]> Untitled Git - lemmy.git/blobdiff - src/scheduled_tasks.rs
Cache & Optimize Woodpecker CI (#3450)
[lemmy.git] / src / scheduled_tasks.rs
index 5d98baf9b4479f3c17b2afd5671179d019c33580..4928fe8dd5c0f836757f2bb1ce92431794f5fd6f 100644 (file)
@@ -1,55 +1,99 @@
+use chrono::NaiveDateTime;
 use clokwerk::{Scheduler, TimeUnits as CTimeUnits};
 use diesel::{
   dsl::{now, IntervalDsl},
+  sql_types::{Integer, Timestamp},
   Connection,
   ExpressionMethods,
+  NullableExpressionMethods,
   QueryDsl,
+  QueryableByName,
 };
 // Import week days and WeekDay
 use diesel::{sql_query, PgConnection, RunQueryDsl};
+use lemmy_api_common::context::LemmyContext;
 use lemmy_db_schema::{
-  schema::{activity, community_person_ban, instance, person},
+  schema::{
+    captcha_answer,
+    comment,
+    community_person_ban,
+    instance,
+    person,
+    post,
+    received_activity,
+    sent_activity,
+  },
   source::instance::{Instance, InstanceForm},
-  utils::naive_now,
+  utils::{naive_now, DELETED_REPLACEMENT_TEXT},
 };
 use lemmy_routes::nodeinfo::NodeInfo;
-use lemmy_utils::{error::LemmyError, REQWEST_TIMEOUT};
+use lemmy_utils::{
+  error::{LemmyError, LemmyResult},
+  REQWEST_TIMEOUT,
+};
 use reqwest::blocking::Client;
 use std::{thread, time::Duration};
-use tracing::info;
+use tracing::{error, info, warn};
 
 /// Schedules various cleanup tasks for lemmy in a background thread
-pub fn setup(db_url: String, user_agent: String) -> Result<(), LemmyError> {
+pub fn setup(
+  db_url: String,
+  user_agent: String,
+  context_1: LemmyContext,
+) -> Result<(), LemmyError> {
   // Setup the connections
   let mut scheduler = Scheduler::new();
 
-  let mut conn = PgConnection::establish(&db_url).expect("could not establish connection");
+  startup_jobs(&db_url);
 
-  let mut conn_2 = PgConnection::establish(&db_url).expect("could not establish connection");
+  // Update active counts every hour
+  let url = db_url.clone();
+  scheduler.every(CTimeUnits::hour(1)).run(move || {
+    let mut conn = PgConnection::establish(&url).expect("could not establish connection");
+    active_counts(&mut conn);
+    update_banned_when_expired(&mut conn);
+  });
 
-  active_counts(&mut conn);
-  update_banned_when_expired(&mut conn);
+  // Update hot ranks every 15 minutes
+  let url = db_url.clone();
+  scheduler.every(CTimeUnits::minutes(15)).run(move || {
+    let mut conn = PgConnection::establish(&url).expect("could not establish connection");
+    update_hot_ranks(&mut conn);
+  });
 
-  // On startup, reindex the tables non-concurrently
-  // TODO remove this for now, since it slows down startup a lot on lemmy.ml
-  reindex_aggregates_tables(&mut conn, true);
-  scheduler.every(CTimeUnits::hour(1)).run(move || {
-    let conn = &mut PgConnection::establish(&db_url)
-      .unwrap_or_else(|_| panic!("Error connecting to {db_url}"));
-    active_counts(conn);
-    update_banned_when_expired(conn);
-    reindex_aggregates_tables(conn, true);
-    drop_ccnew_indexes(conn);
+  // Delete any captcha answers older than ten minutes, every ten minutes
+  let url = db_url.clone();
+  scheduler.every(CTimeUnits::minutes(10)).run(move || {
+    let mut conn = PgConnection::establish(&url).expect("could not establish connection");
+    delete_expired_captcha_answers(&mut conn);
   });
 
-  clear_old_activities(&mut conn);
+  // Clear old activities every week
+  let url = db_url.clone();
   scheduler.every(CTimeUnits::weeks(1)).run(move || {
+    let mut conn = PgConnection::establish(&url).expect("could not establish connection");
     clear_old_activities(&mut conn);
   });
 
