Horton

Re: LogwrtResult contended spinlock

Tom Lane <tgl@sss.pgh.pa.us>

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Alvaro Herrera <alvherre@alvh.no-ip.org>
Cc: Andres Freund <andres@anarazel.de>, pgsql-hackers@lists.postgresql.org, Jaime Casanova <jcasanov@systemguards.com.ec>
Date: 2022-03-22T13:59:16Z
Lists: pgsql-hackers 
Alvaro Herrera <alvherre@alvh.no-ip.org> writes:
> On 2022-Mar-21, Andres Freund wrote:
>> Are you aiming this for v15? Otherwise I'd like to move the entry to the next
>> CF. Marked as waiting-on-author.

> I'd like to get 0001 pushed to pg15, yes.  I'll let 0002 sit here for
> discussion, but I haven't seen any evidence that we need it.  If others
> vouch for it, I can push that one too, but I'd rather have it be a
> separate thing.

I looked briefly at 0001, and I've got to say that I disagree with
your decision to rearrange the representation of the local LogwrtResult
copy.  It clutters the patch tremendously and makes it hard to
understand what the actual functional change is.  Moreover, I'm
not entirely convinced that it's a notational improvement in the
first place.

Perhaps it'd help if you split 0001 into two steps, one to do the
mechanical change of the representation and then a second patch that
converts the shared variable to atomics.  Since you've moved around
the places that read the shared variable, that part is subtler than
one could wish and really needs to be studied on its own.

			regards, tom lane

Commits

  1. Remove bogus assertion in pg_atomic_monotonic_advance_u64

  2. Add XLogCtl->logInsertResult

  3. Operate XLogCtl->log{Write,Flush}Result with atomics

  4. Split XLogCtl->LogwrtResult into separate struct members

  5. Introduce atomic read/write functions with full barrier semantics.

  6. Reduce the number of GetFlushRecPtr() calls done by walsenders.

  7. Remove most volatile qualifiers from xlog.c