Re: SLRU statistics

From: Tomas Vondra <tomas(dot)vondra(at)2ndquadrant(dot)com>
To: Alvaro Herrera <alvherre(at)2ndquadrant(dot)com>
Cc: Masahiko Sawada <masahiko(dot)sawada(at)2ndquadrant(dot)com>, "tsunakawa(dot)takay(at)fujitsu(dot)com" <tsunakawa(dot)takay(at)fujitsu(dot)com>, "pgsql-hackers(at)postgresql(dot)org" <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: SLRU statistics
Date: 2020-02-29 12:24:41
Message-ID: 20200229122441.ow7gnnqgesmojyc3@development
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-hackers

On Fri, Feb 28, 2020 at 08:19:18PM -0300, Alvaro Herrera wrote:
>On 2020-Jan-21, Tomas Vondra wrote:
>> On Tue, Jan 21, 2020 at 05:09:33PM +0900, Masahiko Sawada wrote:
>> > I've not tested the performance impact but perhaps we might want to
>> > disable these counter by default and controlled by a GUC. And similar
>> > to buffer statistics it might be better to inline
>> > pgstat_count_slru_page_xxx function for better performance.
>> Hmmm, yeah. Inlining seems like a good idea, and maybe we should have
>> something like track_slru GUC.
>I disagree with adding a GUC. If a performance impact can be measured
>let's turn the functions to static inline, as already proposed. My
>guess is that pgstat_count_slru_page_hit() is the only candidate for
>that; all the other paths involve I/O or lock acquisition or even WAL
>generation, so the impact won't be measurable anyhow. We removed
>track-enabling GUCs years ago.

Did we actually remove track-enabling GUCs? I think we still have

- track_activities
- track_counts
- track_io_timing
- track_functions

But maybe I'm missing something?

That being said, I'm not sure we need to add a GUC. I'll do some
measurements and we'll see. Maybe the statis inline will me enough.

>BTW, this comment:
> /* update the stats counter of pages found in shared buffers */
>is not strictly true, because we don't use what we normally call "shared
>buffers" for SLRUs.

Oh, right. Will fix.

>Patch applies cleanly. I suggest to move the page_miss() call until
>after SlruRecentlyUsed(), for consistency with the other case.


>I find SlruType pretty odd, and the accompanying "if" list in
>pg_stat_get_slru() correspondingly so. Would it be possible to have
>each SLRU enumerate itself somehow? Maybe add the name in SlruCtlData
>and query that, somehow. (I don't think we have an array of SlruCtlData
>anywhere though, so this might be a useless idea.)

Well, maybe. We could have a system to register SLRUs dynamically, but
the trick here is that by having a fixed predefined number of SLRUs
simplifies serialization in pgstat.c and so on. I don't think the "if"
branches in pg_stat_get_slru() are particularly ugly, but maybe we could
replace the enum with a registry of structs, something like rmgrlist.h.
It seems like an overkill to me, though.


Tomas Vondra
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

In response to


Browse pgsql-hackers by date

  From Date Subject
Next Message Alexey Kondratov 2020-02-29 12:35:27 Re: Allow CLUSTER, VACUUM FULL and REINDEX to change tablespace on the fly
Previous Message Erik Rijkers 2020-02-29 12:20:02 Re: Yet another fast GiST build (typo)