Fixes

pageserver/src/branches.rs

@@ -242,9 +242,9 @@ fn bootstrap_timeline<R: Repository>(
     // Because we know it upfront avoid having an option or dummy zero value by passing it to create_empty_timeline.
     let timeline = repo.create_empty_timeline(tli, lsn)?;
 
-    let page_tline: DatadirTimeline<R> = DatadirTimeline::new(timeline);
+    let mut page_tline: DatadirTimeline<R> = DatadirTimeline::new(timeline);
 
-    import_datadir::import_timeline_from_postgres_datadir(&pgdata_path, &page_tline, lsn)?;
+    import_datadir::import_timeline_from_postgres_datadir(&pgdata_path, &mut page_tline, lsn)?;
     page_tline.tline.checkpoint(CheckpointConfig::Forced)?;
 
     println!(

pageserver/src/import_datadir.rs

@@ -30,7 +30,7 @@ use zenith_utils::lsn::Lsn;
 /// cluster was not shut down cleanly.
 pub fn import_timeline_from_postgres_datadir<R: Repository>(
     path: &Path,
-    tline: &DatadirTimeline<R>,
+    tline: &mut DatadirTimeline<R>,
     lsn: Lsn,
 ) -> Result<()> {
     let mut pg_control: Option<ControlFileData> = None;
@@ -335,7 +335,7 @@ fn import_slru_file<R: Repository>(
 /// 'startpoint' and 'endpoint' into the repository.
 fn import_wal<R: Repository>(
     walpath: &Path,
-    tline: &DatadirTimeline<R>,
+    tline: &mut DatadirTimeline<R>,
     startpoint: Lsn,
     endpoint: Lsn,
 ) -> Result<()> {

pageserver/src/layered_repository.rs

@@ -1028,10 +1028,11 @@ impl LayeredTimeline {
                     if prev_lsn <= cont_lsn {
                         // Didn't make any progress in last iteration. Error out to avoid
                         // getting stuck in the loop.
-                        bail!("could not find layer with more data for key {} at LSN {}, request LSN {}",
+                        bail!("could not find layer with more data for key {} at LSN {}, request LSN {}, ancestor {}",
                           key,
                           Lsn(cont_lsn.0 - 1),
-                          request_lsn)
+                              request_lsn,
+                        timeline.ancestor_lsn)
                     }
                     prev_lsn = cont_lsn;
                 }
@@ -1047,7 +1048,7 @@ impl LayeredTimeline {
 
             // Recurse into ancestor if needed
             if Lsn(cont_lsn.0 - 1) <= timeline.ancestor_lsn {
-                //info!("going into ancestor {}", timeline.ancestor_lsn);
+                info!("going into ancestor {}, cont_lsn is {}", timeline.ancestor_lsn, cont_lsn);
                 let ancestor = timeline.get_ancestor_timeline()?;
                 timeline_owned = ancestor;
                 timeline = &*timeline_owned;
@@ -1060,7 +1061,7 @@ impl LayeredTimeline {
             // Check the open and frozen in-memory layers first
             if let Some(open_layer) = &layers.open_layer {
                 let start_lsn = open_layer.get_lsn_range().start;
-                if cont_lsn >= start_lsn {
+                if cont_lsn > start_lsn {
                     //info!("CHECKING for {} at {} on open layer {}", reconstruct_state.key, reconstruct_state.lsn, open_layer.filename().display());
                     result = open_layer.get_value_reconstruct_data(key, open_layer.get_lsn_range().start..cont_lsn, reconstruct_state)?;
                     cont_lsn = open_layer.get_lsn_range().start;
@@ -1069,7 +1070,7 @@ impl LayeredTimeline {
             }
             if let Some(frozen_layer) = &layers.frozen_layer {
                 let start_lsn = frozen_layer.get_lsn_range().start;
-                if cont_lsn >= start_lsn {
+                if cont_lsn > start_lsn {
                     //info!("CHECKING for {} at {} on frozen layer {}", reconstruct_state.key, reconstruct_state.lsn, frozen_layer.filename().display());
                     result = frozen_layer.get_value_reconstruct_data(key, frozen_layer.get_lsn_range().start..cont_lsn, reconstruct_state)?;
                     cont_lsn = frozen_layer.get_lsn_range().start;
@@ -1594,6 +1595,9 @@ impl LayeredTimeline {
     fn gc(&self) -> Result<GcResult> {
         let now = Instant::now();
         let mut result: GcResult = Default::default();
+        if true {
+            return Ok(result);
+        }
         let disk_consistent_lsn = self.get_disk_consistent_lsn();
         let _checkpoint_cs = self.checkpoint_cs.lock().unwrap();
 

pageserver/src/pgdatadir_mapping.rs

@@ -22,6 +22,7 @@ use std::sync::{Arc, RwLockReadGuard};
 use tracing::{debug, info, warn};
 use zenith_utils::bin_ser::BeSer;
 use zenith_utils::lsn::{Lsn, RecordLsn};
+use zenith_utils::lsn::AtomicLsn;
 
 /// Block number within a relation or SRU. This matches PostgreSQL's BlockNumber type.
 pub type BlockNumber = u32;
@@ -31,7 +32,7 @@ where
     R: Repository,
 {
     pub tline: Arc<R::Timeline>,
-    pub last_partitioning: Option<Lsn>,
+    pub last_partitioning: AtomicLsn,
 }
 
 #[derive(Debug, Serialize, Deserialize)]
@@ -71,7 +72,7 @@ impl<R: Repository> DatadirTimeline<R> {
     pub fn new(tline: Arc<R::Timeline>) -> Self {
         DatadirTimeline {
             tline,
-            last_partitioning: None,
+            last_partitioning: AtomicLsn::new(0),
         }
     }
 
@@ -692,7 +693,7 @@ impl<'a, R: Repository> DatadirTimelineWriter<'a, R> {
     pub fn finish(self) -> Result<()> {
         let writer = self.tline.tline.writer();
 
-        let last_partitioning = self.last_partitioning.unwrap_or(Lsn(0));
+        let last_partitioning = self.last_partitioning.load();
 
         for (key, value) in self.pending_updates {
             writer.put(key, self.lsn, value)?;
@@ -707,6 +708,7 @@ impl<'a, R: Repository> DatadirTimelineWriter<'a, R> {
             let mut partitioning = self.tline.collect_keyspace(self.lsn)?;
             partitioning.repartition(TARGET_FILE_SIZE_BYTES);
             self.tline.tline.hint_partitioning(partitioning)?;
+            self.tline.last_partitioning.store(self.lsn);
         }
 
         Ok(())

pageserver/src/walreceiver.rs

@@ -242,7 +242,7 @@ fn walreceiver_main(
                 let startlsn = Lsn::from(xlog_data.wal_start());
                 let endlsn = startlsn + data.len() as u64;
 
-                trace!("received XLogData between {} and {}", startlsn, endlsn);
+                info!("received XLogData between {} and {}", startlsn, endlsn);
 
                 waldecoder.feed_bytes(data);