Commit 2c55f7d7 authored by rinpatch's avatar rinpatch
Browse files

Remove FedSockets

Current FedSocket implementation has a bunch of problems. It doesn't
have proper error handling (in case of an error the server just doesn't
respond until the connection is closed, while the client doesn't match
any error messages and just assumes there has been an error after 15s)
and the code is full of bad descisions (see: fetch registry which uses
uuids for no reason and waits for a response by recursively querying a
 ets table until the value changes, or double JSON encoding).

Sometime ago I almost completed rewriting fedsockets from scrach to
adress these issues. However, while doing so, I realized that fedsockets
 are just too overkill for what they were trying to accomplish, which is
 reduce the overhead of federation by not signing every message.
This could be done without reimplementing failure states and endpoint
 logic we already have with HTTP by, for example, using TLS cert auth,
or switching to a more performant signature algorithm. I opened
#2262 for further
discussion on alternatives to fedsockets.

From discussions I had with other Pleroma developers it seems like they
 would approve the descision to remove them as well,
therefore I am submitting this patch.
parent 9b9afe6b
Pipeline #32946 passed with stages
in 15 minutes and 13 seconds
......@@ -129,7 +129,6 @@
dispatch: [
{:_,
[
{"/api/fedsocket/v1", Pleroma.Web.FedSockets.IncomingHandler, []},
{"/api/v1/streaming", Pleroma.Web.MastodonAPI.WebsocketHandler, []},
{"/websocket", Phoenix.Endpoint.CowboyWebSocket,
{Phoenix.Transports.WebSocket,
......
......@@ -272,19 +272,6 @@
}
]
},
%{
group: :pleroma,
key: :fed_sockets,
type: :group,
description: "Websocket based federation",
children: [
%{
key: :enabled,
type: :boolean,
description: "Enable FedSockets"
}
]
},
%{
group: :pleroma,
key: Pleroma.Emails.Mailer,
......
......@@ -220,18 +220,6 @@ config :pleroma, :mrf_user_allowlist, %{
* `total_user_limit`: the number of scheduled activities a user is allowed to create in total (Default: `300`)
* `enabled`: whether scheduled activities are sent to the job queue to be executed
## FedSockets
FedSockets is an experimental feature allowing for Pleroma backends to federate using a persistant websocket connection as opposed to making each federation a seperate http connection. This feature is currently off by default. It is configurable throught he following options.
### :fedsockets
* `enabled`: Enables FedSockets for this instance. `false` by default.
* `connection_duration`: Time an idle websocket is kept open.
* `rejection_duration`: Failures to connect via FedSockets will not be retried for this period of time.
* `fed_socket_fetches` and `fed_socket_rejections`: Settings passed to `cachex` for the fetch registry, and rejection stacks. See `Pleroma.Web.FedSockets` for more details.
## Frontends
### :frontend_configurations
This can be used to configure a keyword list that keeps the configuration data for any kind of frontend. By default, settings for `pleroma_fe` and `masto_fe` are configured. You can find the documentation for `pleroma_fe` configuration into [Pleroma-FE configuration and customization for instance administrators](/frontend/CONFIGURATION/#options).
......
......@@ -93,9 +93,4 @@ server {
chunked_transfer_encoding on;
proxy_pass http://phoenix;
}
location /api/fedsocket/v1 {
proxy_request_buffering off;
proxy_pass http://phoenix/api/fedsocket/v1;
}
}
......@@ -207,8 +207,7 @@ defp dont_run_in_test(_) do
name: Pleroma.Web.Streamer.registry(),
keys: :duplicate,
partitions: System.schedulers_online()
]},
Pleroma.Web.FedSockets.Supervisor
]}
]
end
......
