From: | Sergey Konoplev <gray(dot)ru(at)gmail(dot)com> |
---|---|
To: | Andres Freund <andres(at)2ndquadrant(dot)com> |
Cc: | PostgreSQL Hackers <pgsql-hackers(at)postgresql(dot)org>, Maxim Boguk <maxim(dot)boguk(at)gmail(dot)com> |
Subject: | Re: System catalog bloat removing safety |
Date: | 2013-09-18 17:03:26 |
Message-ID: | CAL_0b1tsce2KQmfhpk52v0qhAp_TRaTwm+Oi7SZY3Q0=LU2OnA@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
On Wed, Sep 18, 2013 at 2:06 AM, Andres Freund <andres(at)2ndquadrant(dot)com> wrote:
> On 2013-09-17 23:12:24 -0700, Sergey Konoplev wrote:
>> How safe is it to use the technique described by the link below with
>> system catalog tables to remove bloat?
>> (in a couple of words it is about moving tuples to the beginning of
>> the table with a special way of updating)
>> http://www.depesz.com/2010/10/17/reduce-bloat-of-table-without-longexclusive-locks/
>
> At least for pg_index it's not safe, we use the xmin of rows when
> indcheckxmin = true.
> I am not aware of other cases, but I'd be hesitant to do so.
Thank you, Andres, I did not know this. Just wondering, why it was
made of type bool and not xid?
> You have bloat because of lots of temporary tables IIRC?
Actually I am developing a bloat removing tool based on the updating technique:
https://code.google.com/p/pgtoolkit/source/browse/trunk/README
Now the --system-catalog feature marked as experimental. About
pg_index - I think the solution is to exclude it from processing
strictly.
Anyone has other ideas of what system tables it might be affected
except pg_index?
--
Kind regards,
Sergey Konoplev
PostgreSQL Consultant and DBA
http://www.linkedin.com/in/grayhemp
+1 (415) 867-9984, +7 (901) 903-0499, +7 (988) 888-1979
gray(dot)ru(at)gmail(dot)com
From | Date | Subject | |
---|---|---|---|
Next Message | Stephen Frost | 2013-09-18 17:04:21 | Re: record identical operator |
Previous Message | Fujii Masao | 2013-09-18 17:01:35 | Re: Patch for typo in src/bin/psql/command.c |