NEON: Finish Zenith->Neon rename (#12566)

Even though we're now part of Databricks, let's at least make this part
consistent.

## Summary of changes

- PG14: https://github.com/neondatabase/postgres/pull/669
- PG15: https://github.com/neondatabase/postgres/pull/670
- PG16: https://github.com/neondatabase/postgres/pull/671
- PG17: https://github.com/neondatabase/postgres/pull/672

---------

Co-authored-by: Arpad Müller <arpad-m@users.noreply.github.com>

compute_tools/src/compute.rs

@@ -1040,6 +1040,8 @@ impl ComputeNode {
             PageserverProtocol::Grpc => self.try_get_basebackup_grpc(spec, lsn)?,
         };
 
+        self.fix_zenith_signal_neon_signal()?;
+
         let mut state = self.state.lock().unwrap();
         state.metrics.pageserver_connect_micros =
             connected.duration_since(started).as_micros() as u64;
@@ -1049,6 +1051,27 @@ impl ComputeNode {
         Ok(())
     }
 
+    /// Move the Zenith signal file to Neon signal file location.
+    /// This makes Compute compatible with older PageServers that don't yet
+    /// know about the Zenith->Neon rename.
+    fn fix_zenith_signal_neon_signal(&self) -> Result<()> {
+        let datadir = Path::new(&self.params.pgdata);
+
+        let neonsig = datadir.join("neon.signal");
+
+        if neonsig.is_file() {
+            return Ok(());
+        }
+
+        let zenithsig = datadir.join("zenith.signal");
+
+        if zenithsig.is_file() {
+            fs::copy(zenithsig, neonsig)?;
+        }
+
+        Ok(())
+    }
+
     /// Fetches a basebackup via gRPC. The connstring must use grpc://. Returns the timestamp when
     /// the connection was established, and the (compressed) size of the basebackup.
     fn try_get_basebackup_grpc(&self, spec: &ParsedSpec, lsn: Lsn) -> Result<(Instant, usize)> {

control_plane/src/endpoint.rs

@@ -32,7 +32,8 @@
 //!     config.json                 - passed to `compute_ctl`
 //!     pgdata/
 //!         postgresql.conf       - copy of postgresql.conf created by `compute_ctl`
-//!         zenith.signal
+//!         neon.signal
+//!         zenith.signal         - copy of neon.signal, for backward compatibility
 //!         <other PostgreSQL files>
 //! ```
 //!

docs/core_changes.md

@@ -129,9 +129,10 @@ segment to bootstrap the WAL writing, but it doesn't contain the checkpoint reco
 changes in xlog.c, to allow starting the compute node without reading the last checkpoint record
 from WAL.
 
-This includes code to read the `zenith.signal` file, which tells the startup code the LSN to start
-at. When the `zenith.signal` file is present, the startup uses that LSN instead of the last
-checkpoint's LSN. The system is known to be consistent at that LSN, without any WAL redo.
+This includes code to read the `neon.signal` (also `zenith.signal`) file, which tells the startup 
+code the LSN to start at. When the `neon.signal` file is present, the startup uses that LSN
+instead of the last checkpoint's LSN. The system is known to be consistent at that LSN, without 
+any WAL redo.
 
 
 ### How to get rid of the patch

pageserver/src/basebackup.rs

@@ -114,7 +114,7 @@ where
     // Compute postgres doesn't have any previous WAL files, but the first
     // record that it's going to write needs to include the LSN of the
     // previous record (xl_prev). We include prev_record_lsn in the
-    // "zenith.signal" file, so that postgres can read it during startup.
+    // "neon.signal" file, so that postgres can read it during startup.
     //
     // We don't keep full history of record boundaries in the page server,
     // however, only the predecessor of the latest record on each
@@ -751,34 +751,39 @@ where
 