-  update_instance_software(&mut conn_2, &user_agent);
+  // Remove old rate limit buckets after 1 to 2 hours of inactivity
+  scheduler.every(CTimeUnits::hour(1)).run(move || {
+    let hour = Duration::from_secs(3600);
+    context_1.settings_updated_channel().remove_older_than(hour);
+  });
+
+  // Overwrite deleted & removed posts and comments every day
+  let url = db_url.clone();
+  scheduler.every(CTimeUnits::days(1)).run(move || {
+    let mut conn = PgConnection::establish(&url).expect("could not establish connection");
+    overwrite_deleted_posts_and_comments(&mut conn);
+  });
+
+  // Update the Instance Software
   scheduler.every(CTimeUnits::days(1)).run(move || {
-    update_instance_software(&mut conn_2, &user_agent);
+    let mut conn = PgConnection::establish(&db_url).expect("could not establish connection");
+    update_instance_software(&mut conn, &user_agent)
+      .map_err(|e| warn!("Failed to update instance software: {e}"))
+      .ok();
   });
 
   // Manually run the scheduler in an event loop
@@ -59,34 +103,176 @@ pub fn setup(db_url: String, user_agent: String) -> Result<(), LemmyError> {
   }
 }
 
-/// Reindex the aggregates tables every one hour
-/// This is necessary because hot_rank is actually a mutable function:
-/// https://dba.stackexchange.com/questions/284052/how-to-create-an-index-based-on-a-time-based-function-in-postgres?noredirect=1#comment555727_284052
-fn reindex_aggregates_tables(conn: &mut PgConnection, concurrently: bool) {
-  for table_name in &[
+/// Run these on server startup
+fn startup_jobs(db_url: &str) {
+  let mut conn = PgConnection::establish(db_url).expect("could not establish connection");
+  active_counts(&mut conn);
+  update_hot_ranks(&mut conn);
+  update_banned_when_expired(&mut conn);
+  clear_old_activities(&mut conn);
+  overwrite_deleted_posts_and_comments(&mut conn);
+}
+
+/// Update the hot_rank columns for the aggregates tables
+/// Runs in batches until all necessary rows are updated once
+fn update_hot_ranks(conn: &mut PgConnection) {
+  info!("Updating hot ranks for all history...");
+
+  process_hot_ranks_in_batches(
+    conn,
     "post_aggregates",
+    "a.hot_rank != 0 OR a.hot_rank_active != 0",
+    "SET hot_rank = hot_rank(a.score, a.published),
+         hot_rank_active = hot_rank(a.score, a.newest_comment_time_necro)",
+  );
+
+  process_hot_ranks_in_batches(
+    conn,
     "comment_aggregates",
+    "a.hot_rank != 0",
+    "SET hot_rank = hot_rank(a.score, a.published)",
+  );
+
+  process_hot_ranks_in_batches(
+    conn,
     "community_aggregates",
-  ] {
-    reindex_table(conn, table_name, concurrently);
+    "a.hot_rank != 0",
+    "SET hot_rank = hot_rank(a.subscribers, a.published)",
+  );
+
+  info!("Finished hot ranks update!");
+}
+
+#[derive(QueryableByName)]
+struct HotRanksUpdateResult {
+  #[diesel(sql_type = Timestamp)]
+  published: NaiveDateTime,
+}
+
+/// Runs the hot rank update query in batches until all rows have been processed.
+/// In `where_clause` and `set_clause`, "a" will refer to the current aggregates table.
+/// Locked rows are skipped in order to prevent deadlocks (they will likely get updated on the next
+/// run)
+fn process_hot_ranks_in_batches(
+  conn: &mut PgConnection,
+  table_name: &str,
+  where_clause: &str,
+  set_clause: &str,
+) {
+  let process_start_time = NaiveDateTime::from_timestamp_opt(0, 0).expect("0 timestamp creation");
+
+  let update_batch_size = 1000; // Bigger batches than this tend to cause seq scans
+  let mut processed_rows_count = 0;
+  let mut previous_batch_result = Some(process_start_time);
+  while let Some(previous_batch_last_published) = previous_batch_result {
+    // Raw `sql_query` is used as a performance optimization - Diesel does not support doing this
+    // in a single query (neither as a CTE, nor using a subquery)
+    let result = sql_query(format!(
+      r#"WITH batch AS (SELECT a.id
+               FROM {aggregates_table} a
+               WHERE a.published > $1 AND ({where_clause})
+               ORDER BY a.published
+               LIMIT $2
+               FOR UPDATE SKIP LOCKED)
+         UPDATE {aggregates_table} a {set_clause}
+             FROM batch WHERE a.id = batch.id RETURNING a.published;
+    "#,
+      aggregates_table = table_name,
+      set_clause = set_clause,
+      where_clause = where_clause
+    ))
+    .bind::<Timestamp, _>(previous_batch_last_published)
+    .bind::<Integer, _>(update_batch_size)
+    .get_results::<HotRanksUpdateResult>(conn);
+
+    match result {
+      Ok(updated_rows) => {
+        processed_rows_count += updated_rows.len();
+        previous_batch_result = updated_rows.last().map(|row| row.published);
+      }
+      Err(e) => {
+        error!("Failed to update {} hot_ranks: {}", table_name, e);
+        break;
+      }
+    }
   }
+  info!(
+    "Finished process_hot_ranks_in_batches execution for {} (processed {} rows)",
+    table_name, processed_rows_count
+  );
 }
 
