Do not emit WAL for unlogged BRIN indexes REL_16_STABLE github/REL_16_STABLE
authorHeikki Linnakangas <[email protected]>
Thu, 18 Dec 2025 13:08:48 +0000 (15:08 +0200)
committerHeikki Linnakangas <[email protected]>
Thu, 18 Dec 2025 13:09:21 +0000 (15:09 +0200)
Operations on unlogged relations should not be WAL-logged. The
brin_initialize_empty_new_buffer() function didn't get the memo.

The function is only called when a concurrent update to a brin page
uses up space that we're just about to insert to, which makes it
pretty hard to hit. If you do manage to hit it, a full-page WAL record
is erroneously emitted for the unlogged index. If you then crash,
crash recovery will fail on that record with an error like this:

    FATAL:  could not create file "base/5/32819": File exists

Author: Kirill Reshke <[email protected]>
Discussion: https://www.postgresql.org/message-id/CALdSSPhpZXVFnWjwEBNcySx_vXtXHwB2g99gE6rK0uRJm-3GgQ@mail.gmail.com
Backpatch-through: 14

src/backend/access/brin/brin_pageops.c

index b578d2595451fa4cdbf45e1e3ba17ce714a49a79..27eb2f62ad6d2a1a70e203b19c30568a470bb35a 100644 (file)
@@ -894,7 +894,11 @@ brin_initialize_empty_new_buffer(Relation idxrel, Buffer buffer)
    page = BufferGetPage(buffer);
    brin_page_init(page, BRIN_PAGETYPE_REGULAR);
    MarkBufferDirty(buffer);
-   log_newpage_buffer(buffer, true);
+
+   /* XLOG stuff */
+   if (RelationNeedsWAL(idxrel))
+       log_newpage_buffer(buffer, true);
+
    END_CRIT_SECTION();
 
    /*