     //
     // Add generated pg_control file and bootstrap WAL segment.
-    // Also send zenith.signal file with extra bootstrap data.
+    // Also send neon.signal and zenith.signal file with extra bootstrap data.
     //
     async fn add_pgcontrol_file(
         &mut self,
         pg_control_bytes: Bytes,
         system_identifier: u64,
     ) -> Result<(), BasebackupError> {
-        // add zenith.signal file
-        let mut zenith_signal = String::new();
+        // add neon.signal file
+        let mut neon_signal = String::new();
         if self.prev_record_lsn == Lsn(0) {
             if self.timeline.is_ancestor_lsn(self.lsn) {
-                write!(zenith_signal, "PREV LSN: none")
+                write!(neon_signal, "PREV LSN: none")
                     .map_err(|e| BasebackupError::Server(e.into()))?;
             } else {
-                write!(zenith_signal, "PREV LSN: invalid")
+                write!(neon_signal, "PREV LSN: invalid")
                     .map_err(|e| BasebackupError::Server(e.into()))?;
             }
         } else {
-            write!(zenith_signal, "PREV LSN: {}", self.prev_record_lsn)
+            write!(neon_signal, "PREV LSN: {}", self.prev_record_lsn)
                 .map_err(|e| BasebackupError::Server(e.into()))?;
         }
+
+        // TODO: Remove zenith.signal once all historical computes have been replaced
+        // ... and thus support the neon.signal file.
+        for signalfilename in ["neon.signal", "zenith.signal"] {
             self.ar
                 .append(
-                &new_tar_header("zenith.signal", zenith_signal.len() as u64)?,
-                zenith_signal.as_bytes(),
+                    &new_tar_header(signalfilename, neon_signal.len() as u64)?,
+                    neon_signal.as_bytes(),
                 )
                 .await
-            .map_err(|e| BasebackupError::Client(e, "add_pgcontrol_file,zenith.signal"))?;
+                .map_err(|e| BasebackupError::Client(e, "add_pgcontrol_file,neon.signal"))?;
+        }
 
         //send pg_control
         let header = new_tar_header("global/pg_control", pg_control_bytes.len() as u64)?;

pageserver/src/import_datadir.rs

@@ -610,13 +610,13 @@ async fn import_file(
         debug!("imported twophase file");
     } else if file_path.starts_with("pg_wal") {
         debug!("found wal file in base section. ignore it");
-    } else if file_path.starts_with("zenith.signal") {
+    } else if file_path.starts_with("zenith.signal") || file_path.starts_with("neon.signal") {
         // Parse zenith signal file to set correct previous LSN
         let bytes = read_all_bytes(reader).await?;
-        // zenith.signal format is "PREV LSN: prev_lsn"
+        // neon.signal format is "PREV LSN: prev_lsn"
         // TODO write serialization and deserialization in the same place.
-        let zenith_signal = std::str::from_utf8(&bytes)?.trim();
-        let prev_lsn = match zenith_signal {
+        let neon_signal = std::str::from_utf8(&bytes)?.trim();
+        let prev_lsn = match neon_signal {
             "PREV LSN: none" => Lsn(0),
             "PREV LSN: invalid" => Lsn(0),
             other => {
@@ -624,17 +624,17 @@ async fn import_file(
                 split[1]
                     .trim()
                     .parse::<Lsn>()
-                    .context("can't parse zenith.signal")?
+                    .context("can't parse neon.signal")?
             }
         };
 
-        // zenith.signal is not necessarily the last file, that we handle
+        // neon.signal is not necessarily the last file, that we handle
         // but it is ok to call `finish_write()`, because final `modification.commit()`
         // will update lsn once more to the final one.
         let writer = modification.tline.writer().await;
         writer.finish_write(prev_lsn);
 
-        debug!("imported zenith signal {}", prev_lsn);
+        debug!("imported neon signal {}", prev_lsn);
     } else if file_path.starts_with("pg_tblspc") {
         // TODO Backups exported from neon won't have pg_tblspc, but we will need
         // this to import arbitrary postgres databases.

pgxn/neon_test_utils/neontest.c

@@ -236,13 +236,13 @@ clear_buffer_cache(PG_FUNCTION_ARGS)
 	bool		save_neon_test_evict;
 
 	/*
-	 * Temporarily set the zenith_test_evict GUC, so that when we pin and
+	 * Temporarily set the neon_test_evict GUC, so that when we pin and
 	 * unpin a buffer, the buffer is evicted. We use that hack to evict all
 	 * buffers, as there is no explicit "evict this buffer" function in the
 	 * buffer manager.
 	 */
-	save_neon_test_evict = zenith_test_evict;
-	zenith_test_evict = true;
+	save_neon_test_evict = neon_test_evict;
+	neon_test_evict = true;
 	PG_TRY();
 	{
 		/* Scan through all the buffers */
@@ -273,7 +273,7 @@ clear_buffer_cache(PG_FUNCTION_ARGS)
 