-fn reindex_table(conn: &mut PgConnection, table_name: &str, concurrently: bool) {
-  let concurrently_str = if concurrently { "concurrently" } else { "" };
-  info!("Reindexing table {} {} ...", concurrently_str, table_name);
-  let query = format!("reindex table {concurrently_str} {table_name}");
-  sql_query(query).execute(conn).expect("reindex table");
-  info!("Done.");
+fn delete_expired_captcha_answers(conn: &mut PgConnection) {
+  match diesel::delete(
+    captcha_answer::table.filter(captcha_answer::published.lt(now - IntervalDsl::minutes(10))),
+  )
+  .execute(conn)
+  {
+    Ok(_) => {
+      info!("Done.");
+    }
+    Err(e) => {
+      error!("Failed to clear old captcha answers: {}", e)
+    }
+  }
 }
 
 /// Clear old activities (this table gets very large)
 fn clear_old_activities(conn: &mut PgConnection) {
   info!("Clearing old activities...");
-  diesel::delete(activity::table.filter(activity::published.lt(now - 6.months())))
+  diesel::delete(sent_activity::table.filter(sent_activity::published.lt(now - 3.months())))
     .execute(conn)
-    .expect("clear old activities");
-  info!("Done.");
+    .map_err(|e| error!("Failed to clear old sent activities: {}", e))
+    .ok();
+
+  diesel::delete(
+    received_activity::table.filter(received_activity::published.lt(now - 3.months())),
+  )
+  .execute(conn)
+  .map_err(|e| error!("Failed to clear old received activities: {}", e))
+  .ok();
+}
+
+/// overwrite posts and comments 30d after deletion
+fn overwrite_deleted_posts_and_comments(conn: &mut PgConnection) {
+  info!("Overwriting deleted posts...");
+  match diesel::update(
+    post::table
+      .filter(post::deleted.eq(true))
+      .filter(post::updated.lt(now.nullable() - 1.months()))
+      .filter(post::body.ne(DELETED_REPLACEMENT_TEXT)),
+  )
+  .set((
+    post::body.eq(DELETED_REPLACEMENT_TEXT),
+    post::name.eq(DELETED_REPLACEMENT_TEXT),
+  ))
+  .execute(conn)
+  {
+    Ok(_) => {
+      info!("Done.");
+    }
+    Err(e) => {
+      error!("Failed to overwrite deleted posts: {}", e)
+    }
+  }
+
+  info!("Overwriting deleted comments...");
+  match diesel::update(
+    comment::table
+      .filter(comment::deleted.eq(true))
+      .filter(comment::updated.lt(now.nullable() - 1.months()))
+      .filter(comment::content.ne(DELETED_REPLACEMENT_TEXT)),
+  )
+  .set(comment::content.eq(DELETED_REPLACEMENT_TEXT))
+  .execute(conn)
+  {
+    Ok(_) => {
+      info!("Done.");
+    }
+    Err(e) => {
+      error!("Failed to overwrite deleted comments: {}", e)
+    }
+  }
 }
 
 /// Re-calculate the site and community active counts every 12 hours
@@ -102,17 +288,23 @@ fn active_counts(conn: &mut PgConnection) {
 
   for i in &intervals {
     let update_site_stmt = format!(
-      "update site_aggregates set users_active_{} = (select * from site_aggregates_activity('{}'))",
+      "update site_aggregates set users_active_{} = (select * from site_aggregates_activity('{}')) where site_id = 1",
       i.1, i.0
     );
-    sql_query(update_site_stmt)
-      .execute(conn)
-      .expect("update site stats");
+    match sql_query(update_site_stmt).execute(conn) {
+      Ok(_) => {}
+      Err(e) => {
+        error!("Failed to update site stats: {}", e)
+      }
+    }
 
     let update_community_stmt = format!("update community_aggregates ca set users_active_{} = mv.count_ from community_aggregates_activity('{}') mv where ca.community_id = mv.community_id_", i.1, i.0);
-    sql_query(update_community_stmt)
-      .execute(conn)
-      .expect("update community stats");
+    match sql_query(update_community_stmt).execute(conn) {
+      Ok(_) => {}
+      Err(e) => {
+        error!("Failed to update community stats: {}", e)
+      }
+    }
   }
 
   info!("Done.");
@@ -122,74 +314,96 @@ fn active_counts(conn: &mut PgConnection) {
 fn update_banned_when_expired(conn: &mut PgConnection) {
   info!("Updating banned column if it expires ...");
 
