Re: Avoiding smgrimmedsync() during nbtree index builds

From: Justin Pryzby <pryzby(at)telsasoft(dot)com>
To: Melanie Plageman <melanieplageman(at)gmail(dot)com>
Cc: Andres Freund <andres(at)anarazel(dot)de>, Heikki Linnakangas <hlinnaka(at)iki(dot)fi>, pgsql-hackers(at)postgresql(dot)org, Peter Geoghegan <pg(at)bowt(dot)ie>, Thomas Munro <thomas(dot)munro(at)gmail(dot)com>
Subject: Re: Avoiding smgrimmedsync() during nbtree index builds
Date: 2022-03-03 01:09:49
Message-ID: 20220303010949.GC27651@telsasoft.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Rebased to appease cfbot.

I ran these paches under a branch which shows code coverage in cirrus. It
looks good to my eyes.
https://api.cirrus-ci.com/v1/artifact/task/5212346552418304/coverage/coverage/00-index.html

Are these patches being considered for v15 ?

Attachment Content-Type Size
0001-Add-unbuffered-IO-API.patch text/x-diff 31.8 KB
0002-Avoid-immediate-fsync-for-unbuffered-IO.patch text/x-diff 12.2 KB
0003-BTree-index-use-unbuffered-IO-optimization.patch text/x-diff 2.1 KB
0004-Use-shared-buffers-when-possible-for-index-build.patch text/x-diff 19.1 KB

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Michael Paquier 2022-03-03 01:10:36 Re: pg_stop_backup() v2 incorrectly marked as proretset
Previous Message Melanie Plageman 2022-03-03 01:07:14 Why do spgbuildempty(), btbuildempty(), and blbuildempty() use smgrwrite()?