Merge branch 'main' into migration-runner

Cargo.lock

@@ -2737,7 +2737,7 @@ checksum = "830d08ce1d1d941e6b30645f1a0eb5643013d835ce3779a5fc208261dbe10f55"
 
 [[package]]
 name = "lemmy_api"
-version = "0.19.4-rc.2"
+version = "0.19.4-rc.3"
 dependencies = [
  "activitypub_federation",
  "actix-web",
@@ -2766,7 +2766,7 @@ dependencies = [
 
 [[package]]
 name = "lemmy_api_common"
-version = "0.19.4-rc.2"
+version = "0.19.4-rc.3"
 dependencies = [
  "activitypub_federation",
  "actix-web",
@@ -2804,7 +2804,7 @@ dependencies = [
 
 [[package]]
 name = "lemmy_api_crud"
-version = "0.19.4-rc.2"
+version = "0.19.4-rc.3"
 dependencies = [
  "accept-language",
  "activitypub_federation",
@@ -2827,7 +2827,7 @@ dependencies = [
 
 [[package]]
 name = "lemmy_apub"
-version = "0.19.4-rc.2"
+version = "0.19.4-rc.3"
 dependencies = [
  "activitypub_federation",
  "actix-web",
@@ -2865,7 +2865,7 @@ dependencies = [
 
 [[package]]
 name = "lemmy_db_perf"
-version = "0.19.4-rc.2"
+version = "0.19.4-rc.3"
 dependencies = [
  "anyhow",
  "clap",
@@ -2880,7 +2880,7 @@ dependencies = [
 
 [[package]]
 name = "lemmy_db_schema"
-version = "0.19.4-rc.2"
+version = "0.19.4-rc.3"
 dependencies = [
  "activitypub_federation",
  "anyhow",
@@ -2921,7 +2921,7 @@ dependencies = [
 
 [[package]]
 name = "lemmy_db_views"
-version = "0.19.4-rc.2"
+version = "0.19.4-rc.3"
 dependencies = [
  "actix-web",
  "chrono",
@@ -2943,7 +2943,7 @@ dependencies = [
 
 [[package]]
 name = "lemmy_db_views_actor"
-version = "0.19.4-rc.2"
+version = "0.19.4-rc.3"
 dependencies = [
  "chrono",
  "diesel",
@@ -2963,7 +2963,7 @@ dependencies = [
 
 [[package]]
 name = "lemmy_db_views_moderator"
-version = "0.19.4-rc.2"
+version = "0.19.4-rc.3"
 dependencies = [
  "diesel",
  "diesel-async",
@@ -2975,7 +2975,7 @@ dependencies = [
 
 [[package]]
 name = "lemmy_federate"
-version = "0.19.4-rc.2"
+version = "0.19.4-rc.3"
 dependencies = [
  "activitypub_federation",
  "anyhow",
@@ -2987,10 +2987,12 @@ dependencies = [
  "lemmy_apub",
  "lemmy_db_schema",
  "lemmy_db_views_actor",
+ "lemmy_utils",
  "moka",
  "once_cell",
  "reqwest 0.11.27",
  "serde_json",
+ "serial_test",
  "tokio",
  "tokio-util",
  "tracing",
@@ -2998,7 +3000,7 @@ dependencies = [
 
 [[package]]
 name = "lemmy_routes"
-version = "0.19.4-rc.2"
+version = "0.19.4-rc.3"
 dependencies = [
  "activitypub_federation",
  "actix-web",
@@ -3023,7 +3025,7 @@ dependencies = [
 
 [[package]]
 name = "lemmy_server"
-version = "0.19.4-rc.2"
+version = "0.19.4-rc.3"
 dependencies = [
  "activitypub_federation",
  "actix-cors",
@@ -3066,7 +3068,7 @@ dependencies = [
 
 [[package]]
 name = "lemmy_utils"
-version = "0.19.4-rc.2"
+version = "0.19.4-rc.3"
 dependencies = [
  "actix-web",
  "anyhow",

Cargo.toml

@@ -1,5 +1,5 @@
 [workspace.package]
-version = "0.19.4-rc.2"
+version = "0.19.4-rc.3"
 edition = "2021"
 description = "A link aggregator for the fediverse"
 license = "AGPL-3.0"
@@ -88,17 +88,17 @@ unused_self = "deny"
 unwrap_used = "deny"
 
 [workspace.dependencies]
-lemmy_api = { version = "=0.19.4-rc.2", path = "./crates/api" }
+lemmy_api = { version = "=0.19.4-rc.3", path = "./crates/api" }
-lemmy_api_crud = { version = "=0.19.4-rc.2", path = "./crates/api_crud" }
+lemmy_api_crud = { version = "=0.19.4-rc.3", path = "./crates/api_crud" }
-lemmy_apub = { version = "=0.19.4-rc.2", path = "./crates/apub" }
+lemmy_apub = { version = "=0.19.4-rc.3", path = "./crates/apub" }
-lemmy_utils = { version = "=0.19.4-rc.2", path = "./crates/utils", default-features = false }
+lemmy_utils = { version = "=0.19.4-rc.3", path = "./crates/utils", default-features = false }
-lemmy_db_schema = { version = "=0.19.4-rc.2", path = "./crates/db_schema" }
+lemmy_db_schema = { version = "=0.19.4-rc.3", path = "./crates/db_schema" }
-lemmy_api_common = { version = "=0.19.4-rc.2", path = "./crates/api_common" }
+lemmy_api_common = { version = "=0.19.4-rc.3", path = "./crates/api_common" }
-lemmy_routes = { version = "=0.19.4-rc.2", path = "./crates/routes" }
+lemmy_routes = { version = "=0.19.4-rc.3", path = "./crates/routes" }
-lemmy_db_views = { version = "=0.19.4-rc.2", path = "./crates/db_views" }
+lemmy_db_views = { version = "=0.19.4-rc.3", path = "./crates/db_views" }
-lemmy_db_views_actor = { version = "=0.19.4-rc.2", path = "./crates/db_views_actor" }
+lemmy_db_views_actor = { version = "=0.19.4-rc.3", path = "./crates/db_views_actor" }
-lemmy_db_views_moderator = { version = "=0.19.4-rc.2", path = "./crates/db_views_moderator" }
+lemmy_db_views_moderator = { version = "=0.19.4-rc.3", path = "./crates/db_views_moderator" }
-lemmy_federate = { version = "=0.19.4-rc.2", path = "./crates/federate" }
+lemmy_federate = { version = "=0.19.4-rc.3", path = "./crates/federate" }
 activitypub_federation = { version = "0.5.6", default-features = false, features = [
   "actix-web",
 ] }

crates/api/src/post/like.rs

@@ -68,7 +68,6 @@ pub async fn like_post(
       .with_lemmy_type(LemmyErrorType::CouldntLikePost)?;
   }
 