......@@ -12,7 +12,6 @@ defmodule Pleroma.Object.Fetcher do
alias Pleroma.Web.ActivityPub.ObjectValidator
alias Pleroma.Web.ActivityPub.Transmogrifier
alias Pleroma.Web.Federator
alias Pleroma.Web.FedSockets
require Logger
require Pleroma.Constants
......@@ -183,16 +182,16 @@ defp maybe_date_fetch(headers, date) do
end
end
def fetch_and_contain_remote_object_from_id(prm, opts \\ [])
def fetch_and_contain_remote_object_from_id(id)
def fetch_and_contain_remote_object_from_id(%{"id" => id}, opts),
do: fetch_and_contain_remote_object_from_id(id, opts)
def fetch_and_contain_remote_object_from_id(%{"id" => id}),
do: fetch_and_contain_remote_object_from_id(id)
def fetch_and_contain_remote_object_from_id(id, opts) when is_binary(id) do
def fetch_and_contain_remote_object_from_id(id) when is_binary(id) do
Logger.debug("Fetching object #{id} via AP")
with {:scheme, true} <- {:scheme, String.starts_with?(id, "http")},
{:ok, body} <- get_object(id, opts),
{:ok, body} <- get_object(id),
{:ok, data} <- safe_json_decode(body),
:ok <- Containment.contain_origin_from_id(id, data) do
{:ok, data}
......@@ -208,22 +207,10 @@ def fetch_and_contain_remote_object_from_id(id, opts) when is_binary(id) do
end
end
def fetch_and_contain_remote_object_from_id(_id, _opts),
def fetch_and_contain_remote_object_from_id(_id),
do: {:error, "id must be a string"}
defp get_object(id, opts) do
with false <- Keyword.get(opts, :force_http, false),
{:ok, fedsocket} <- FedSockets.get_or_create_fed_socket(id) do
Logger.debug("fetching via fedsocket - #{inspect(id)}")
FedSockets.fetch(fedsocket, id)
else
_other ->
Logger.debug("fetching via http - #{inspect(id)}")
get_object_http(id)
end
end
defp get_object_http(id) do
defp get_object(id) do
date = Pleroma.Signature.signed_date()
headers =
......
......@@ -39,7 +39,7 @@ def key_id_to_actor_id(key_id) do
def fetch_public_key(conn) do
with %{"keyId" => kid} <- HTTPSignatures.signature_for_conn(conn),
{:ok, actor_id} <- key_id_to_actor_id(kid),
{:ok, public_key} <- User.get_public_key_for_ap_id(actor_id, force_http: true) do
{:ok, public_key} <- User.get_public_key_for_ap_id(actor_id) do
{:ok, public_key}
else
e ->
......@@ -50,8 +50,8 @@ def fetch_public_key(conn) do
def refetch_public_key(conn) do
with %{"keyId" => kid} <- HTTPSignatures.signature_for_conn(conn),
{:ok, actor_id} <- key_id_to_actor_id(kid),
{:ok, _user} <- ActivityPub.make_user_from_ap_id(actor_id, force_http: true),
{:ok, public_key} <- User.get_public_key_for_ap_id(actor_id, force_http: true) do
{:ok, _user} <- ActivityPub.make_user_from_ap_id(actor_id),
{:ok, public_key} <- User.get_public_key_for_ap_id(actor_id) do
{:ok, public_key}
else
e ->
......
......@@ -1772,12 +1772,12 @@ def html_filter_policy(%User{no_rich_text: true}) do
def html_filter_policy(_), do: Config.get([:markup, :scrub_policy])
def fetch_by_ap_id(ap_id, opts \\ []), do: ActivityPub.make_user_from_ap_id(ap_id, opts)
def fetch_by_ap_id(ap_id), do: ActivityPub.make_user_from_ap_id(ap_id)
def get_or_fetch_by_ap_id(ap_id, opts \\ []) do
def get_or_fetch_by_ap_id(ap_id) do
cached_user = get_cached_by_ap_id(ap_id)
maybe_fetched_user = needs_update?(cached_user) && fetch_by_ap_id(ap_id, opts)
maybe_fetched_user = needs_update?(cached_user) && fetch_by_ap_id(ap_id)
case {cached_user, maybe_fetched_user} do
{_, {:ok, %User{} = user}} ->
......@@ -1850,8 +1850,8 @@ def public_key(%{public_key: public_key_pem}) when is_binary(public_key_pem) do
def public_key(_), do: {:error, "key not found"}
def get_public_key_for_ap_id(ap_id, opts \\ []) do
with {:ok, %User{} = user} <- get_or_fetch_by_ap_id(ap_id, opts),
def get_public_key_for_ap_id(ap_id) do
with {:ok, %User{} = user} <- get_or_fetch_by_ap_id(ap_id),
{:ok, public_key} <- public_key(user) do
{:ok, public_key}
else
......
