Re: wal stats questions

Kyotaro Horiguchi <horikyota.ntt@gmail.com>

From: Kyotaro Horiguchi <horikyota.ntt@gmail.com>
To: andres@anarazel.de
Cc: ikedamsh@oss.nttdata.com, pgsql-hackers@postgresql.org, masao.fujii@oss.nttdata.com
Date: 2021-03-29T02:09:00Z
Lists: pgsql-hackers
At Fri, 26 Mar 2021 10:07:45 -0700, Andres Freund <andres@anarazel.de> wrote in 
> Hi,
> 
> On 2021-03-25 16:37:10 +0900, Kyotaro Horiguchi wrote:
> > On the other hand, the counters are incremented in XLogInsertRecord()
> > and I think we don't want add instructions there.
> 
> I don't really buy this. Setting a boolean to true, in a cacheline
> you're already touching, isn't that much compared to all the other stuff
> in there. The branch to check if wal stats timing etc is enabled is much
> more expensive.  I think we should just set a boolean to true and leave
> it at that.

Hmm. Yes, I agree to you in that opinion.  I (remember I) was told not
to add even a cycle to the hot path as far as we can avoid when I
tried something like that.

So I'm happy to +1 for that if it is the consensus here, since it is
cleaner.

regards.

-- 
Kyotaro Horiguchi
NTT Open Source Software Center



Commits

  1. Fix issues in pg_stat_wal.

  2. Change data type of counters in BufferUsage and WalUsage from long to int64.