Do not return Error in case of duplicate layer detetion

pageserver/src/tenant/layer_map.rs

@@ -51,11 +51,12 @@ use crate::keyspace::KeyPartitioning;
 use crate::repository::Key;
 use crate::tenant::storage_layer::InMemoryLayer;
 use crate::tenant::storage_layer::Layer;
-use anyhow::{bail, Result};
+use anyhow::Result;
 use std::collections::VecDeque;
 use std::ops::Range;
 use std::sync::Arc;
 use utils::lsn::Lsn;
+use tracing::*;
 
 use historic_layer_coverage::BufferedHistoricLayerCoverage;
 pub use historic_layer_coverage::Replacement;
@@ -276,17 +277,17 @@ where
     pub(self) fn insert_historic_noflush(&mut self, layer: Arc<L>) -> anyhow::Result<()> {
         let key = historic_layer_coverage::LayerKey::from(&*layer);
         if self.historic.contains(&key) {
-            bail!(
+            error!(
                 "Attempt to insert duplicate layer {} in layer map",
                 layer.short_id()
             );
-        }
-        self.historic.insert(key, Arc::clone(&layer));
-
-        if Self::is_l0(&layer) {
-            self.l0_delta_layers.push(layer);
-        }
+        } else {
+			self.historic.insert(key, Arc::clone(&layer));
 
+			if Self::is_l0(&layer) {
+				self.l0_delta_layers.push(layer);
+			}
+		}
         Ok(())
     }
 

pageserver/src/tenant/timeline.rs

@@ -3300,6 +3300,10 @@ impl Timeline {
 
         drop(all_keys_iter); // So that deltas_to_compact is no longer borrowed
 
+        fail_point!("compact-level0-phase1-finish", |_| {
+            Err(anyhow::anyhow!("failpoint compact-level0-phase1-finish").into())
+        });
+
         Ok(CompactLevel0Phase1Result {
             new_layers,
             deltas_to_compact,

test_runner/regress/test_duplicate_layers.py

@@ -0,0 +1,41 @@
+import pytest
+import time
+from fixtures.neon_fixtures import NeonEnvBuilder, PgBin
+
+
+# Test duplicate layer detection
+#
+# This test sets fail point at the end of first compaction phase:
+# after flushing new L1 layers but before deletion of L0 layes
+# It should cause generation of duplicate L1 layer by compaction after restart
+@pytest.mark.timeout(600)
+def test_duplicate_layers(neon_env_builder: NeonEnvBuilder, pg_bin: PgBin):
+    env = neon_env_builder.init_start()
+
+    # These warnings are expected, when the pageserver is restarted abruptly
+    env.pageserver.allowed_errors.append(".*found future image layer.*")
+    env.pageserver.allowed_errors.append(".*found future delta layer.*")
+    env.pageserver.allowed_errors.append(".*Attempt to insert duplicate layer.*")
+
+    pageserver_http = env.pageserver.http_client()
+
+    # Use aggressive compaction and checkpoint settings
+    tenant_id, _ = env.neon_cli.create_tenant(
+        conf={
+            "checkpoint_distance": f"{1024 ** 2}",
+            "compaction_target_size": f"{1024 ** 2}",
+            "compaction_period": "1 s",
+            "compaction_threshold": "3",
+        }
+    )
+    endpoint = env.endpoints.create_start("main", tenant_id=tenant_id)
+    connstr = endpoint.connstr(options="-csynchronous_commit=off")
+    pg_bin.run_capture(["pgbench", "-i", "-s10", connstr])
+
+    pageserver_http.configure_failpoints(("compact-level0-phase1-finish", "exit"))
+
+    with pytest.raises(Exception):
+        pg_bin.run_capture(["pgbench", "-P1", "-N", "-c5", "-T500", "-Mprepared", connstr])
+    env.pageserver.stop()
+    env.pageserver.start()
+    time.sleep(10)  # let compaction to be performed