $1 - the filename, can be a pattern .
+## $2 - the cache directory.
+## $3 - (optional) the number of parallel processes to run for grep.
+get_cache_files() {
+ local max_parallel=${3-16}
+ find $2 -maxdepth 2 -type d | xargs -P $max_parallel -n 1 grep -E Rl "^KEY:.*$1" | sort -u
+}
+
+## Removes an item from the given cache zone.
+## $1 - the filename, can be a pattern .
+## $2 - the cache directory.
+purge_item() {
+ for f in $(get_cache_files $1 $2); do
+ echo "found file: $f"
+ [ -f $f ] || continue
+ echo "Deleting $f from $2."
+ rm $f
+ done
+} # purge_item
+
+purge() {
+ for url in "$@"
+ do
+ echo "$SCRIPTNAME delete \`$url\` from cache ($CACHE_DIRECTORY)"
+ purge_item $url $CACHE_DIRECTORY
+ done
+
+}
+
+purge $1
diff --git a/lib/pleroma/object.ex b/lib/pleroma/object.ex
index ab16bf2db..546c4ea01 100644
--- a/lib/pleroma/object.ex
+++ b/lib/pleroma/object.ex
@@ -9,11 +9,13 @@ defmodule Pleroma.Object do
import Ecto.Changeset
alias Pleroma.Activity
+ alias Pleroma.Config
alias Pleroma.Object
alias Pleroma.Object.Fetcher
alias Pleroma.ObjectTombstone
alias Pleroma.Repo
alias Pleroma.User
+ alias Pleroma.Workers.AttachmentsCleanupWorker
require Logger
@@ -188,27 +190,37 @@ defmodule Pleroma.Object do
def delete(%Object{data: %{"id" => id}} = object) do
with {:ok, _obj} = swap_object_with_tombstone(object),
deleted_activity = Activity.delete_all_by_object_ap_id(id),
- {:ok, true} <- Cachex.del(:object_cache, "object:#{id}"),
- {:ok, _} <- Cachex.del(:web_resp_cache, URI.parse(id).path) do
- with true <- Pleroma.Config.get([:instance, :cleanup_attachments]) do
- {:ok, _} =
- Pleroma.Workers.AttachmentsCleanupWorker.enqueue("cleanup_attachments", %{
- "object" => object
- })
- end
+ {:ok, _} <- invalid_object_cache(object) do
+ cleanup_attachments(
+ Config.get([:instance, :cleanup_attachments]),
+ %{"object" => object}
+ )
{:ok, object, deleted_activity}
end
end
- def prune(%Object{data: %{"id" => id}} = object) do
+ @spec cleanup_attachments(boolean(), %{required(:object) => map()}) ::
+ {:ok, Oban.Job.t() | nil}
+ def cleanup_attachments(true, %{"object" => _} = params) do
+ AttachmentsCleanupWorker.enqueue("cleanup_attachments", params)
+ end
+
+ def cleanup_attachments(_, _), do: {:ok, nil}
+
+ def prune(%Object{data: %{"id" => _id}} = object) do
with {:ok, object} <- Repo.delete(object),
- {:ok, true} <- Cachex.del(:object_cache, "object:#{id}"),
- {:ok, _} <- Cachex.del(:web_resp_cache, URI.parse(id).path) do
+ {:ok, _} <- invalid_object_cache(object) do
{:ok, object}
end
end
+ def invalid_object_cache(%Object{data: %{"id" => id}}) do
+ with {:ok, true} <- Cachex.del(:object_cache, "object:#{id}") do
+ Cachex.del(:web_resp_cache, URI.parse(id).path)
+ end
+ end
+
def set_cache(%Object{data: %{"id" => ap_id}} = object) do
Cachex.put(:object_cache, "object:#{ap_id}", object)
{:ok, object}
diff --git a/lib/pleroma/web/media_proxy/invalidation.ex b/lib/pleroma/web/media_proxy/invalidation.ex
new file mode 100644
index 000000000..c037ff13e
--- /dev/null
+++ b/lib/pleroma/web/media_proxy/invalidation.ex
@@ -0,0 +1,26 @@
+# Pleroma: A lightweight social networking server
+# Copyright © 2017-2020 Pleroma Authors <https://pleroma.social/>
+# SPDX-License-Identifier: AGPL-3.0-only
+
+defmodule Pleroma.Web.MediaProxy.Invalidation do
+ @moduledoc false
+
+ @callback purge(list(String.t()), map()) :: {:ok, String.t()} | {:error, String.t()}
+
+ alias Pleroma.Config
+
