From ba4a96fdb1cc08098c57e5b4f75492c5ea30345b Mon Sep 17 00:00:00 2001
From: Arthur Petukhovsky <petuhovskiy@yandex.ru>
Date: Thu, 6 Apr 2023 20:57:06 +0300
Subject: [PATCH] Eagerly update wal_backup_lsn after each segment offload
 (#3976)

Otherwise it can lag a lot, preventing WAL segments cleanup. Also max
wal_backup_lsn on update, pulling it down is pointless.

Should help with https://github.com/neondatabase/neon/issues/3957, but
will not fix it completely.
---
 safekeeper/src/timeline.rs   |  3 ++-
 safekeeper/src/wal_backup.rs | 26 +++++++++++++-------------
 2 files changed, 15 insertions(+), 14 deletions(-)

diff --git a/safekeeper/src/timeline.rs b/safekeeper/src/timeline.rs
index 9dd8a63cf0..8097c863fa 100644
--- a/safekeeper/src/timeline.rs
+++ b/safekeeper/src/timeline.rs
@@ -674,7 +674,8 @@ impl Timeline {
             bail!(TimelineError::Cancelled(self.ttid));
         }
 
-        self.write_shared_state().sk.inmem.backup_lsn = backup_lsn;
+        let mut state = self.write_shared_state();
+        state.sk.inmem.backup_lsn = max(state.sk.inmem.backup_lsn, backup_lsn);
         // we should check whether to shut down offloader, but this will be done
         // soon by peer communication anyway.
         Ok(())
diff --git a/safekeeper/src/wal_backup.rs b/safekeeper/src/wal_backup.rs
index 798b9abaf3..163ac99be8 100644
--- a/safekeeper/src/wal_backup.rs
+++ b/safekeeper/src/wal_backup.rs
@@ -323,7 +323,8 @@ impl WalBackupTask {
             }
 
             match backup_lsn_range(
-                backup_lsn,
+                &self.timeline,
+                &mut backup_lsn,
                 commit_lsn,
                 self.wal_seg_size,
                 &self.timeline_dir,
@@ -331,13 +332,7 @@ impl WalBackupTask {
             )
             .await
             {
-                Ok(backup_lsn_result) => {
-                    backup_lsn = backup_lsn_result;
-                    let res = self.timeline.set_wal_backup_lsn(backup_lsn_result);
-                    if let Err(e) = res {
-                        error!("failed to set wal_backup_lsn: {}", e);
-                        return;
-                    }
+                Ok(()) => {
                     retry_attempt = 0;
                 }
                 Err(e) => {
@@ -354,20 +349,25 @@ impl WalBackupTask {
 }
 
 pub async fn backup_lsn_range(
-    start_lsn: Lsn,
+    timeline: &Arc<Timeline>,
+    backup_lsn: &mut Lsn,
     end_lsn: Lsn,
     wal_seg_size: usize,
     timeline_dir: &Path,
     workspace_dir: &Path,
-) -> Result<Lsn> {
-    let mut res = start_lsn;
+) -> Result<()> {
+    let start_lsn = *backup_lsn;
     let segments = get_segments(start_lsn, end_lsn, wal_seg_size);
     for s in &segments {
         backup_single_segment(s, timeline_dir, workspace_dir)
             .await
             .with_context(|| format!("offloading segno {}", s.seg_no))?;
 
-        res = s.end_lsn;
+        let new_backup_lsn = s.end_lsn;
+        timeline
+            .set_wal_backup_lsn(new_backup_lsn)
+            .context("setting wal_backup_lsn")?;
+        *backup_lsn = new_backup_lsn;
     }
     info!(
         "offloaded segnos {:?} up to {}, previous backup_lsn {}",
@@ -375,7 +375,7 @@ pub async fn backup_lsn_range(
         end_lsn,
         start_lsn,
     );
-    Ok(res)
+    Ok(())
 }
 
 async fn backup_single_segment(