Horton

Re: VACUUM and ANALYZE disagreeing on what reltuples means

Tomas Vondra <tomas.vondra@2ndquadrant.com>

View thread View on postgr.es
From: Tomas Vondra <tomas.vondra@2ndquadrant.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: pgsql-hackers@postgresql.org
Date: 2017-07-25T11:33:28Z
Lists: pgsql-hackers

Attachments

On 7/25/17 12:55 AM, Tom Lane wrote:
> Tomas Vondra <tomas.vondra@2ndquadrant.com> writes:
>> It seems to me that VACUUM and ANALYZE somewhat disagree on what
>> exactly reltuples means. VACUUM seems to be thinking that reltuples
>> = live + dead while ANALYZE apparently believes that reltuples =
>> live
> 
>> The question is - which of the reltuples definitions is the right
>> one? I've always assumed that "reltuples = live + dead" but perhaps
>> not?
> 
> I think the planner basically assumes that reltuples is the live
> tuple count, so maybe we'd better change VACUUM to get in step.
> 

Attached is a patch that (I think) does just that. The disagreement was 
caused by VACUUM treating recently dead tuples as live, while ANALYZE 
treats both of those as dead.

At first I was worried that this will negatively affect plans in the 
long-running transaction, as it will get underestimates (due to 
reltuples not including rows it can see). But that's a problem we 
already have anyway, you just need to run ANALYZE in the other session.

regards

-- 
Tomas Vondra                  http://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

Commits

  1. Sync up our various ways of estimating pg_class.reltuples.

  2. Fix tuple counting in SP-GiST index build.

  3. Fix errors in contrib/bloom index build.

  4. When updating reltuples after ANALYZE, just extrapolate from our sample.

  5. Avoid holding AutovacuumScheduleLock while rechecking table statistics.