 			/*
 			 * Pin the buffer, and release it again. Because we have
-			 * zenith_test_evict==true, this will evict the page from the
+			 * neon_test_evict==true, this will evict the page from the
 			 * buffer cache if no one else is holding a pin on it.
 			 */
 			if (isvalid)
@@ -286,7 +286,7 @@ clear_buffer_cache(PG_FUNCTION_ARGS)
 	PG_FINALLY();
 	{
 		/* restore the GUC */
-		zenith_test_evict = save_neon_test_evict;
+		neon_test_evict = save_neon_test_evict;
 	}
 	PG_END_TRY();
 

pgxn/typedefs.list

@@ -2953,17 +2953,17 @@ XmlTableBuilderData
 YYLTYPE
 YYSTYPE
 YY_BUFFER_STATE
-ZenithErrorResponse
-ZenithExistsRequest
-ZenithExistsResponse
-ZenithGetPageRequest
-ZenithGetPageResponse
-ZenithMessage
-ZenithMessageTag
-ZenithNblocksRequest
-ZenithNblocksResponse
-ZenithRequest
-ZenithResponse
+NeonErrorResponse
+NeonExistsRequest
+NeonExistsResponse
+NeonGetPageRequest
+NeonGetPageResponse
+NeonMessage
+NeonMessageTag
+NeonNblocksRequest
+NeonNblocksResponse
+NeonRequest
+NeonResponse
 _SPI_connection
 _SPI_plan
 __AssignProcessToJobObject

test_runner/fixtures/neon_fixtures.py

@@ -5409,6 +5409,7 @@ SKIP_FILES = frozenset(
     (
         "pg_internal.init",
         "pg.log",
+        "neon.signal",
         "zenith.signal",
         "pg_hba.conf",
         "postgresql.conf",

test_runner/regress/test_timeline_detach_ancestor.py

@@ -209,9 +209,9 @@ def test_ancestor_detach_branched_from(
     client.timeline_delete(env.initial_tenant, env.initial_timeline)
     wait_timeline_detail_404(client, env.initial_tenant, env.initial_timeline)
 
-    # because we do the fullbackup from ancestor at the branch_lsn, the zenith.signal is always different
-    # as there is always "PREV_LSN: invalid" for "before"
-    skip_files = {"zenith.signal"}
+    # because we do the fullbackup from ancestor at the branch_lsn, the neon.signal and/or zenith.signal is always
+    # different as there is always "PREV_LSN: invalid" for "before"
+    skip_files = {"zenith.signal", "neon.signal"}
 
     assert_pageserver_backups_equal(fullbackup_before, fullbackup_after, skip_files)
 
@@ -767,7 +767,7 @@ def test_compaction_induced_by_detaches_in_history(
         env.pageserver, env.initial_tenant, branch_timeline_id, branch_lsn, fullbackup_after
     )
 
-    # we don't need to skip any files, because zenith.signal will be identical
+    # we don't need to skip any files, because neon.signal will be identical
     assert_pageserver_backups_equal(fullbackup_before, fullbackup_after, set())
 
 

vendor/revisions.json

@@ -1,18 +1,18 @@
 {
   "v17": [
     "17.5",
-    "db424d42d748f8ad91ac00e28db2c7f2efa42f7f"
+    "353c725b0c76cc82b15af21d8360d03391dc6814"
   ],
   "v16": [
     "16.9",
-    "7a4c0eacaeb9b97416542fa19103061c166460b1"
+    "e08c8d5f1576ca0487d14d154510499c5f12adfb"
   ],
   "v15": [
     "15.13",
-    "8c3249f36c7df6ac0efb8ee9f1baf4aa1b83e5c9"
+    "afd46987f3da50c9146a8aa59380052df0862c06"
   ],
   "v14": [
     "14.18",
-    "9085654ee8022d5cc4ca719380a1dc53e5e3246f"
+    "8ce1f52303aec29e098309347b57c01a1962e221"
   ]
 }