feat: 410 Gone Handling

changelog.d/2355.feature.md

@@ -0,0 +1 @@
+Added support for servers returning 410 gone to indicate they are no longer serving a Matrix server (MSC4188). Contributed by @nex and @silversys

src/service/federation/execute.rs

@@ -1,4 +1,9 @@
-use std::{borrow::Cow, fmt::Debug, mem, time::Instant};
+use std::{
+	borrow::Cow,
+	fmt::Debug,
+	mem,
+	time::{Duration, Instant},
+};
 
 use conduwuit::{
 	Err, Error, Result, debug, debug_error, debug_info, debug_warn, err, metrics::KeyValue,
@@ -133,12 +138,20 @@ impl super::Service {
 
 		self.ensure_remote_is_healthy(dest)?;
 
+		/// How long to mark a server as unhealthy before attempting
+		/// re-resolution
+		const UNHEALTHY_TIMEOUT: Duration = Duration::from_hours(24 * 90);
+
 		let actual = self
 			.services
 			.client
 			.matrix_resolver
 			.resolve_server(dest.as_str())
-			.await?;
+			.await
+			.inspect_err(|_| {
+				self.hit_unhealthy(dest.to_owned(), Some(UNHEALTHY_TIMEOUT));
+				self.mark_destination_stale(dest);
+			})?;
 
 		let request = Request::try_from(request.try_into_http_request::<Vec<u8>>(
 			actual.base_url().as_str(),
@@ -210,7 +223,7 @@ impl super::Service {
 				if self.should_mark_stale(&wrapped) {
 					debug_info!("{dest} is unhealthy & stale due to a connect error");
 					self.mark_destination_stale(dest);
-					self.hit_unhealthy(dest.to_owned());
+					self.hit_unhealthy(dest.to_owned(), None);
 				}
 				let Error::Reqwest(unwrapped) = wrapped else {
 					unreachable!("wrapped reqwest error must unwrap to a reqwest error");

src/service/federation/mod.rs

@@ -92,7 +92,7 @@ impl Service {
 	/// retry count is incremented and next retry window is updated.
 	///
 	/// Does not update the marker if the backoff period is already in effect.
-	pub fn hit_unhealthy(&self, server_name: OwnedServerName) {
+	pub fn hit_unhealthy(&self, server_name: OwnedServerName, next_attempt: Option<Duration>) {
 		if server_name == self.services.server.name {
 			return;
 		}
@@ -109,7 +109,8 @@ impl Service {
 		}
 
 		*retries = retries.saturating_add(1);
-		let next_interval = self.services.server.config.sender_retry_backoff(*retries);
+		let next_interval = next_attempt
+			.unwrap_or_else(|| self.services.server.config.sender_retry_backoff(*retries));
 		*next_retry = millis_from_now(next_interval);
 		debug!(
 			"{} is (now) unhealthy ({} retries, blocked until: {})",
@@ -200,14 +201,14 @@ impl Service {
 		}
 
 		match error.status_code() {
-			// Some special servers account for this specifically
-			| StatusCode::MISDIRECTED_REQUEST
-			// Common error codes observed for misdirected requests
-			// | StatusCode::NOT_FOUND  This one can be encountered naturally
-			| StatusCode::METHOD_NOT_ALLOWED
-			| StatusCode::IM_A_TEAPOT => true,
-			_ => false,
-		}
+            // Some special servers account for this specifically
+            | StatusCode::MISDIRECTED_REQUEST
+            // Common error codes observed for misdirected requests
+            // | StatusCode::NOT_FOUND  This one can be encountered naturally
+            | StatusCode::METHOD_NOT_ALLOWED
+            | StatusCode::IM_A_TEAPOT => true,
+            _ => false,
+        }
 	}
 
 	/// Returns a clone of the internal remote health tracking map.

src/service/sending/sender.rs

@@ -149,10 +149,10 @@ impl Service {
 					{
 						debug!("{dest} is now unhealthy & stale due to a connect error: {e:?}");
 						self.services.federation.mark_destination_stale(dest);
-						self.services.federation.hit_unhealthy(dest.clone());
+						self.services.federation.hit_unhealthy(dest.clone(), None);
 					} else if e.status_code() != StatusCode::OK {
 						debug!("{dest} is now unhealthy due to error response: {e:?}");
-						self.services.federation.hit_unhealthy(dest.clone());
+						self.services.federation.hit_unhealthy(dest.clone(), None);
 					}
 				}
 				Self::handle_response_err(dest, statuses, &e);