Re: FSM corruption leading to errors

From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: Alvaro Herrera <alvherre(at)2ndquadrant(dot)com>
Cc: Michael Paquier <michael(dot)paquier(at)gmail(dot)com>, Jim Nasby <Jim(dot)Nasby(at)bluetreble(dot)com>, Pavan Deolasee <pavan(dot)deolasee(at)gmail(dot)com>, Heikki Linnakangas <hlinnaka(at)iki(dot)fi>, pgsql-hackers <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: FSM corruption leading to errors
Date: 2016-10-24 18:46:20
Message-ID: 16429.1477334780@sss.pgh.pa.us
Views: Raw Message | Whole Thread | Download mbox
Thread:
Lists: pgsql-hackers

Alvaro Herrera <alvherre(at)2ndquadrant(dot)com> writes:
> Tom Lane wrote:
>> Also, we could at least discount the FSM root page and first intermediate
>> page, no? That is, the upper limit could be
>>
>> pg_relation_size(oid::regclass, 'fsm') / 2 - 2*current_setting('block_size')::BIGINT
>>
>> I think this is a worthwhile improvement because it reduces the time spent
>> on small relations. For me, the query as given takes 9 seconds to examine
>> the regression database, which seems like a lot. Discounting two pages
>> reduces that to 20 ms.

> Hah, good one. We spent some time thinking about subtracting some value
> to make the value more accurate but it didn't occur to me to just use
> constant two.

I got the arithmetic wrong in the above, it should be like

(pg_relation_size(oid::regclass, 'fsm') - 2*current_setting('block_size')::BIGINT) / 2

With that, the runtime on HEAD's regression DB is about 700 ms, which is
still a nice win over 9000 ms.

I've put up draft wiki pages about these two problems at

https://wiki.postgresql.org/wiki/Free_Space_Map_Problems
https://wiki.postgresql.org/wiki/Visibility_Map_Problems

(thanks to Michael Paquier for initial work on the first one).
They're meant to be reasonably generic about FSM/VM problems
rather than only being about our current bugs. Please review.

regards, tom lane

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Robert Haas 2016-10-24 18:49:00 Re: Press Release Draft - 2016-10-27 Cumulative Update
Previous Message Alvaro Herrera 2016-10-24 18:31:03 Re: [COMMITTERS] pgsql: Remove extra comma at end of enum list