fix(pageserver): add the walreceiver state to tenant timeline GET api endpoint (#5196)

Add a `walreceiver_state` field to `TimelineInfo` (response of `GET /v1/tenant/:tenant_id/timeline/:timeline_id`) and while doing that, refactor out a common `Timeline::walreceiver_state(..)`. No OpenAPI changes, because this is an internal debugging addition.

Fixes #3115.

Co-authored-by: Joonas Koivunen <joonas.koivunen@gmail.com>

libs/pageserver_api/src/models.rs

@@ -381,6 +381,8 @@ pub struct TimelineInfo {
     pub pg_version: u32,
 
     pub state: TimelineState,
+
+    pub walreceiver_status: String,
 }
 
 #[derive(Debug, Clone, Serialize)]

pageserver/src/http/routes.rs

@@ -285,6 +285,8 @@ async fn build_timeline_info_common(
     let state = timeline.current_state();
     let remote_consistent_lsn = timeline.get_remote_consistent_lsn().unwrap_or(Lsn(0));
 
+    let walreceiver_status = timeline.walreceiver_status();
+
     let info = TimelineInfo {
         tenant_id: timeline.tenant_id,
         timeline_id: timeline.timeline_id,
@@ -305,6 +307,8 @@ async fn build_timeline_info_common(
         pg_version: timeline.pg_version,
 
         state,
+
+        walreceiver_status,
     };
     Ok(info)
 }

pageserver/src/tenant/timeline.rs

@@ -585,15 +585,7 @@ impl Timeline {
             Err(e) => {
                 // don't count the time spent waiting for lock below, and also in walreceiver.status(), towards the wait_lsn_time_histo
                 drop(_timer);
-                let walreceiver_status = {
-                    match &*self.walreceiver.lock().unwrap() {
-                        None => "stopping or stopped".to_string(),
-                        Some(walreceiver) => match walreceiver.status() {
-                            Some(status) => status.to_human_readable_string(),
-                            None => "Not active".to_string(),
-                        },
-                    }
-                };
+                let walreceiver_status = self.walreceiver_status();
                 Err(anyhow::Error::new(e).context({
                     format!(
                         "Timed out while waiting for WAL record at LSN {} to arrive, last_record_lsn {} disk consistent LSN={}, WalReceiver status: {}",
@@ -607,6 +599,16 @@ impl Timeline {
         }
     }
 
+    pub(crate) fn walreceiver_status(&self) -> String {
+        match &*self.walreceiver.lock().unwrap() {
+            None => "stopping or stopped".to_string(),
+            Some(walreceiver) => match walreceiver.status() {
+                Some(status) => status.to_human_readable_string(),
+                None => "Not active".to_string(),
+            },
+        }
+    }
+
     /// Check that it is valid to request operations with that lsn.
     pub fn check_lsn_is_in_scope(
         &self,

pageserver/src/tenant/timeline/walreceiver.rs

@@ -135,7 +135,7 @@ impl WalReceiver {
         .await;
     }
 
-    pub(super) fn status(&self) -> Option<ConnectionManagerStatus> {
+    pub(crate) fn status(&self) -> Option<ConnectionManagerStatus> {
         self.manager_status.read().unwrap().clone()
     }
 }