Re: Index corruption with CREATE INDEX CONCURRENTLY

From: Vladimir Borodin <root(at)simply(dot)name>
To: Peter Geoghegan <pg(at)bowt(dot)ie>
Cc: Tomas Vondra <tomas(dot)vondra(at)2ndquadrant(dot)com>, Robert Haas <robertmhaas(at)gmail(dot)com>, Martín Marqués <martin(at)2ndquadrant(dot)com>, Michael Paquier <michael(dot)paquier(at)gmail(dot)com>, Pavel Stehule <pavel(dot)stehule(at)gmail(dot)com>, Pavan Deolasee <pavan(dot)deolasee(at)gmail(dot)com>, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>, Alvaro Herrera <alvherre(at)2ndquadrant(dot)com>, pgsql-hackers <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: Index corruption with CREATE INDEX CONCURRENTLY
Date: 2017-02-06 08:48:28
Message-ID: 2A9246F1-28EA-4406-AC0C-D1B643DD2DEC@simply.name
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers


> 6 февр. 2017 г., в 4:57, Peter Geoghegan <pg(at)bowt(dot)ie> написал(а):
>
> I meant that I find the fact that there were no user reports in all
> these years to be a good reason to not proceed for now in this
> instance.

Well, we had some strange situations with indexes (see below for example) but I couldn’t even think that CIC is the problem.

And it is really difficult to give diagnostics for problems of such kind. Because 1. you may see the problem several days after last major change in the database and 2. you don’t even know how to start investigating the problem.

xdb314g/maildb M # show enable_indexscan ;
enable_indexscan
------------------
off
(1 row)

Time: 0.120 ms
xdb314g/maildb M # select * from mail.folders where uid=448300241 and fid=1;
-[ RECORD 1 ]---+------------------------------
uid | 448300241
fid | 1
<...>

Time: 30398.637 ms
xdb314g/maildb M # set enable_indexscan to true;
SET
Time: 0.111 ms
xdb314g/maildb M # select * from mail.folders where uid=448300241 and fid=1;
(0 rows)

Time: 0.312 ms

xdb314g/maildb M #

The row of course hasn’t been deleted.

--
May the force be with you…
https://simply.name

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Amit Langote 2017-02-06 09:01:58 Re: contrib modules and relkind check
Previous Message Tsunakawa, Takayuki 2017-02-06 07:44:21 Re: [RFC] Should I embed or parameterize syscall/Win32 function names from error messages?