-  // Mark the post as read
   mark_post_as_read(person_id, post_id, &mut context.pool()).await?;
 
   let community = Community::read(&mut context.pool(), post.community_id)

crates/api/src/post/save.rs

@@ -38,7 +38,6 @@ pub async fn save_post(
     .await?
     .ok_or(LemmyErrorType::CouldntFindPost)?;
 
-  // Mark the post as read
   mark_post_as_read(person_id, post_id, &mut context.pool()).await?;
 
   Ok(Json(PostResponse { post_view }))

crates/api_common/src/utils.rs

@@ -6,6 +6,7 @@ use crate::{
 use chrono::{DateTime, Days, Local, TimeZone, Utc};
 use enum_map::{enum_map, EnumMap};
 use lemmy_db_schema::{
+  aggregates::structs::{PersonPostAggregates, PersonPostAggregatesForm},
   newtypes::{CommunityId, DbUrl, InstanceId, PersonId, PostId},
   source::{
     comment::{Comment, CommentUpdateForm},
@@ -139,13 +140,7 @@ pub fn is_top_mod(
   }
 }
 
-#[tracing::instrument(skip_all)]
+/// Marks a post as read for a given person.
-pub async fn get_post(post_id: PostId, pool: &mut DbPool<'_>) -> LemmyResult<Post> {
-  Post::read(pool, post_id)
-    .await?
-    .ok_or(LemmyErrorType::CouldntFindPost.into())
-}
-
 #[tracing::instrument(skip_all)]
 pub async fn mark_post_as_read(
   person_id: PersonId,
@@ -158,6 +153,28 @@ pub async fn mark_post_as_read(
   Ok(())
 }
 
+/// Updates the read comment count for a post. Usually done when reading or creating a new comment.
+#[tracing::instrument(skip_all)]
+pub async fn update_read_comments(
+  person_id: PersonId,
+  post_id: PostId,
+  read_comments: i64,
+  pool: &mut DbPool<'_>,
+) -> LemmyResult<()> {
+  let person_post_agg_form = PersonPostAggregatesForm {
+    person_id,
+    post_id,
+    read_comments,
+    ..PersonPostAggregatesForm::default()
+  };
+
+  PersonPostAggregates::upsert(pool, &person_post_agg_form)
+    .await
+    .with_lemmy_type(LemmyErrorType::CouldntFindPost)?;
+
+  Ok(())
+}
+
 pub fn check_user_valid(person: &Person) -> LemmyResult<()> {
   // Check for a site ban
   if person.banned {

crates/api_crud/src/comment/create.rs

@@ -9,11 +9,11 @@ use lemmy_api_common::{
     check_community_user_action,
     check_post_deleted_or_removed,
     generate_local_apub_endpoint,
-    get_post,
     get_url_blocklist,
     is_mod_or_admin,
     local_site_to_slur_regex,
     process_markdown,
+    update_read_comments,
     EndpointType,
   },
 };
@@ -28,7 +28,7 @@ use lemmy_db_schema::{
   },
   traits::{Crud, Likeable},
 };
-use lemmy_db_views::structs::LocalUserView;
+use lemmy_db_views::structs::{LocalUserView, PostView};
 use lemmy_utils::{
   error::{LemmyErrorExt, LemmyErrorType, LemmyResult},
   utils::{mention::scrape_text_for_mentions, validation::is_valid_body_field},
@@ -51,8 +51,19 @@ pub async fn create_comment(
 
   // Check for a community ban
   let post_id = data.post_id;
-  let post = get_post(post_id, &mut context.pool()).await?;
+
-  let community_id = post.community_id;
+  // Read the full post view in order to get the comments count.
+  let post_view = PostView::read(
+    &mut context.pool(),
+    post_id,
+    Some(local_user_view.person.id),
+    true,
+  )
+  .await?
+  .ok_or(LemmyErrorType::CouldntFindPost)?;
+
+  let post = post_view.post;
+  let community_id = post_view.community.id;
 
   check_community_user_action(&local_user_view.person, community_id, &mut context.pool()).await?;
   check_post_deleted_or_removed(&post)?;
@@ -164,6 +175,15 @@ pub async fn create_comment(
   )
   .await?;
 
+  // Update the read comments, so your own new comment doesn't appear as a +1 unread
+  update_read_comments(
+    local_user_view.person.id,
+    post_id,
+    post_view.counts.comments + 1,
+    &mut context.pool(),
+  )
+  .await?;
+
   // If we're responding to a comment where we're the recipient,
   // (ie we're the grandparent, or the recipient of the parent comment_reply),
   // then mark the parent as read.

crates/api_crud/src/post/create.rs

@@ -176,7 +176,6 @@ pub async fn create_post(
     .await
     .with_lemmy_type(LemmyErrorType::CouldntLikePost)?;
 
-  // Mark the post as read
   mark_post_as_read(person_id, post_id, &mut context.pool()).await?;
 
   if let Some(url) = updated_post.url.clone() {

crates/api_crud/src/post/read.rs

@@ -2,10 +2,9 @@ use actix_web::web::{Data, Json, Query};
 use lemmy_api_common::{
   context::LemmyContext,
   post::{GetPost, GetPostResponse},
-  utils::{check_private_instance, is_mod_or_admin_opt, mark_post_as_read},
+  utils::{check_private_instance, is_mod_or_admin_opt, mark_post_as_read, update_read_comments},
 };
 use lemmy_db_schema::{
-  aggregates::structs::{PersonPostAggregates, PersonPostAggregatesForm},
   source::{comment::Comment, post::Post},
   traits::Crud,
 };
@@ -14,7 +13,7 @@ use lemmy_db_views::{
   structs::{LocalUserView, PostView, SiteView},
 };
 use lemmy_db_views_actor::structs::{CommunityModeratorView, CommunityView};
-use lemmy_utils::error::{LemmyErrorExt, LemmyErrorType, LemmyResult};
+use lemmy_utils::error::{LemmyErrorType, LemmyResult};
 