......@@ -1289,12 +1289,10 @@ defp object_to_user_data(data) do
def fetch_follow_information_for_user(user) do
with {:ok, following_data} <-
Fetcher.fetch_and_contain_remote_object_from_id(user.following_address,
force_http: true
),
Fetcher.fetch_and_contain_remote_object_from_id(user.following_address),
{:ok, hide_follows} <- collection_private(following_data),
{:ok, followers_data} <-
Fetcher.fetch_and_contain_remote_object_from_id(user.follower_address, force_http: true),
Fetcher.fetch_and_contain_remote_object_from_id(user.follower_address),
{:ok, hide_followers} <- collection_private(followers_data) do
{:ok,
%{
......@@ -1368,8 +1366,8 @@ def user_data_from_user_object(data) do
end
end
def fetch_and_prepare_user_from_ap_id(ap_id, opts \\ []) do
with {:ok, data} <- Fetcher.fetch_and_contain_remote_object_from_id(ap_id, opts),
def fetch_and_prepare_user_from_ap_id(ap_id) do
with {:ok, data} <- Fetcher.fetch_and_contain_remote_object_from_id(ap_id),
{:ok, data} <- user_data_from_user_object(data) do
{:ok, maybe_update_follow_information(data)}
else
......@@ -1412,13 +1410,13 @@ def maybe_handle_clashing_nickname(data) do
end
end
def make_user_from_ap_id(ap_id, opts \\ []) do
def make_user_from_ap_id(ap_id) do
user = User.get_cached_by_ap_id(ap_id)
if user && !User.ap_enabled?(user) do
Transmogrifier.upgrade_user_from_ap_id(ap_id)
else
with {:ok, data} <- fetch_and_prepare_user_from_ap_id(ap_id, opts) do
with {:ok, data} <- fetch_and_prepare_user_from_ap_id(ap_id) do
if user do
user
|> User.remote_user_changeset(data)
......
......@@ -13,7 +13,6 @@ defmodule Pleroma.Web.ActivityPub.Publisher do
alias Pleroma.User
alias Pleroma.Web.ActivityPub.Relay
alias Pleroma.Web.ActivityPub.Transmogrifier
alias Pleroma.Web.FedSockets
require Pleroma.Constants
......@@ -50,28 +49,6 @@ def is_representable?(%Activity{} = activity) do
"""
def publish_one(%{inbox: inbox, json: json, actor: %User{} = actor, id: id} = params) do
Logger.debug("Federating #{id} to #{inbox}")
case FedSockets.get_or_create_fed_socket(inbox) do
{:ok, fedsocket} ->
Logger.debug("publishing via fedsockets - #{inspect(inbox)}")
FedSockets.publish(fedsocket, json)
_ ->
Logger.debug("publishing via http - #{inspect(inbox)}")
http_publish(inbox, actor, json, params)
end
end
def publish_one(%{actor_id: actor_id} = params) do
actor = User.get_cached_by_id(actor_id)
params
|> Map.delete(:actor_id)
|> Map.put(:actor, actor)
|> publish_one()
end
defp http_publish(inbox, actor, json, params) do
uri = %{path: path} = URI.parse(inbox)
digest = "SHA-256=" <> (:crypto.hash(:sha256, json) |> Base.encode64())
......@@ -110,6 +87,15 @@ defp http_publish(inbox, actor, json, params) do
end
end
def publish_one(%{actor_id: actor_id} = params) do
actor = User.get_cached_by_id(actor_id)
params
|> Map.delete(:actor_id)
|> Map.put(:actor, actor)
|> publish_one()
end
defp signature_host(%URI{port: port, scheme: scheme, host: host}) do
if port == URI.default_port(scheme) do
host
......
......@@ -1008,7 +1008,7 @@ def perform(:user_upgrade, user) do
def upgrade_user_from_ap_id(ap_id) do
with %User{local: false} = user <- User.get_cached_by_ap_id(ap_id),
{:ok, data} <- ActivityPub.fetch_and_prepare_user_from_ap_id(ap_id, force_http: true),
{:ok, data} <- ActivityPub.fetch_and_prepare_user_from_ap_id(ap_id),
{:ok, user} <- update_user(user, data) do
TransmogrifierWorker.enqueue("user_upgrade", %{"user_id" => user.id})
{:ok, user}
......
# Pleroma: A lightweight social networking server
# Copyright © 2017-2020 Pleroma Authors <https://pleroma.social/>
# SPDX-License-Identifier: AGPL-3.0-only
defmodule Pleroma.Web.FedSockets do
@moduledoc """
This documents the FedSockets framework. A framework for federating
ActivityPub objects between servers via persistant WebSocket connections.
FedSockets allow servers to authenticate on first contact and maintain that
connection, eliminating the need to authenticate every time data needs to be shared.
