Re: [HACKERS] Proposal: GetOldestXminExtend for ignoring arbitraryvacuum flags - Mailing list pgsql-hackers

From Simon Riggs
Subject Re: [HACKERS] Proposal: GetOldestXminExtend for ignoring arbitraryvacuum flags
Date
Msg-id CANP8+jK1brbuX2Jy=NQYvudtS3oHZFG+o4_g5qLYVogALGo0_w@mail.gmail.com
Whole thread Raw
In response to [HACKERS] Proposal: GetOldestXminExtend for ignoring arbitrary vacuum flags  ("Seki, Eiji" <seki.eiji@jp.fujitsu.com>)
Responses Re: [HACKERS] Proposal: GetOldestXminExtend for ignoring arbitraryvacuum flags  (Michael Paquier <michael.paquier@gmail.com>)
List pgsql-hackers
On 14 February 2017 at 06:19, Seki, Eiji <seki.eiji@jp.fujitsu.com> wrote:
> In our benchmark, we found that waiting an ANALYZE process created by autovacuum daemon often has a significant
impactto the performance although the waited process do only reading as to the table.
 
...
> I'm not sure that this feature is useful in general.
> Please let me know your opinion if you are interested.

You mention the above problem and hypothesise a solution.

IMHO the discussion on this is the wrong way around. First we must be
certain that the solution is effective and has no problems, then we
decide how to code it or discuss APIs.

If you do this, ANALYZE will see an inconsistent view of data in the
table, biasing its view towards data not recently updated, which might
also have a negative performance effect.

Please persuade us with measurements that allowing this impact on
ANALYZE would really improve performance at least in your case, and
also examine the effect of this on the accuracy and usefulness of the
gathered statistics.

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



pgsql-hackers by date:

Previous
From: Masahiko Sawada
Date:
Subject: Re: [HACKERS] GUC for cleanup indexes threshold.
Next
From: Alexander Korotkov
Date:
Subject: Re: [HACKERS] Should we cacheline align PGXACT?