 #[tracing::instrument(skip(context))]
 pub async fn get_post(
@@ -60,10 +59,17 @@ pub async fn get_post(
     .await?
     .ok_or(LemmyErrorType::CouldntFindPost)?;
 
-  // Mark the post as read
   let post_id = post_view.post.id;
   if let Some(person_id) = person_id {
     mark_post_as_read(person_id, post_id, &mut context.pool()).await?;
+
+    update_read_comments(
+      person_id,
+      post_id,
+      post_view.counts.comments,
+      &mut context.pool(),
+    )
+    .await?;
   }
 
   // Necessary for the sidebar subscribed
@@ -76,21 +82,6 @@ pub async fn get_post(
   .await?
   .ok_or(LemmyErrorType::CouldntFindCommunity)?;
 
-  // Insert into PersonPostAggregates
-  // to update the read_comments count
-  if let Some(person_id) = person_id {
-    let read_comments = post_view.counts.comments;
-    let person_post_agg_form = PersonPostAggregatesForm {
-      person_id,
-      post_id,
-      read_comments,
-      ..PersonPostAggregatesForm::default()
-    };
-    PersonPostAggregates::upsert(&mut context.pool(), &person_post_agg_form)
-      .await
-      .with_lemmy_type(LemmyErrorType::CouldntFindPost)?;
-  }
-
   let moderators = CommunityModeratorView::for_community(&mut context.pool(), community_id).await?;
 
   // Fetch the cross_posts

crates/db_schema/src/impls/instance.rs

@@ -94,11 +94,15 @@ impl Instance {
       .await
   }
 
-  #[cfg(test)]
+  /// Only for use in tests
   pub async fn delete_all(pool: &mut DbPool<'_>) -> Result<usize, Error> {
     let conn = &mut get_conn(pool).await?;
+    diesel::delete(federation_queue_state::table)
+      .execute(conn)
+      .await?;
     diesel::delete(instance::table).execute(conn).await
   }
+
   pub async fn allowlist(pool: &mut DbPool<'_>) -> Result<Vec<Self>, Error> {
     let conn = &mut get_conn(pool).await?;
     instance::table

crates/db_schema/src/newtypes.rs

@@ -127,11 +127,13 @@ pub struct LanguageId(pub i32);
 /// The comment reply id.
 pub struct CommentReplyId(i32);
 
-#[derive(Debug, Copy, Clone, Hash, Eq, PartialEq, Serialize, Deserialize, Default)]
+#[derive(
+  Debug, Copy, Clone, Hash, Eq, PartialEq, Serialize, Deserialize, Default, Ord, PartialOrd,
+)]
 #[cfg_attr(feature = "full", derive(DieselNewType, TS))]
 #[cfg_attr(feature = "full", ts(export))]
 /// The instance id.
-pub struct InstanceId(i32);
+pub struct InstanceId(pub i32);
 
 #[derive(
   Debug, Copy, Clone, Hash, Eq, PartialEq, Serialize, Deserialize, Default, PartialOrd, Ord,

crates/federate/Cargo.toml

@@ -19,6 +19,7 @@ lemmy_api_common.workspace = true
 lemmy_apub.workspace = true
 lemmy_db_schema = { workspace = true, features = ["full"] }
 lemmy_db_views_actor.workspace = true
+lemmy_utils.workspace = true
 
 activitypub_federation.workspace = true
 anyhow.workspace = true
@@ -33,3 +34,6 @@ tokio = { workspace = true, features = ["full"] }
 tracing.workspace = true
 moka.workspace = true
 tokio-util = "0.7.11"
+
+[dev-dependencies]
+serial_test = { workspace = true }

crates/federate/src/lib.rs

@@ -1,20 +1,22 @@
 use crate::{util::CancellableTask, worker::InstanceWorker};
 use activitypub_federation::config::FederationConfig;
-use chrono::{Local, Timelike};
+use lemmy_api_common::context::LemmyContext;
-use lemmy_api_common::{context::LemmyContext, federate_retry_sleep_duration};
 use lemmy_db_schema::{
   newtypes::InstanceId,
   source::{federation_queue_state::FederationQueueState, instance::Instance},
-  utils::{ActualDbPool, DbPool},
 };
+use lemmy_utils::error::LemmyResult;
+use stats::receive_print_stats;
 use std::{collections::HashMap, time::Duration};
 use tokio::{
-  sync::mpsc::{unbounded_channel, UnboundedReceiver},
+  sync::mpsc::{unbounded_channel, UnboundedSender},
+  task::JoinHandle,
   time::sleep,
 };
 use tokio_util::sync::CancellationToken;
 use tracing::info;
 
+mod stats;
 mod util;
 mod worker;
 
@@ -32,34 +34,60 @@ pub struct Opts {
   pub process_index: i32,
 }
 
-async fn start_stop_federation_workers(
+pub struct SendManager {
   opts: Opts,
-  pool: ActualDbPool,
+  workers: HashMap<InstanceId, CancellableTask>,
-  federation_config: FederationConfig<LemmyContext>,
+  context: FederationConfig<LemmyContext>,
-  cancel: CancellationToken,
+  stats_sender: UnboundedSender<(InstanceId, FederationQueueState)>,
-) -> anyhow::Result<()> {
+  exit_print: JoinHandle<()>,
-  let mut workers = HashMap::<InstanceId, CancellableTask>::new();
+}
+
+impl SendManager {
+  pub fn new(opts: Opts, context: FederationConfig<LemmyContext>) -> Self {
+    assert!(opts.process_count > 0);
+    assert!(opts.process_index > 0);
+    assert!(opts.process_index <= opts.process_count);
 