+ @spec purge(list(String.t())) :: {:ok, String.t()} | {:error, String.t()}
+ def purge(urls) do
+ [:media_proxy, :invalidation, :enabled]
+ |> Config.get()
+ |> do_purge(urls)
+ end
+
+ defp do_purge(true, urls) do
+ provider = Config.get([:media_proxy, :invalidation, :provider])
+ options = Config.get(provider)
+ provider.purge(urls, options)
+ end
+
+ defp do_purge(_, _), do: :ok
+end
diff --git a/lib/pleroma/web/media_proxy/invalidations/http.ex b/lib/pleroma/web/media_proxy/invalidations/http.ex
new file mode 100644
index 000000000..07248df6e
--- /dev/null
+++ b/lib/pleroma/web/media_proxy/invalidations/http.ex
@@ -0,0 +1,40 @@
+# Pleroma: A lightweight social networking server
+# Copyright © 2017-2020 Pleroma Authors <https://pleroma.social/>
+# SPDX-License-Identifier: AGPL-3.0-only
+
+defmodule Pleroma.Web.MediaProxy.Invalidation.Http do
+ @moduledoc false
+ @behaviour Pleroma.Web.MediaProxy.Invalidation
+
+ require Logger
+
+ @impl Pleroma.Web.MediaProxy.Invalidation
+ def purge(urls, opts) do
+ method = Map.get(opts, :method, :purge)
+ headers = Map.get(opts, :headers, [])
+ options = Map.get(opts, :options, [])
+
+ Logger.debug("Running cache purge: #{inspect(urls)}")
+
+ Enum.each(urls, fn url ->
+ with {:error, error} <- do_purge(method, url, headers, options) do
+ Logger.error("Error while cache purge: url - #{url}, error: #{inspect(error)}")
+ end
+ end)
+
+ {:ok, "success"}
+ end
+
+ defp do_purge(method, url, headers, options) do
+ case Pleroma.HTTP.request(method, url, "", headers, options) do
+ {:ok, %{status: status} = env} when 400 <= status and status < 500 ->
+ {:error, env}
+
+ {:error, error} = error ->
+ error
+
+ _ ->
+ {:ok, "success"}
+ end
+ end
+end
diff --git a/lib/pleroma/web/media_proxy/invalidations/script.ex b/lib/pleroma/web/media_proxy/invalidations/script.ex
new file mode 100644
index 000000000..6be782132
--- /dev/null
+++ b/lib/pleroma/web/media_proxy/invalidations/script.ex
@@ -0,0 +1,41 @@
+# Pleroma: A lightweight social networking server
+# Copyright © 2017-2020 Pleroma Authors <https://pleroma.social/>
+# SPDX-License-Identifier: AGPL-3.0-only
+
+defmodule Pleroma.Web.MediaProxy.Invalidation.Script do
+ @moduledoc false
+
+ @behaviour Pleroma.Web.MediaProxy.Invalidation
+
+ require Logger
+
+ @impl Pleroma.Web.MediaProxy.Invalidation
+ def purge(urls, %{script_path: script_path} = _options) do
+ args =
+ urls
+ |> List.wrap()
+ |> Enum.uniq()
+ |> Enum.join(" ")
+
+ path = Path.expand(script_path)
+
+ Logger.debug("Running cache purge: #{inspect(urls)}, #{path}")
+
+ case do_purge(path, [args]) do
+ {result, exit_status} when exit_status > 0 ->
+ Logger.error("Error while cache purge: #{inspect(result)}")
+ {:error, inspect(result)}
+
+ _ ->
+ {:ok, "success"}
+ end
+ end
+
+ def purge(_, _), do: {:error, "not found script path"}
+
+ defp do_purge(path, args) do
+ System.cmd(path, args)
+ rescue
+ error -> {inspect(error), 1}
+ end
+end
diff --git a/lib/pleroma/workers/attachments_cleanup_worker.ex b/lib/pleroma/workers/attachments_cleanup_worker.ex
index 3c5820a86..49352db2a 100644
--- a/lib/pleroma/workers/attachments_cleanup_worker.ex
+++ b/lib/pleroma/workers/attachments_cleanup_worker.ex
@@ -27,8 +27,20 @@ defmodule Pleroma.Workers.AttachmentsCleanupWorker do
uploader = Pleroma.Config.get([Pleroma.Upload, :uploader])
+ prefix =
+ case Pleroma.Config.get([Pleroma.Upload, :base_url]) do
+ nil -> "media"
+ _ -> ""
+ end
+
+ base_url =
+ String.trim_trailing(