-  diesel::update(
+  match diesel::update(
     person::table
       .filter(person::banned.eq(true))
       .filter(person::ban_expires.lt(now)),
   )
   .set(person::banned.eq(false))
   .execute(conn)
-  .expect("update person.banned when expires");
-
-  diesel::delete(community_person_ban::table.filter(community_person_ban::expires.lt(now)))
-    .execute(conn)
-    .expect("remove community_ban expired rows");
-}
-
-/// Drops the phantom CCNEW indexes created by postgres
-/// https://github.com/LemmyNet/lemmy/issues/2431
-fn drop_ccnew_indexes(conn: &mut PgConnection) {
-  info!("Dropping phantom ccnew indexes...");
-  let drop_stmt = "select drop_ccnew_indexes()";
-  sql_query(drop_stmt)
+  {
+    Ok(_) => {}
+    Err(e) => {
+      error!("Failed to update person.banned when expires: {}", e)
+    }
+  }
+  match diesel::delete(community_person_ban::table.filter(community_person_ban::expires.lt(now)))
     .execute(conn)
-    .expect("drop ccnew indexes");
+  {
+    Ok(_) => {}
+    Err(e) => {
+      error!("Failed to remove community_ban expired rows: {}", e)
+    }
+  }
 }
 
 /// Updates the instance software and version
-fn update_instance_software(conn: &mut PgConnection, user_agent: &str) {
+///
+/// TODO: this should be async
+/// TODO: if instance has been dead for a long time, it should be checked less frequently
+fn update_instance_software(conn: &mut PgConnection, user_agent: &str) -> LemmyResult<()> {
   info!("Updating instances software and versions...");
 
   let client = Client::builder()
     .user_agent(user_agent)
     .timeout(REQWEST_TIMEOUT)
-    .build()
-    .expect("couldnt build reqwest client");
+    .build()?;
 
-  let instances = instance::table
-    .get_results::<Instance>(conn)
-    .expect("no instances found");
+  let instances = instance::table.get_results::<Instance>(conn)?;
 
   for instance in instances {
     let node_info_url = format!("https://{}/nodeinfo/2.0.json", instance.domain);
 
-    // Skip it if it can't connect
-    let res = client
-      .get(&node_info_url)
-      .send()
-      .ok()
-      .and_then(|t| t.json::<NodeInfo>().ok());
-
-    if let Some(node_info) = res {
-      let software = node_info.software.as_ref();
-      let form = InstanceForm::builder()
-        .domain(instance.domain)
-        .software(software.and_then(|s| s.name.clone()))
-        .version(software.and_then(|s| s.version.clone()))
-        .updated(Some(naive_now()))
-        .build();
-
+    // The `updated` column is used to check if instances are alive. If it is more than three days
+    // in the past, no outgoing activities will be sent to that instance. However not every
+    // Fediverse instance has a valid Nodeinfo endpoint (its not required for Activitypub). That's
+    // why we always need to mark instances as updated if they are alive.
+    let default_form = InstanceForm::builder()
+      .domain(instance.domain.clone())
+      .updated(Some(naive_now()))
+      .build();
+    let form = match client.get(&node_info_url).send() {
+      Ok(res) if res.status().is_client_error() => {
+        // Instance doesnt have nodeinfo but sent a response, consider it alive
+        Some(default_form)
+      }
+      Ok(res) => match res.json::<NodeInfo>() {
+        Ok(node_info) => {
+          // Instance sent valid nodeinfo, write it to db
+          Some(
+            InstanceForm::builder()
+              .domain(instance.domain)
+              .updated(Some(naive_now()))
+              .software(node_info.software.and_then(|s| s.name))
+              .version(node_info.version.clone())
+              .build(),
+          )
+        }
+        Err(_) => {
+          // No valid nodeinfo but valid HTTP response, consider instance alive
+          Some(default_form)
+        }
+      },
+      Err(_) => {
+        // dead instance, do nothing
+        None
+      }
+    };
+    if let Some(form) = form {
       diesel::update(instance::table.find(instance.id))
         .set(form)
-        .execute(conn)
-        .expect("update site instance software");
+        .execute(conn)?;
     }
   }
-  info!("Done.");
+  info!("Finished updating instances software and versions...");
+  Ok(())
 }
 
 #[cfg(test)]
 mod tests {
+  #![allow(clippy::unwrap_used)]
+  #![allow(clippy::indexing_slicing)]
+
   use lemmy_routes::nodeinfo::NodeInfo;
   use reqwest::Client;