     let (stats_sender, stats_receiver) = unbounded_channel();
-  let exit_print = tokio::spawn(receive_print_stats(pool.clone(), stats_receiver));
+    Self {
-  let pool2 = &mut DbPool::Pool(&pool);
+      opts,
-  let process_index = opts.process_index - 1;
+      workers: HashMap::new(),
-  let local_domain = federation_config.settings().get_hostname_without_port()?;
+      stats_sender,
+      exit_print: tokio::spawn(receive_print_stats(
+        context.inner_pool().clone(),
+        stats_receiver,
+      )),
+      context,
+    }
+  }
+
+  pub fn run(mut self) -> CancellableTask {
+    CancellableTask::spawn(WORKER_EXIT_TIMEOUT, move |cancel| async move {
+      self.do_loop(cancel).await?;
+      self.cancel().await?;
+      Ok(())
+    })
+  }
+
+  async fn do_loop(&mut self, cancel: CancellationToken) -> LemmyResult<()> {
+    let process_index = self.opts.process_index - 1;
     info!(
       "Starting federation workers for process count {} and index {}",
-    opts.process_count, process_index
+      self.opts.process_count, process_index
     );
+    let local_domain = self.context.settings().get_hostname_without_port()?;
+    let mut pool = self.context.pool();
     loop {
       let mut total_count = 0;
       let mut dead_count = 0;
       let mut disallowed_count = 0;
       for (instance, allowed, is_dead) in
-      Instance::read_federated_with_blocked_and_dead(pool2).await?
+        Instance::read_federated_with_blocked_and_dead(&mut pool).await?
       {
         if instance.domain == local_domain {
           continue;
         }
-      if instance.id.inner() % opts.process_count != process_index {
+        if instance.id.inner() % self.opts.process_count != process_index {
           continue;
         }
         total_count += 1;
@@ -71,136 +99,228 @@ async fn start_stop_federation_workers(
         }
         let should_federate = allowed && !is_dead;
         if should_federate {
-        if workers.contains_key(&instance.id) {
+          if self.workers.contains_key(&instance.id) {
             // worker already running
             continue;
           }
           // create new worker
-        let config = federation_config.clone();
-        let stats_sender = stats_sender.clone();
-        let pool = pool.clone();
-        workers.insert(
-          instance.id,
-          CancellableTask::spawn(WORKER_EXIT_TIMEOUT, move |stop| {
           let instance = instance.clone();
-            let req_data = config.clone().to_request_data();
+          let req_data = self.context.to_request_data();
-            let stats_sender = stats_sender.clone();
+          let stats_sender = self.stats_sender.clone();
-            let pool = pool.clone();
+          self.workers.insert(
-            async move {
+            instance.id,
-              InstanceWorker::init_and_loop(
+            CancellableTask::spawn(WORKER_EXIT_TIMEOUT, move |stop| async move {
-                instance,
+              InstanceWorker::init_and_loop(instance, req_data, stop, stats_sender).await?;
-                req_data,
+              Ok(())
-                &mut DbPool::Pool(&pool),
-                stop,
-                stats_sender,
-              )
-              .await
-            }
             }),
           );
         } else if !should_federate {
-        if let Some(worker) = workers.remove(&instance.id) {
+          if let Some(worker) = self.workers.remove(&instance.id) {
             if let Err(e) = worker.cancel().await {
               tracing::error!("error stopping worker: {e}");
             }
           }
         }
       }
-    let worker_count = workers.len();
+      let worker_count = self.workers.len();
       tracing::info!("Federating to {worker_count}/{total_count} instances ({dead_count} dead, {disallowed_count} disallowed)");
       tokio::select! {
         () = sleep(INSTANCES_RECHECK_DELAY) => {},
-      _ = cancel.cancelled() => { break; }
+        _ = cancel.cancelled() => { return Ok(()) }
       }
     }
-  drop(stats_sender);
+  }
+
+  pub async fn cancel(self) -> LemmyResult<()> {
+    drop(self.stats_sender);
     tracing::warn!(
       "Waiting for {} workers ({:.2?} max)",
-    workers.len(),
+      self.workers.len(),
       WORKER_EXIT_TIMEOUT
     );
     // the cancel futures need to be awaited concurrently for the shutdown processes to be triggered
     // concurrently
-  futures::future::join_all(workers.into_values().map(util::CancellableTask::cancel)).await;
+    futures::future::join_all(
-  exit_print.await?;
+      self
+        .workers
+        .into_values()
+        .map(util::CancellableTask::cancel),
+    )
+    .await;
+    self.exit_print.await?;
     Ok(())
+  }
 }
 
-/// starts and stops federation workers depending on which instances are on db
+#[cfg(test)]
-/// await the returned future to stop/cancel all workers gracefully
+#[allow(clippy::unwrap_used)]
-pub fn start_stop_federation_workers_cancellable(
+#[allow(clippy::indexing_slicing)]
-  opts: Opts,
+mod test {
-  pool: ActualDbPool,
+
-  config: FederationConfig<LemmyContext>,
+  use super::*;
-) -> CancellableTask {
+  use activitypub_federation::config::Data;
-  CancellableTask::spawn(WORKER_EXIT_TIMEOUT, move |stop| {
+  use chrono::DateTime;
-    let opts = opts.clone();
+  use lemmy_db_schema::source::{
-    let pool = pool.clone();
+    federation_allowlist::FederationAllowList,
-    let config = config.clone();
+    federation_blocklist::FederationBlockList,
-    async move { start_stop_federation_workers(opts, pool, config, stop).await }
+    instance::InstanceForm,
+  };
+  use lemmy_utils::error::LemmyError;
+  use serial_test::serial;
+  use std::{
+    collections::HashSet,
+    sync::{Arc, Mutex},
+  };
+  use tokio::{spawn, time::sleep};
+
+  struct TestData {
+    send_manager: SendManager,
+    context: Data<LemmyContext>,
+    instances: Vec<Instance>,
+  }
+  impl TestData {
+    async fn init(process_count: i32, process_index: i32) -> LemmyResult<Self> {
+      let context = LemmyContext::init_test_context().await;
+      let opts = Opts {
+        process_count,
+        process_index,
+      };
+      let federation_config = FederationConfig::builder()
+        .domain("local.com")
+        .app_data(context.clone())
+        .build()
+        .await?;
+
+      let pool = &mut context.pool();
+      let instances = vec![
+        Instance::read_or_create(pool, "alpha.com".to_string()).await?,
+        Instance::read_or_create(pool, "beta.com".to_string()).await?,
+        Instance::read_or_create(pool, "gamma.com".to_string()).await?,
+      ];
+
+      let send_manager = SendManager::new(opts, federation_config);
+      Ok(Self {
+        send_manager,
+        context,
+        instances,
       })
-}
+    }
 