+ Pleroma.Config.get([Pleroma.Upload, :base_url], Pleroma.Web.base_url()),
+ "/"
+ )
+
# find all objects for copies of the attachments, name and actor doesn't matter here
- delete_ids =
+ object_ids_and_hrefs =
from(o in Object,
where:
fragment(
@@ -67,29 +79,28 @@ defmodule Pleroma.Workers.AttachmentsCleanupWorker do
|> Enum.map(fn {href, %{id: id, count: count}} ->
# only delete files that have single instance
with 1 <- count do
- prefix =
- case Pleroma.Config.get([Pleroma.Upload, :base_url]) do
- nil -> "media"
- _ -> ""
- end
-
- base_url =
- String.trim_trailing(
- Pleroma.Config.get([Pleroma.Upload, :base_url], Pleroma.Web.base_url()),
- "/"
- )
-
- file_path = String.trim_leading(href, "#{base_url}/#{prefix}")
+ href
+ |> String.trim_leading("#{base_url}/#{prefix}")
+ |> uploader.delete_file()
- uploader.delete_file(file_path)
+ {id, href}
+ else
+ _ -> {id, nil}
end
-
- id
end)
- from(o in Object, where: o.id in ^delete_ids)
+ object_ids = Enum.map(object_ids_and_hrefs, fn {id, _} -> id end)
+
+ from(o in Object, where: o.id in ^object_ids)
|> Repo.delete_all()
+
+ object_ids_and_hrefs
+ |> Enum.filter(fn {_, href} -> not is_nil(href) end)
+ |> Enum.map(&elem(&1, 1))
+ |> Pleroma.Web.MediaProxy.Invalidation.purge()
+
+ {:ok, :success}
end
- def perform(%{"op" => "cleanup_attachments", "object" => _object}, _job), do: :ok
+ def perform(%{"op" => "cleanup_attachments", "object" => _object}, _job), do: {:ok, :skip}
end
diff --git a/test/web/media_proxy/invalidations/http_test.exs b/test/web/media_proxy/invalidations/http_test.exs
new file mode 100644
index 000000000..8a3b4141c
--- /dev/null
+++ b/test/web/media_proxy/invalidations/http_test.exs
@@ -0,0 +1,35 @@
+defmodule Pleroma.Web.MediaProxy.Invalidation.HttpTest do
+ use ExUnit.Case
+ alias Pleroma.Web.MediaProxy.Invalidation
+
+ import ExUnit.CaptureLog
+ import Tesla.Mock
+
+ test "logs hasn't error message when request is valid" do
+ mock(fn
+ %{method: :purge, url: "http://example.com/media/example.jpg"} ->
+ %Tesla.Env{status: 200}
+ end)
+
+ refute capture_log(fn ->
+ assert Invalidation.Http.purge(
+ ["http://example.com/media/example.jpg"],
+ %{}
+ ) == {:ok, "success"}
+ end) =~ "Error while cache purge"
+ end
+
+ test "it write error message in logs when request invalid" do
+ mock(fn
+ %{method: :purge, url: "http://example.com/media/example1.jpg"} ->
+ %Tesla.Env{status: 404}
+ end)
+
+ assert capture_log(fn ->
+ assert Invalidation.Http.purge(
+ ["http://example.com/media/example1.jpg"],
+ %{}
+ ) == {:ok, "success"}
+ end) =~ "Error while cache purge: url - http://example.com/media/example1.jpg"
+ end
+end
diff --git a/test/web/media_proxy/invalidations/script_test.exs b/test/web/media_proxy/invalidations/script_test.exs
new file mode 100644
index 000000000..1358963ab
--- /dev/null
+++ b/test/web/media_proxy/invalidations/script_test.exs
@@ -0,0 +1,20 @@
+defmodule Pleroma.Web.MediaProxy.Invalidation.ScriptTest do
+ use ExUnit.Case
+ alias Pleroma.Web.MediaProxy.Invalidation
+
+ import ExUnit.CaptureLog
+
+ test "it logger error when script not found" do
+ assert capture_log(fn ->
+ assert Invalidation.Script.purge(
+ ["http://example.com/media/example.jpg"],
+ %{script_path: "./example"}
+ ) == {:error, "\"%ErlangError{original: :enoent}\""}
+ end) =~ "Error while cache purge: \"%ErlangError{original: :enoent}\""
+
+ assert Invalidation.Script.purge(
+ ["http://example.com/media/example.jpg"],
+ %{}
+ ) == {:error, "not found script path"}
+ end
+end