Fix unlogged sequence corruption after standby promotion

Previously, if an unlogged sequence was created on the primary and
replicated to a standby, reading the sequence after promoting the
standby (for example, with nextval()) could trigger the following
assertion failure:

    TRAP: failed Assert("((const PageHeaderData *) page)->pd_special >= SizeOfPageHeaderData")

In non-assert builds, the same operation could instead fail with an
error such as:

    ERROR:  bad magic number in sequence

The problem was that seq_redo() updated the init fork page in shared
buffers but did not flush it to disk. During promotion,
ResetUnloggedRelations() recreates the main fork of unlogged
relations by copying the init fork from disk, bypassing shared
buffers. As a result, the main fork could be recreated from a stale
init fork instead of the WAL-replayed page.

Fix this by introducing a helper to flush init fork buffers
immediately, and make seq_redo() use it. As a result, the main fork
of an unlogged sequence is recreated from the up-to-date init fork on
disk, allowing the unlogged sequence to be read successfully after
standby promotion.

Backpatch to v15, where unlogged sequences were introduced.

Author: Fujii Masao <masao.fujii@gmail.com>
Reviewed-by: vignesh C <vignesh21@gmail.com>
Discussion: https://postgr.es/m/CAHGQGwH1Ssze3XM6wjoTjSLVOR041c6xP+vsdLP951=w8oG8bA@mail.gmail.com
Backpatch-through: 15

diff --git a/src/backend/access/hash/hash_xlog.c b/src/backend/access/hash/hash_xlog.c
index 8d97067fe54036906c2a5c8a4ee474268ba71ea0..d4cb6246b4811a90e6b263f7f6587181cbc7ce22 100644 (file)
--- a/src/backend/access/hash/hash_xlog.c
+++ b/src/backend/access/hash/hash_xlog.c
@@ -29,7 +29,6 @@ hash_xlog_init_meta_page(XLogReaderState *record)
        XLogRecPtr      lsn = record->EndRecPtr;
        Page            page;
        Buffer          metabuf;
-       ForkNumber      forknum;
 
        xl_hash_init_meta_page *xlrec = (xl_hash_init_meta_page *) XLogRecGetData(record);
 
@@ -41,16 +40,7 @@ hash_xlog_init_meta_page(XLogReaderState *record)
        page = (Page) BufferGetPage(metabuf);
        PageSetLSN(page, lsn);
        MarkBufferDirty(metabuf);
-
-       /*
-        * Force the on-disk state of init forks to always be in sync with the
-        * state in shared buffers.  See XLogReadBufferForRedoExtended.  We need
-        * special handling for init forks as create index operations don't log a
-        * full page image of the metapage.
-        */
-       XLogRecGetBlockTag(record, 0, NULL, &forknum, NULL);
-       if (forknum == INIT_FORKNUM)
-               FlushOneBuffer(metabuf);
+       XLogFlushBufferForRedoIfInit(record, 0, metabuf);
 
        /* all done */
        UnlockReleaseBuffer(metabuf);
@@ -68,7 +58,6 @@ hash_xlog_init_bitmap_page(XLogReaderState *record)
        Page            page;
        HashMetaPage metap;
        uint32          num_buckets;
-       ForkNumber      forknum;
 
        xl_hash_init_bitmap_page *xlrec = (xl_hash_init_bitmap_page *) XLogRecGetData(record);
 
@@ -79,16 +68,7 @@ hash_xlog_init_bitmap_page(XLogReaderState *record)
        _hash_initbitmapbuffer(bitmapbuf, xlrec->bmsize, true);
        PageSetLSN(BufferGetPage(bitmapbuf), lsn);
        MarkBufferDirty(bitmapbuf);
-
-       /*
-        * Force the on-disk state of init forks to always be in sync with the
-        * state in shared buffers.  See XLogReadBufferForRedoExtended.  We need
-        * special handling for init forks as create index operations don't log a
-        * full page image of the metapage.
-        */
-       XLogRecGetBlockTag(record, 0, NULL, &forknum, NULL);
-       if (forknum == INIT_FORKNUM)
-               FlushOneBuffer(bitmapbuf);
+       XLogFlushBufferForRedoIfInit(record, 0, bitmapbuf);
        UnlockReleaseBuffer(bitmapbuf);
 
        /* add the new bitmap page to the metapage's list of bitmaps */
@@ -109,10 +89,7 @@ hash_xlog_init_bitmap_page(XLogReaderState *record)
 
                PageSetLSN(page, lsn);
                MarkBufferDirty(metabuf);
-
-               XLogRecGetBlockTag(record, 1, NULL, &forknum, NULL);
-               if (forknum == INIT_FORKNUM)
-                       FlushOneBuffer(metabuf);
+               XLogFlushBufferForRedoIfInit(record, 1, metabuf);
        }
        if (BufferIsValid(metabuf))
                UnlockReleaseBuffer(metabuf);

diff --git a/src/backend/access/transam/xlogutils.c b/src/backend/access/transam/xlogutils.c
index db5a314edf86957d59f867c3be3e7a14debb2ae6..0d67f256afedff3d120619dce8e0d47ae871c0cf 100644 (file)
--- a/src/backend/access/transam/xlogutils.c
+++ b/src/backend/access/transam/xlogutils.c
@@ -321,6 +321,28 @@ XLogInitBufferForRedo(XLogReaderState *record, uint8 block_id)
        return buf;
 }
 