## Protocol
FedSockets currently support 2 types of data transfer:
* `publish` method which doesn't require a response
* `fetch` method requires a response be sent
### Publish
The publish operation sends a json encoded map of the shape:
%{action: :publish, data: json}
and accepts (but does not require) a reply of form:
%{"action" => "publish_reply"}
The outgoing params represent
* data: ActivityPub object encoded into json
### Fetch
The fetch operation sends a json encoded map of the shape:
%{action: :fetch, data: id, uuid: fetch_uuid}
and requires a reply of form:
%{"action" => "fetch_reply", "uuid" => uuid, "data" => data}
The outgoing params represent
* id: an ActivityPub object URI
* uuid: a unique uuid generated by the sender
The reply params represent
* data: an ActivityPub object encoded into json
* uuid: the uuid sent along with the fetch request
## Examples
Clients of FedSocket transfers shouldn't need to use any of the functions outside of this module.
A typical publish operation can be performed through the following code, and a fetch operation in a similar manner.
case FedSockets.get_or_create_fed_socket(inbox) do
{:ok, fedsocket} ->
FedSockets.publish(fedsocket, json)
_ ->
alternative_publish(inbox, actor, json, params)
end
## Configuration
FedSockets have the following config settings
config :pleroma, :fed_sockets,
enabled: true,
ping_interval: :timer.seconds(15),
connection_duration: :timer.hours(1),
rejection_duration: :timer.hours(1),
fed_socket_fetches: [
default: 12_000,
interval: 3_000,
lazy: false
]
* enabled - turn FedSockets on or off with this flag. Can be toggled at runtime.
* connection_duration - How long a FedSocket can sit idle before it's culled.
* rejection_duration - After failing to make a FedSocket connection a host will be excluded
from further connections for this amount of time
* fed_socket_fetches - Use these parameters to pass options to the Cachex queue backing the FetchRegistry
* fed_socket_rejections - Use these parameters to pass options to the Cachex queue backing the FedRegistry
Cachex options are
* default: the minimum amount of time a fetch can wait before it times out.
* interval: the interval between checks for timed out entries. This plus the default represent the maximum time allowed
* lazy: leave at false for consistant and fast lookups, set to true for stricter timeout enforcement
"""
require Logger
alias Pleroma.Web.FedSockets.FedRegistry
alias Pleroma.Web.FedSockets.FedSocket
alias Pleroma.Web.FedSockets.SocketInfo
@doc """
returns a FedSocket for the given origin. Will reuse an existing one or create a new one.
address is expected to be a fully formed URL such as:
"http://www.example.com" or "http://www.example.com:8080"
It can and usually does include additional path parameters,
but these are ignored as the FedSockets are organized by host and port info alone.
"""
def get_or_create_fed_socket(address) do
with {:cache, {:error, :missing}} <- {:cache, get_fed_socket(address)},
{:connect, {:ok, _pid}} <- {:connect, FedSocket.connect_to_host(address)},
{:cache, {:ok, fed_socket}} <- {:cache, get_fed_socket(address)} do
Logger.debug("fedsocket created for - #{inspect(address)}")
{:ok, fed_socket}
else
{:cache, {:ok, socket}} ->
Logger.debug("fedsocket found in cache - #{inspect(address)}")
{:ok, socket}
{:cache, {:error, :rejected} = e} ->
e
{:connect, {:error, _host}} ->
Logger.debug("set host rejected for - #{inspect(address)}")
FedRegistry.set_host_rejected(address)
{:error, :rejected}
{_, {:error, :disabled}} ->
{:error, :disabled}
{_, {:error, reason}} ->
Logger.warn("get_or_create_fed_socket error - #{inspect(reason)}")
{:error, reason}
end
end
@doc """
returns a FedSocket for the given origin. Will not create a new FedSocket if one does not exist.
address is expected to be a fully formed URL such as:
"http://www.example.com" or "http://www.example.com:8080"
"""
def get_fed_socket(address) do
origin = SocketInfo.origin(address)
with {:config, true} <- {:config, Pleroma.Config.get([:fed_sockets, :enabled], false)},
{:ok, socket} <- FedRegistry.get_fed_socket(origin) do
{:ok, socket}
else
{:config, _} ->
{:error, :disabled}
{:error, :rejected} ->
Logger.debug("FedSocket previously rejected - #{inspect(origin)}")
{:error, :rejected}
{:error, reason} ->
{:error, reason}
end
end
@doc """
Sends the supplied data via the publish protocol.
It will not block waiting for a reply.
Returns :ok but this is not an indication of a successful transfer.
the data is expected to be JSON encoded binary data.
"""
def publish(%SocketInfo{} = fed_socket, json) do
FedSocket.publish(fed_socket, json)
end
@doc """
Sends the supplied data via the fetch protocol.