-/// every 60s, print the state for every instance. exits if the receiver is done (all senders
+    async fn run(&mut self) -> LemmyResult<()> {
-/// dropped)
+      // start it and cancel after workers are running
-async fn receive_print_stats(
+      let cancel = CancellationToken::new();
-  pool: ActualDbPool,
+      let cancel_ = cancel.clone();
-  mut receiver: UnboundedReceiver<(String, FederationQueueState)>,
+      spawn(async move {
-) {
+        sleep(Duration::from_millis(100)).await;
-  let pool = &mut DbPool::Pool(&pool);
+        cancel_.cancel();
-  let mut printerval = tokio::time::interval(Duration::from_secs(60));
+      });
-  printerval.tick().await; // skip first
+      self.send_manager.do_loop(cancel.clone()).await?;
-  let mut stats = HashMap::new();
+      Ok(())
-  loop {
-    tokio::select! {
-      ele = receiver.recv() => {
-        let Some((domain, ele)) = ele else {
-          print_stats(pool, &stats).await;
-          return;
-        };
-        stats.insert(domain, ele);
-      },
-      _ = printerval.tick() => {
-        print_stats(pool, &stats).await;
     }
-    }
-  }
-}
 
-async fn print_stats(pool: &mut DbPool<'_>, stats: &HashMap<String, FederationQueueState>) {
+    async fn cleanup(self) -> LemmyResult<()> {
-  let last_id = crate::util::get_latest_activity_id(pool).await;
+      self.send_manager.cancel().await?;
-  let Ok(last_id) = last_id else {
+      Instance::delete_all(&mut self.context.pool()).await?;
-    tracing::error!("could not get last id");
+      Ok(())
-    return;
+    }
+  }
+
+  /// Basic test with default params and only active/allowed instances
+  #[tokio::test]
+  #[serial]
+  async fn test_send_manager() -> LemmyResult<()> {
+    let mut data = TestData::init(1, 1).await?;
+
+    data.run().await?;
+    assert_eq!(3, data.send_manager.workers.len());
+    let workers: HashSet<_> = data.send_manager.workers.keys().cloned().collect();
+    let instances: HashSet<_> = data.instances.iter().map(|i| i.id).collect();
+    assert_eq!(instances, workers);
+
+    data.cleanup().await?;
+    Ok(())
+  }
+
+  /// Running with multiple processes should start correct workers
+  #[tokio::test]
+  #[serial]
+  async fn test_send_manager_processes() -> LemmyResult<()> {
+    let active = Arc::new(Mutex::new(vec![]));
+    let execute = |count, index, active: Arc<Mutex<Vec<InstanceId>>>| async move {
+      let mut data = TestData::init(count, index).await?;
+      data.run().await?;
+      assert_eq!(1, data.send_manager.workers.len());
+      for k in data.send_manager.workers.keys() {
+        active.lock().unwrap().push(*k);
+      }
+      data.cleanup().await?;
+      Ok::<(), LemmyError>(())
     };
-  // it's expected that the values are a bit out of date, everything < SAVE_STATE_EVERY should be
+    execute(3, 1, active.clone()).await?;
-  // considered up to date
+    execute(3, 2, active.clone()).await?;
-  tracing::info!(
+    execute(3, 3, active.clone()).await?;
-    "Federation state as of {}:",
+
-    Local::now()
+    // Should run exactly three workers
-      .with_nanosecond(0)
+    assert_eq!(3, active.lock().unwrap().len());
-      .expect("0 is valid nanos")
+
-      .to_rfc3339()
+    Ok(())
-  );
-  // todo: more stats (act/sec, avg http req duration)
-  let mut ok_count = 0;
-  let mut behind_count = 0;
-  for (domain, stat) in stats {
-    let behind = last_id.0 - stat.last_successful_id.map(|e| e.0).unwrap_or(0);
-    if stat.fail_count > 0 {
-      tracing::info!(
-        "{}: Warning. {} behind, {} consecutive fails, current retry delay {:.2?}",
-        domain,
-        behind,
-        stat.fail_count,
-        federate_retry_sleep_duration(stat.fail_count)
-      );
-    } else if behind > 0 {
-      tracing::debug!("{}: Ok. {} activities behind", domain, behind);
-      behind_count += 1;
-    } else {
-      ok_count += 1;
   }
+
+  /// Use blocklist, should not send to blocked instances
+  #[tokio::test]
+  #[serial]
+  async fn test_send_manager_blocked() -> LemmyResult<()> {
+    let mut data = TestData::init(1, 1).await?;
+
+    let domain = data.instances[0].domain.clone();
+    FederationBlockList::replace(&mut data.context.pool(), Some(vec![domain])).await?;
+    data.run().await?;
+    let workers = &data.send_manager.workers;
+    assert_eq!(2, workers.len());
+    assert!(workers.contains_key(&data.instances[1].id));
+    assert!(workers.contains_key(&data.instances[2].id));
+
+    data.cleanup().await?;
+    Ok(())
+  }
+
+  /// Use allowlist, should only send to allowed instance
+  #[tokio::test]
+  #[serial]
+  async fn test_send_manager_allowed() -> LemmyResult<()> {
+    let mut data = TestData::init(1, 1).await?;
+
+    let domain = data.instances[0].domain.clone();
+    FederationAllowList::replace(&mut data.context.pool(), Some(vec![domain])).await?;
+    data.run().await?;
+    let workers = &data.send_manager.workers;
+    assert_eq!(1, workers.len());
+    assert!(workers.contains_key(&data.instances[0].id));
+
+    data.cleanup().await?;
+    Ok(())
+  }
+
+  /// Mark instance as dead, there should be no worker created for it
+  #[tokio::test]
+  #[serial]
+  async fn test_send_manager_dead() -> LemmyResult<()> {
+    let mut data = TestData::init(1, 1).await?;
+
+    let instance = &data.instances[0];
+    let form = InstanceForm::builder()
+      .domain(instance.domain.clone())
+      .updated(DateTime::from_timestamp(0, 0))
+      .build();
+    Instance::update(&mut data.context.pool(), instance.id, form).await?;
+
+    data.run().await?;
+    let workers = &data.send_manager.workers;
+    assert_eq!(2, workers.len());
+    assert!(workers.contains_key(&data.instances[1].id));
+    assert!(workers.contains_key(&data.instances[2].id));
+
+    data.cleanup().await?;
+    Ok(())
   }
-  tracing::info!("{ok_count} others up to date. {behind_count} instances behind.");
 }

