Re: Improve WALRead() to suck data directly from WAL buffers when possible

On Tue, Feb 28, 2023 at 6:14 AM Nathan Bossart <nathandbossart@gmail.com> wrote:
>
> On Wed, Feb 08, 2023 at 08:00:00PM +0530, Bharath Rupireddy wrote:
> > +                     /*
> > +                      * We read some of the requested bytes. Continue to read remaining
> > +                      * bytes.
> > +                      */
> > +                     ptr += nread;
> > +                     nbytes -= nread;
> > +                     dst += nread;
> > +                     *read_bytes += nread;
>
> Why do we only read a page at a time in XLogReadFromBuffersGuts()?  What is
> preventing us from copying all the data we need in one go?

Note that most of the WALRead() callers request a single page of
XLOG_BLCKSZ bytes even if the server has less or more available WAL
pages. It's the streaming replication wal sender that can request less
than XLOG_BLCKSZ bytes and upto MAX_SEND_SIZE (16 * XLOG_BLCKSZ). And,
if we read, say, MAX_SEND_SIZE at once while holding
WALBufMappingLock, that might impact concurrent inserters (at least, I
can say it in theory) - one of the main intentions of this patch is
not to impact inserters much.

Therefore, I feel reading one WAL buffer page at a time, which works
for most of the cases, without impacting concurrent inserters much is
better - https://www.postgresql.org/message-id/CALj2ACWXHP6Ha1BfDB14txm%3DXP272wCbOV00mcPg9c6EXbnp5A%40mail.gmail.com.

--
Bharath Rupireddy
PostgreSQL Contributors Team
RDS Open Source Databases
Amazon Web Services: https://aws.amazon.com