Re: 回复: Why is XLOG_FPI_FOR_HINT always need backups? - Mailing list pgsql-hackers

From Kyotaro Horiguchi
Subject Re: 回复: Why is XLOG_FPI_FOR_HINT always need backups?
Date
Msg-id 20210707.161125.1190797247155264523.horikyota.ntt@gmail.com
Whole thread Raw
In response to 回复: Why is XLOG_FPI_FOR_HINT always need backups?  ("zwj" <757634191@qq.com>)
Responses Re: 回复: Why is XLOG_FPI_FOR_HINT always need backups?  (Fujii Masao <masao.fujii@oss.nttdata.com>)
List pgsql-hackers
Hello.

At Tue, 6 Jul 2021 20:42:23 +0800, "zwj" <757634191@qq.com> wrote in 
> But I wonder whether it is necessary or not while my file system can protect the blocks of database to be torn. And I
reada comment in function MarkBufferDirtyHint:
 
> 
> /*
>  * If we need to protect hint bit updates from torn writes, WAL-log a
>  * full page image of the page. This full page image is only necessary
>  * if the hint bit update is the first change to the page since the
>  * last checkpoint.
>  *
>  * We don't check full_page_writes here because that logic is included
>  * when we call XLogInsert() since the value changes dynamically.
>  */
> 
> However, the code tell me it has nothing to do with full_page_writes. I can't figure it out.

The doc of wal_log_hints says that "*even* for non-critical
modifications of so-called hint bits", which seems to me implies it is
following full_page_writes (and I think it is nonsense otherwise, as
you suspect).

XLogSaveBufferForHint sets REGBUF_FORCE_IMAGE since 2c03216d83116 when
the symbol was introduced. As my understanding XLogInsert did not have
an ability to enforce FPIs before the commit. The code comment above
is older than that commit. So it seems to me a thinko that
XLogSaveBufferForHint sets REGBUF_FORCE_IMAGE.

I think the attached fixes that thinko.

regards.

-- 
Kyotaro Horiguchi
NTT Open Source Software Center
From d0d3d52953eda6451c212f4994fdc21a8284f6c1 Mon Sep 17 00:00:00 2001
From: Kyotaro Horiguchi <horikyota.ntt@gmail.com>
Date: Wed, 7 Jul 2021 15:34:41 +0900
Subject: [PATCH] Make FPI_FOR_HINT follow standard FPI emitting policy

Commit 2c03216d831160bedd72d45f712601b6f7d03f1c changed
XLogSaveBufferForHint to enforce FPI but the caller didn't intend to
do so.  Restore the intended behavior that FPI_FOR_HINT follows
full_page_writes.
---
 src/backend/access/transam/xloginsert.c | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 3d2c9c3e8c..e596a0470a 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -287,8 +287,6 @@ XLogRegisterBlock(uint8 block_id, RelFileNode *rnode, ForkNumber forknum,
 {
     registered_buffer *regbuf;
 
-    /* This is currently only used to WAL-log a full-page image of a page */
-    Assert(flags & REGBUF_FORCE_IMAGE);
     Assert(begininsert_called);
 
     if (block_id >= max_registered_block_id)
@@ -995,7 +993,7 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 
     if (lsn <= RedoRecPtr)
     {
-        int            flags;
+        int            flags = 0;
         PGAlignedBlock copied_buffer;
         char       *origdata = (char *) BufferGetBlock(buffer);
         RelFileNode rnode;
@@ -1022,7 +1020,6 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 
         XLogBeginInsert();
 
-        flags = REGBUF_FORCE_IMAGE;
         if (buffer_std)
             flags |= REGBUF_STANDARD;
 
-- 
2.27.0


pgsql-hackers by date:

Previous
From: Tatsuo Ishii
Date:
Subject: Re: [HACKERS] WIP aPatch: Pgbench Serialization and deadlock errors
Next
From: Amit Langote
Date:
Subject: Re: ExecRTCheckPerms() and many prunable partitions