crates/federate/src/stats.rs

@@ -0,0 +1,97 @@
+use crate::util::get_latest_activity_id;
+use chrono::Local;
+use diesel::result::Error::NotFound;
+use lemmy_api_common::federate_retry_sleep_duration;
+use lemmy_db_schema::{
+  newtypes::InstanceId,
+  source::{federation_queue_state::FederationQueueState, instance::Instance},
+  utils::{ActualDbPool, DbPool},
+};
+use lemmy_utils::{error::LemmyResult, CACHE_DURATION_FEDERATION};
+use moka::future::Cache;
+use once_cell::sync::Lazy;
+use std::{collections::HashMap, time::Duration};
+use tokio::{sync::mpsc::UnboundedReceiver, time::interval};
+use tracing::{debug, info, warn};
+
+/// every 60s, print the state for every instance. exits if the receiver is done (all senders
+/// dropped)
+pub(crate) async fn receive_print_stats(
+  pool: ActualDbPool,
+  mut receiver: UnboundedReceiver<(InstanceId, FederationQueueState)>,
+) {
+  let pool = &mut DbPool::Pool(&pool);
+  let mut printerval = interval(Duration::from_secs(60));
+  let mut stats = HashMap::new();
+  loop {
+    tokio::select! {
+      ele = receiver.recv() => {
+        match ele {
+          // update stats for instance
+          Some((instance_id, ele)) => {stats.insert(instance_id, ele);},
+          // receiver closed, print stats and exit
+          None => {
+            print_stats(pool, &stats).await;
+            return;
+          }
+        }
+      },
+      _ = printerval.tick() => {
+        print_stats(pool, &stats).await;
+      }
+    }
+  }
+}
+
+async fn print_stats(pool: &mut DbPool<'_>, stats: &HashMap<InstanceId, FederationQueueState>) {
+  let res = print_stats_with_error(pool, stats).await;
+  if let Err(e) = res {
+    warn!("Failed to print stats: {e}");
+  }
+}
+
+async fn print_stats_with_error(
+  pool: &mut DbPool<'_>,
+  stats: &HashMap<InstanceId, FederationQueueState>,
+) -> LemmyResult<()> {
+  static INSTANCE_CACHE: Lazy<Cache<(), Vec<Instance>>> = Lazy::new(|| {
+    Cache::builder()
+      .max_capacity(1)
+      .time_to_live(CACHE_DURATION_FEDERATION)
+      .build()
+  });
+  let instances = INSTANCE_CACHE
+    .try_get_with((), async { Instance::read_all(pool).await })
+    .await?;
+
+  let last_id = get_latest_activity_id(pool).await?;
+
+  // it's expected that the values are a bit out of date, everything < SAVE_STATE_EVERY should be
+  // considered up to date
+  info!("Federation state as of {}:", Local::now().to_rfc3339());
+  // todo: more stats (act/sec, avg http req duration)
+  let mut ok_count = 0;
+  let mut behind_count = 0;
+  for (instance_id, stat) in stats {
+    let domain = &instances
+      .iter()
+      .find(|i| &i.id == instance_id)
+      .ok_or(NotFound)?
+      .domain;
+    let behind = last_id.0 - stat.last_successful_id.map(|e| e.0).unwrap_or(0);
+    if stat.fail_count > 0 {
+      info!(
+        "{domain}: Warning. {behind} behind, {} consecutive fails, current retry delay {:.2?}",
+        stat.fail_count,
+        federate_retry_sleep_duration(stat.fail_count)
+      );
+    } else if behind > 0 {
+      debug!("{}: Ok. {} activities behind", domain, behind);
+      behind_count += 1;
+    } else {
+      ok_count += 1;
+    }
+  }
+  info!("{ok_count} others up to date. {behind_count} instances behind.");
+  Ok(())
+}

crates/federate/src/util.rs

@@ -17,6 +17,7 @@ use lemmy_db_schema::{
   traits::ApubActor,
   utils::{get_conn, DbPool},
 };
+use lemmy_utils::error::LemmyResult;
 use moka::future::Cache;
 use once_cell::sync::Lazy;
 use reqwest::Url;
@@ -24,6 +25,7 @@ use serde_json::Value;
 use std::{fmt::Debug, future::Future, pin::Pin, sync::Arc, time::Duration};
 use tokio::{task::JoinHandle, time::sleep};
 use tokio_util::sync::CancellationToken;
+use tracing::error;
 
 /// Decrease the delays of the federation queue.
 /// Should only be used for federation tests since it significantly increases CPU and DB load of the
@@ -59,36 +61,29 @@ impl CancellableTask {
   /// spawn a task but with graceful shutdown
   pub fn spawn<F, R: Debug>(
     timeout: Duration,
-    task: impl Fn(CancellationToken) -> F + Send + 'static,
+    task: impl FnOnce(CancellationToken) -> F + Send + 'static,
   ) -> CancellableTask
   where
-    F: Future<Output = R> + Send + 'static,
+    F: Future<Output = LemmyResult<R>> + Send + 'static,
+    R: Send + 'static,
   {
     let stop = CancellationToken::new();
     let stop2 = stop.clone();
-    let task: JoinHandle<()> = tokio::spawn(async move {
+    let task: JoinHandle<LemmyResult<R>> = tokio::spawn(task(stop2));
-      loop {
-        let res = task(stop2.clone()).await;
-        if stop2.is_cancelled() {
-          return;
-        } else {
-          tracing::warn!("task exited, restarting: {res:?}");
-        }
-      }
-    });
     let abort = task.abort_handle();
     CancellableTask {
       f: Box::pin(async move {
         stop.cancel();
         tokio::select! {
             r = task => {
-              r.context("could not join")?;
+              if let Err(ref e) = r? {
+                error!("CancellableTask threw error: {e}");
+              }
               Ok(())
             },
             _ = sleep(timeout) => {
                 abort.abort();
-                tracing::warn!("Graceful shutdown timed out, aborting task");
+                Err(anyhow!("CancellableTask aborted due to shutdown timeout"))
-                Err(anyhow!("task aborted due to timeout"))
             }
         }
       }),

