perform async shutdown for database pool after services stop

Signed-off-by: Jason Volk <jason@zemos.net>

src/database/engine.rs

@@ -148,6 +148,8 @@ impl Engine {
 			.expect("column was created and exists")
 	}
 
+	pub async fn shutdown_pool(&self) { self.pool.shutdown().await; }
+
 	pub fn flush(&self) -> Result<()> { result(DBCommon::flush_wal(&self.db, false)) }
 
 	pub fn sync(&self) -> Result<()> { result(DBCommon::flush_wal(&self.db, true)) }
@@ -325,8 +327,7 @@ impl Drop for Engine {
 	fn drop(&mut self) {
 		const BLOCKING: bool = true;
 
-		debug!("Shutting down request pool...");
-		self.pool.close();
+		debug_assert!(!self.pool.close(), "request pool was not closed");
 
 		debug!("Waiting for background tasks to finish...");
 		self.db.cancel_all_background_work(BLOCKING);

src/database/pool.rs

@@ -76,10 +76,8 @@ pub(crate) async fn new(server: &Arc<Server>, opts: &Opts) -> Result<Arc<Self>>
 }
 
 #[implement(Pool)]
-pub(crate) async fn _shutdown(self: &Arc<Self>) {
-	if !self.queue.is_closed() {
+pub(crate) async fn shutdown(self: &Arc<Self>) {
 	self.close();
-	}
 
 	let workers = take(&mut *self.workers.lock().await);
 	debug!(workers = workers.len(), "Waiting for workers to join...");
@@ -89,18 +87,19 @@ pub(crate) async fn _shutdown(self: &Arc<Self>) {
 }
 
 #[implement(Pool)]
-pub(crate) fn close(&self) {
-	debug_assert!(!self.queue.is_closed(), "channel already closed");
+pub(crate) fn close(&self) -> bool {
+	if !self.queue.close() {
+		return false;
+	}
+
+	std::thread::yield_now();
 	debug!(
 		senders = self.queue.sender_count(),
 		receivers = self.queue.receiver_count(),
-		"Closing pool channel"
+		"Closed pool channel"
 	);
 
-	let closing = self.queue.close();
-	debug_assert!(closing, "channel is not closing");
-
-	std::thread::yield_now();
+	true
 }
 
 #[implement(Pool)]

src/router/run.rs

@@ -78,6 +78,11 @@ pub(crate) async fn stop(services: Arc<Services>) -> Result<()> {
 	// unload and explode.
 	services.stop().await;
 
+	// Check that Services and Database will drop as expected, The complex of Arc's
+	// used for various components can easily lead to references being held
+	// somewhere improperly; this can hang shutdowns.
+	debug!("Cleaning up...");
+	let db = Arc::downgrade(&services.db);
 	if let Err(services) = Arc::try_unwrap(services) {
 		debug_error!(
 			"{} dangling references to Services after shutdown",
@@ -85,7 +90,12 @@ pub(crate) async fn stop(services: Arc<Services>) -> Result<()> {
 		);
 	}
 
-	debug!("Cleaning up...");
+	// The db threadpool requires async join if we use tokio/spawn_blocking to
+	// manage the threads. Without async-drop we have to wait here; for symmetry
+	// with Services construction it can't be done in services.stop().
+	if let Some(db) = db.upgrade() {
+		db.db.shutdown_pool().await;
+	}
 
 	#[cfg(feature = "systemd")]
 	sd_notify::notify(true, &[sd_notify::NotifyState::Stopping]).expect("failed to notify systemd of stopping state");