+/*
+ * If a redo routine modified an init fork, flush the buffer immediately.
+ *
+ * At the end of crash recovery the init forks of unlogged relations are
+ * copied to the main fork directly from disk, without going through shared
+ * buffers. Therefore, redo routines that update init forks without
+ * restoring a full-page image must call this after setting the page LSN and
+ * marking the buffer dirty.
+ */
+void
+XLogFlushBufferForRedoIfInit(XLogReaderState *record, uint8 block_id,
+                                                        Buffer buffer)
+{
+       ForkNumber      forknum;
+
+       Assert(BufferIsValid(buffer));
+
+       XLogRecGetBlockTag(record, block_id, NULL, &forknum, NULL);
+       if (forknum == INIT_FORKNUM)
+               FlushOneBuffer(buffer);
+}
+
 /*
  * XLogReadBufferForRedoExtended
  *             Like XLogReadBufferForRedo, but with extra options.
@@ -398,7 +420,9 @@ XLogReadBufferForRedoExtended(XLogReaderState *record,
                 * At the end of crash recovery the init forks of unlogged relations
                 * are copied, without going through shared buffers. So we need to
                 * force the on-disk state of init forks to always be in sync with the
-                * state in shared buffers.
+                * state in shared buffers. Use XLogFlushBufferForRedoIfInit() for
+                * redo routines that dirty init-fork buffers without restoring a
+                * full-page image.
                 */
                if (forknum == INIT_FORKNUM)
                        FlushOneBuffer(*buf);

diff --git a/src/backend/commands/sequence.c b/src/backend/commands/sequence.c
index a79ef0651a9e0cb4534c31be939cbd60d51afab6..c1ad656397a8df3c5ec8e6625358fca85bfc07e6 100644 (file)
--- a/src/backend/commands/sequence.c
+++ b/src/backend/commands/sequence.c
@@ -1933,6 +1933,7 @@ seq_redo(XLogReaderState *record)
 
        memcpy(page, localpage, BufferGetPageSize(buffer));
        MarkBufferDirty(buffer);
+       XLogFlushBufferForRedoIfInit(record, 0, buffer);
        UnlockReleaseBuffer(buffer);
 
        pfree(localpage);

diff --git a/src/include/access/xlogutils.h b/src/include/access/xlogutils.h
index a1870d8e5aa170805f925b315c61ed3de3b76014..7639bd523e1269bc4fdaf964c10600516d036089 100644 (file)
--- a/src/include/access/xlogutils.h
+++ b/src/include/access/xlogutils.h
@@ -87,6 +87,8 @@ typedef struct ReadLocalXLogPageNoWaitPrivate
 extern XLogRedoAction XLogReadBufferForRedo(XLogReaderState *record,
                                                                                        uint8 block_id, Buffer *buf);
 extern Buffer XLogInitBufferForRedo(XLogReaderState *record, uint8 block_id);
+extern void XLogFlushBufferForRedoIfInit(XLogReaderState *record,
+                                                                                uint8 block_id, Buffer buffer);
 extern XLogRedoAction XLogReadBufferForRedoExtended(XLogReaderState *record,
                                                                                                        uint8 block_id,
                                                                                                        ReadBufferMode mode, bool get_cleanup_lock,

diff --git a/src/test/recovery/meson.build b/src/test/recovery/meson.build
index 5245fdde43cd7afed0563ee00ef7dee05f4e96ad..38e1e43e0415487b485387cb0643f0b33c6fe0a6 100644 (file)
--- a/src/test/recovery/meson.build
+++ b/src/test/recovery/meson.build
@@ -58,6 +58,7 @@ tests += {
       't/047_checkpoint_physical_slot.pl',
       't/048_vacuum_horizon_floor.pl',
       't/053_standby_login_event_trigger.pl',
+      't/054_unlogged_sequence_promotion.pl',
     ],
   },
 }

diff --git a/src/test/recovery/t/054_unlogged_sequence_promotion.pl b/src/test/recovery/t/054_unlogged_sequence_promotion.pl
new file mode 100644 (file)
index 0000000..96d1e4b
--- /dev/null
+++ b/src/test/recovery/t/054_unlogged_sequence_promotion.pl
@@ -0,0 +1,34 @@
+# Copyright (c) 2026, PostgreSQL Global Development Group
+
+# Test that unlogged sequences created on a primary can be read after
+# promotion of a standby that replayed their init fork.
+
+use strict;
+use warnings FATAL => 'all';
+use PostgreSQL::Test::Cluster;
+use PostgreSQL::Test::Utils;
+use Test::More;
+
+my $node_primary = PostgreSQL::Test::Cluster->new('primary');
+$node_primary->init(allows_streaming => 1);
+$node_primary->start;
+
+my $backup_name = 'my_backup';
+$node_primary->backup($backup_name);
+
+my $node_standby = PostgreSQL::Test::Cluster->new('standby');
+$node_standby->init_from_backup($node_primary, $backup_name,
+       has_streaming => 1);
+$node_standby->start;
+
+# Create the unlogged sequence after the standby has started, so its init fork
+# is generated by WAL replay on the standby.
+$node_primary->safe_psql('postgres', "CREATE UNLOGGED SEQUENCE ulseq");
+$node_primary->wait_for_replay_catchup($node_standby);
+
+$node_standby->promote;
+
+is($node_standby->safe_psql('postgres', "SELECT nextval('ulseq')"),
+       1, 'unlogged sequence can be read after standby promotion');
+
+done_testing();