crates/federate/src/worker.rs

@@ -22,7 +22,7 @@ use lemmy_db_schema::{
     instance::{Instance, InstanceForm},
     site::Site,
   },
-  utils::{naive_now, DbPool},
+  utils::naive_now,
 };
 use lemmy_db_views_actor::structs::CommunityFollowerView;
 use once_cell::sync::Lazy;
@@ -75,7 +75,7 @@ pub(crate) struct InstanceWorker {
   followed_communities: HashMap<CommunityId, HashSet<Url>>,
   stop: CancellationToken,
   context: Data<LemmyContext>,
-  stats_sender: UnboundedSender<(String, FederationQueueState)>,
+  stats_sender: UnboundedSender<(InstanceId, FederationQueueState)>,
   last_full_communities_fetch: DateTime<Utc>,
   last_incremental_communities_fetch: DateTime<Utc>,
   state: FederationQueueState,
@@ -86,12 +86,11 @@ impl InstanceWorker {
   pub(crate) async fn init_and_loop(
     instance: Instance,
     context: Data<LemmyContext>,
-    pool: &mut DbPool<'_>, /* in theory there's a ref to the pool in context, but i couldn't get
-                            * that to work wrt lifetimes */
     stop: CancellationToken,
-    stats_sender: UnboundedSender<(String, FederationQueueState)>,
+    stats_sender: UnboundedSender<(InstanceId, FederationQueueState)>,
   ) -> Result<(), anyhow::Error> {
-    let state = FederationQueueState::load(pool, instance.id).await?;
+    let mut pool = context.pool();
+    let state = FederationQueueState::load(&mut pool, instance.id).await?;
     let mut worker = InstanceWorker {
       instance,
       site_loaded: false,
@@ -105,32 +104,29 @@ impl InstanceWorker {
       state,
       last_state_insert: Utc.timestamp_nanos(0),
     };
-    worker.loop_until_stopped(pool).await
+    worker.loop_until_stopped().await
   }
   /// loop fetch new activities from db and send them to the inboxes of the given instances
   /// this worker only returns if (a) there is an internal error or (b) the cancellation token is
   /// cancelled (graceful exit)
-  pub(crate) async fn loop_until_stopped(
+  pub(crate) async fn loop_until_stopped(&mut self) -> Result<(), anyhow::Error> {
-    &mut self,
-    pool: &mut DbPool<'_>,
-  ) -> Result<(), anyhow::Error> {
     debug!("Starting federation worker for {}", self.instance.domain);
     let save_state_every = chrono::Duration::from_std(SAVE_STATE_EVERY_TIME).expect("not negative");
 
-    self.update_communities(pool).await?;
+    self.update_communities().await?;
     self.initial_fail_sleep().await?;
     while !self.stop.is_cancelled() {
-      self.loop_batch(pool).await?;
+      self.loop_batch().await?;
       if self.stop.is_cancelled() {
         break;
       }
       if (Utc::now() - self.last_state_insert) > save_state_every {
-        self.save_and_send_state(pool).await?;
+        self.save_and_send_state().await?;
       }
-      self.update_communities(pool).await?;
+      self.update_communities().await?;
     }
     // final update of state in db
-    self.save_and_send_state(pool).await?;
+    self.save_and_send_state().await?;
     Ok(())
   }
 
@@ -155,8 +151,8 @@ impl InstanceWorker {
     Ok(())
   }
   /// send out a batch of CHECK_SAVE_STATE_EVERY_IT activities
-  async fn loop_batch(&mut self, pool: &mut DbPool<'_>) -> Result<()> {
+  async fn loop_batch(&mut self) -> Result<()> {
-    let latest_id = get_latest_activity_id(pool).await?;
+    let latest_id = get_latest_activity_id(&mut self.context.pool()).await?;
     let mut id = if let Some(id) = self.state.last_successful_id {
       id
     } else {
@@ -166,7 +162,7 @@ impl InstanceWorker {
       // skip all past activities:
       self.state.last_successful_id = Some(latest_id);
       // save here to ensure it's not read as 0 again later if no activities have happened
-      self.save_and_send_state(pool).await?;
+      self.save_and_send_state().await?;
       latest_id
     };
     if id >= latest_id {
@@ -184,7 +180,7 @@ impl InstanceWorker {
     {
       id = ActivityId(id.0 + 1);
       processed_activities += 1;
-      let Some(ele) = get_activity_cached(pool, id)
+      let Some(ele) = get_activity_cached(&mut self.context.pool(), id)
         .await
         .context("failed reading activity from db")?
       else {
@@ -192,7 +188,7 @@ impl InstanceWorker {
         self.state.last_successful_id = Some(id);
         continue;
       };
-      if let Err(e) = self.send_retry_loop(pool, &ele.0, &ele.1).await {
+      if let Err(e) = self.send_retry_loop(&ele.0, &ele.1).await {
         warn!(
           "sending {} errored internally, skipping activity: {:?}",
           ele.0.ap_id, e
@@ -213,12 +209,11 @@ impl InstanceWorker {
   // and will return an error if an internal error occurred (send errors cause an infinite loop)
   async fn send_retry_loop(
     &mut self,
-    pool: &mut DbPool<'_>,
     activity: &SentActivity,
     object: &SharedInboxActivities,
   ) -> Result<()> {
     let inbox_urls = self
-      .get_inbox_urls(pool, activity)
+      .get_inbox_urls(activity)
       .await
       .context("failed figuring out inbox urls")?;
     if inbox_urls.is_empty() {
@@ -230,7 +225,7 @@ impl InstanceWorker {
     let Some(actor_apub_id) = &activity.actor_apub_id else {
       return Ok(()); // activity was inserted before persistent queue was activated
     };
-    let actor = get_actor_cached(pool, activity.actor_type, actor_apub_id)
+    let actor = get_actor_cached(&mut self.context.pool(), activity.actor_type, actor_apub_id)
       .await
       .context("failed getting actor instance (was it marked deleted / removed?)")?;
 
