Re: [HACKERS] [PATCH] Vacuum: Update FSM more frequently

From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: Claudio Freire <klaussfreire(at)gmail(dot)com>
Cc: Masahiko Sawada <sawada(dot)mshk(at)gmail(dot)com>, PostgreSQL-Dev <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: [HACKERS] [PATCH] Vacuum: Update FSM more frequently
Date: 2018-03-29 22:55:04
Message-ID: 25356.1522364104@sss.pgh.pa.us
Views: Raw Message | Whole Thread | Download mbox
Thread:
Lists: pgsql-hackers

I wrote:
> I have to go do something
> else right now, but I'll take a closer look at 0004 later.

OK, so after studying 0004, it seems to me that we could do it more
simply as attached; that is, move the IndexFreeSpaceMapVacuum calls
into btvacuumscan/spgvacuumscan, do them only if we found any recyclable
pages, and drop the calls in btvacuumcleanup/spgvacuumcleanup altogether.

The reason I think this is sufficient is that the scans find and record
every reusable page in the index, no matter whether they were recorded
before or not. Therefore, if we don't find any such pages, there's
nothing useful in the FSM and no particular urgency about making its
upper pages up-to-date. It's true that if the FSM is actually corrupt,
leaving that to be fixed retail by searches is probably less efficient
than doing an IndexFreeSpaceMapVacuum call would be --- but *only* if
you assume that the problem is just in the upper pages and the leaf
pages are all fine. That doesn't seem to be a case we should optimize
for.

I realized that the reason BRIN doesn't go through indexfsm.c is
that it's actually interested in less-than-page-size free space.
So it's using the right API. However, it still looks to me like
we could win something there by replacing FreeSpaceMapVacuum calls
with FreeSpaceMapVacuumRange calls. I've not wrapped my head around
the logic completely, but it looks like there are several places
where it calls FreeSpaceMapVacuum to bubble up a free-space update
for just a single page. If so, that's much less efficient than it
could be.

regards, tom lane

Attachment Content-Type Size
0004-update-index-fsm-more-often-v3.patch text/x-diff 3.8 KB

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Andres Freund 2018-03-29 23:14:25 Re: JIT compiling with LLVM v12.2
Previous Message Tatsuo Ishii 2018-03-29 22:49:37 Creating streaming replication standby