Release job map lock before queue sends

A full state-machine queue caused controller RPC handlers to wait while
holding the global job map mutex. One stalled job could therefore prevent
messages from reaching every other job.

Clone the selected job's sender under the map lock, then release the lock
before waiting for channel capacity. This preserves per-job backpressure
without making it controller-wide.

Author: cmackenzie1

crates/arroyo-controller/src/lib.rs

@@ -651,22 +651,28 @@ impl ControllerServer {
     }
 
     async fn send_to_job_queue(&self, job_id: &str, msg: JobMessage) -> Result<(), Status> {
-        let mut jobs = self.job_state.lock().await;
-
-        if let Some(sm) = jobs.get_mut(job_id) {
-            if let Err(e) = sm.send(msg).await {
-                Err(Status::failed_precondition(format!(
-                    "Cannot handle message for {job_id}: {e}"
-                )))
-            } else {
-                Ok(())
-            }
-        } else {
-            warn!(message = "Received message for unknown job id", job_id);
-            Err(Status::failed_precondition(format!(
-                "No job with id {job_id}"
-            )))
-        }
+        // Keep per-job backpressure from holding the global job map lock.
+        let tx = {
+            let jobs = self.job_state.lock().await;
+            let Some(sm) = jobs.get(job_id) else {
+                warn!(message = "Received message for unknown job id", job_id);
+                return Err(Status::failed_precondition(format!(
+                    "No job with id {job_id}"
+                )));
+            };
+
+            sm.sender().ok_or_else(|| {
+                Status::failed_precondition(format!(
+                    "Cannot handle message for {job_id}: State machine is inactive"
+                ))
+            })?
+        };
+
+        tx.send(msg).await.map_err(|_| {
+            Status::failed_precondition(format!(
+                "Cannot handle message for {job_id}: State machine is inactive"
+            ))
+        })
     }
 
     fn start_updater(&self, guard: ShutdownGuard) {

crates/arroyo-controller/src/states/mod.rs

@@ -1207,6 +1207,10 @@ impl StateMachine {
         }
     }
 
+    pub(crate) fn sender(&self) -> Option<Sender<JobMessage>> {
+        self.tx.clone()
+    }
+
     pub fn done(&self) -> bool {
         if let Some(tx) = &self.tx {
             tx.is_closed()