@@ -249,7 +244,7 @@ impl InstanceWorker {
           "{}: retrying {:?} attempt {} with delay {retry_delay:.2?}. ({e})",
           self.instance.domain, activity.id, self.state.fail_count
         );
-        self.save_and_send_state(pool).await?;
+        self.save_and_send_state().await?;
         tokio::select! {
           () = sleep(retry_delay) => {},
           () = self.stop.cancelled() => {
@@ -268,7 +263,7 @@ impl InstanceWorker {
           .domain(self.instance.domain.clone())
           .updated(Some(naive_now()))
           .build();
-        Instance::update(pool, self.instance.id, form).await?;
+        Instance::update(&mut self.context.pool(), self.instance.id, form).await?;
       }
     }
     Ok(())
@@ -278,16 +273,12 @@ impl InstanceWorker {
   /// most often this will return 0 values (if instance doesn't care about the activity)
   /// or 1 value (the shared inbox)
   /// > 1 values only happens for non-lemmy software
-  async fn get_inbox_urls(
+  async fn get_inbox_urls(&mut self, activity: &SentActivity) -> Result<HashSet<Url>> {
-    &mut self,
-    pool: &mut DbPool<'_>,
-    activity: &SentActivity,
-  ) -> Result<HashSet<Url>> {
     let mut inbox_urls: HashSet<Url> = HashSet::new();
 
     if activity.send_all_instances {
       if !self.site_loaded {
-        self.site = Site::read_from_instance_id(pool, self.instance.id).await?;
+        self.site = Site::read_from_instance_id(&mut self.context.pool(), self.instance.id).await?;
         self.site_loaded = true;
       }
       if let Some(site) = &self.site {
@@ -312,22 +303,18 @@ impl InstanceWorker {
     Ok(inbox_urls)
   }
 
-  async fn update_communities(&mut self, pool: &mut DbPool<'_>) -> Result<()> {
+  async fn update_communities(&mut self) -> Result<()> {
     if (Utc::now() - self.last_full_communities_fetch) > *FOLLOW_REMOVALS_RECHECK_DELAY {
       // process removals every hour
       (self.followed_communities, self.last_full_communities_fetch) = self
-        .get_communities(pool, self.instance.id, Utc.timestamp_nanos(0))
+        .get_communities(self.instance.id, Utc.timestamp_nanos(0))
         .await?;
       self.last_incremental_communities_fetch = self.last_full_communities_fetch;
     }
     if (Utc::now() - self.last_incremental_communities_fetch) > *FOLLOW_ADDITIONS_RECHECK_DELAY {
       // process additions every minute
       let (news, time) = self
-        .get_communities(
+        .get_communities(self.instance.id, self.last_incremental_communities_fetch)
-          pool,
-          self.instance.id,
-          self.last_incremental_communities_fetch,
-        )
         .await?;
       self.followed_communities.extend(news);
       self.last_incremental_communities_fetch = time;
@@ -339,7 +326,6 @@ impl InstanceWorker {
   /// them
   async fn get_communities(
     &mut self,
-    pool: &mut DbPool<'_>,
     instance_id: InstanceId,
     last_fetch: DateTime<Utc>,
   ) -> Result<(HashMap<CommunityId, HashSet<Url>>, DateTime<Utc>)> {
@@ -347,7 +333,11 @@ impl InstanceWorker {
       Utc::now() - chrono::TimeDelta::try_seconds(10).expect("TimeDelta out of bounds"); // update to time before fetch to ensure overlap. subtract 10s to ensure overlap even if
                                                                                          // published date is not exact
     Ok((
-      CommunityFollowerView::get_instance_followed_community_inboxes(pool, instance_id, last_fetch)
+      CommunityFollowerView::get_instance_followed_community_inboxes(
+        &mut self.context.pool(),
+        instance_id,
+        last_fetch,
+      )
       .await?
       .into_iter()
       .fold(HashMap::new(), |mut map, (c, u)| {
@@ -357,12 +347,12 @@ impl InstanceWorker {
       new_last_fetch,
     ))
   }
-  async fn save_and_send_state(&mut self, pool: &mut DbPool<'_>) -> Result<()> {
+  async fn save_and_send_state(&mut self) -> Result<()> {
     self.last_state_insert = Utc::now();
-    FederationQueueState::upsert(pool, &self.state).await?;
+    FederationQueueState::upsert(&mut self.context.pool(), &self.state).await?;
     self
       .stats_sender
-      .send((self.instance.domain.clone(), self.state.clone()))?;
+      .send((self.instance.id, self.state.clone()))?;
     Ok(())
   }
 }

crates/utils/translations

@@ -1 +1 @@
-Subproject commit 4e7c7ad4fcb4d8618d93ec17d72379367aa085b1
+Subproject commit e9b3b25fa1af7e06c4ffab86624d95da0836ef36

scripts/test.sh

@@ -20,11 +20,10 @@ then
   cargo test -p $PACKAGE --all-features --no-fail-fast $TEST
 else
   cargo test --workspace --no-fail-fast
+  # Testing lemmy utils all features in particular (for ts-rs bindings)
+  cargo test -p lemmy_utils --all-features --no-fail-fast
 fi
 
-# Testing lemmy utils all features in particular (for ts-rs bindings)
-cargo test -p lemmy_utils --all-features --no-fail-fast
-
 # Add this to do printlns: -- --nocapture
 
 pg_ctl stop --silent

src/lib.rs

@@ -41,7 +41,7 @@ use lemmy_apub::{
   FEDERATION_HTTP_FETCH_LIMIT,
 };
 use lemmy_db_schema::{schema_setup, source::secret::Secret, utils::build_db_pool};
-use lemmy_federate::{start_stop_federation_workers_cancellable, Opts};
+use lemmy_federate::{Opts, SendManager};
 use lemmy_routes::{feeds, images, nodeinfo, webfinger};
 use lemmy_utils::{
   error::LemmyResult,
@@ -253,14 +253,14 @@ pub async fn start_lemmy_server(args: CmdArgs) -> LemmyResult<()> {
     None
   };
   let federate = (!args.disable_activity_sending).then(|| {
-    start_stop_federation_workers_cancellable(
+    let task = SendManager::new(
       Opts {
         process_index: args.federate_process_index,
         process_count: args.federate_process_count,
       },
-      pool.clone(),
+      federation_config,
-      federation_config.clone(),
+    );
-    )
+    task.run()
   });
   let mut interrupt = tokio::signal::unix::signal(SignalKind::interrupt())?;
   let mut terminate = tokio::signal::unix::signal(SignalKind::terminate())?;