It will block waiting for a reply or timeout.
Returns {:ok, object} where object is the requested object (or nil)
{:error, :timeout} in the event the message was not responded to
the id is expected to be the URI of an ActivityPub object.
"""
def fetch(%SocketInfo{} = fed_socket, id) do
FedSocket.fetch(fed_socket, id)
end
@doc """
Disconnect all and restart FedSockets.
This is mainly used in development and testing but could be useful in production.
"""
def reset do
FedRegistry
|> Process.whereis()
|> Process.exit(:testing)
end
def uri_for_origin(origin),
do: "ws://#{origin}/api/fedsocket/v1"
end
# Pleroma: A lightweight social networking server
# Copyright © 2017-2020 Pleroma Authors <https://pleroma.social/>
# SPDX-License-Identifier: AGPL-3.0-only
defmodule Pleroma.Web.FedSockets.FedRegistry do
@moduledoc """
The FedRegistry stores the active FedSockets for quick retrieval.
The storage and retrieval portion of the FedRegistry is done in process through
elixir's `Registry` module for speed and its ability to monitor for terminated processes.
Dropped connections will be caught by `Registry` and deleted. Since the next
message will initiate a new connection there is no reason to try and reconnect at that point.
Normally outside modules should have no need to call or use the FedRegistry themselves.
"""
alias Pleroma.Web.FedSockets.FedSocket
alias Pleroma.Web.FedSockets.SocketInfo
require Logger
@default_rejection_duration 15 * 60 * 1000
@rejections :fed_socket_rejections
@doc """
Retrieves a FedSocket from the Registry given it's origin.
The origin is expected to be a string identifying the endpoint "example.com" or "example2.com:8080"
Will return:
* {:ok, fed_socket} for working FedSockets
* {:error, :rejected} for origins that have been tried and refused within the rejection duration interval
* {:error, some_reason} usually :missing for unknown origins
"""
def get_fed_socket(origin) do
case get_registry_data(origin) do
{:error, reason} ->
{:error, reason}
{:ok, %{state: :connected} = socket_info} ->
{:ok, socket_info}
end
end
@doc """
Adds a connected FedSocket to the Registry.
Always returns {:ok, fed_socket}
"""
def add_fed_socket(origin, pid \\ nil) do
origin
|> SocketInfo.build(pid)
|> SocketInfo.connect()
|> add_socket_info
end
defp add_socket_info(%{origin: origin, state: :connected} = socket_info) do
case Registry.register(FedSockets.Registry, origin, socket_info) do
{:ok, _owner} ->
clear_prior_rejection(origin)
Logger.debug("fedsocket added: #{inspect(origin)}")
{:ok, socket_info}
{:error, {:already_registered, _pid}} ->
FedSocket.close(socket_info)
existing_socket_info = Registry.lookup(FedSockets.Registry, origin)
{:ok, existing_socket_info}
_ ->
{:error, :error_adding_socket}
end
end
@doc """
Mark this origin as having rejected a connection attempt.
This will keep it from getting additional connection attempts
for a period of time specified in the config.
Always returns {:ok, new_reg_data}
"""
def set_host_rejected(uri) do
new_reg_data =
uri
|> SocketInfo.origin()
|> get_or_create_registry_data()
|> set_to_rejected()
|> save_registry_data()
{:ok, new_reg_data}
end
@doc """
Retrieves the FedRegistryData from the Registry given it's origin.
The origin is expected to be a string identifying the endpoint "example.com" or "example2.com:8080"
Will return:
* {:ok, fed_registry_data} for known origins
* {:error, :missing} for uniknown origins
* {:error, :cache_error} indicating some low level runtime issues
"""
def get_registry_data(origin) do
case Registry.lookup(FedSockets.Registry, origin) do
[] ->
if is_rejected?(origin) do
Logger.debug("previously rejected fedsocket requested")
{:error, :rejected}
else
{:error, :missing}
end
[{_pid, %{state: :connected} = socket_info}] ->
{:ok, socket_info}
_ ->
{:error, :cache_error}
end
end
@doc """
Retrieves a map of all sockets from the Registry. The keys are the origins and the values are the corresponding SocketInfo
"""
def list_all do
(list_all_connected() ++ list_all_rejected())
|> Enum.into(%{})
end
defp list_all_connected do
FedSockets.Registry
|> Registry.select([{{:"$1", :_, :"$3"}, [], [{{:"$1", :"$3"}}]}])
end
defp list_all_rejected do