Re: [HACKERS] Re: [HACKERS] GSoC 2015 proposal: Improve the performance of “ALTER TABLE .. SET LOGGED / UNLOGGED” statement
On 2015-07-09 10:39:35 -0300, Fabrízio de Royes Mello wrote: If the wal_level=minimal we don't need to force the wal log of the contents. If the wal_level != minimal we need just to xlog all the pages, but in both cases we don't need the extra job to create a new datafiles and copy the contents between them. So we'll improve performance, or am I missing something? Sure. It'll be a bit faster. I just don't see the peformance increase in not that common situations being worth the price we'll pay in development, code review, debugging and then maintaining some nontrivial code. If this were likely to be a 15 line patch, I'd think differently. -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
[HACKERS] Re: [HACKERS] Re: [HACKERS] GSoC 2015 proposal: Improve the performance of “ALTER TABLE .. SET LOGGED / UNLOGGED” statement
On Thu, Jul 9, 2015 at 10:56 AM, Andres Freund and...@anarazel.de wrote: On 2015-07-09 10:39:35 -0300, Fabrízio de Royes Mello wrote: If the wal_level=minimal we don't need to force the wal log of the contents. If the wal_level != minimal we need just to xlog all the pages, but in both cases we don't need the extra job to create a new datafiles and copy the contents between them. So we'll improve performance, or am I missing something? Sure. It'll be a bit faster. I just don't see the peformance increase in not that common situations being worth the price we'll pay in development, code review, debugging and then maintaining some nontrivial code. If this were likely to be a 15 line patch, I'd think differently. Well, this will not be a 15 line patch, but I'll try to simplify as much as possible. Regards, -- Fabrízio de Royes Mello Consultoria/Coaching PostgreSQL Timbira: http://www.timbira.com.br Blog: http://fabriziomello.github.io Linkedin: http://br.linkedin.com/in/fabriziomello Twitter: http://twitter.com/fabriziomello Github: http://github.com/fabriziomello
Re: [HACKERS] Re: [HACKERS] GSoC 2015 proposal: Improve the performance of “ALTER TABLE .. SET LOGGED / UNLOGGED” statement
Andres Freund and...@anarazel.de writes: On 2015-07-09 10:39:35 -0300, Fabrízio de Royes Mello wrote: If the wal_level=minimal we don't need to force the wal log of the contents. If the wal_level != minimal we need just to xlog all the pages, but in both cases we don't need the extra job to create a new datafiles and copy the contents between them. So we'll improve performance, or am I missing something? Sure. It'll be a bit faster. I just don't see the peformance increase in not that common situations being worth the price we'll pay in development, code review, debugging and then maintaining some nontrivial code. If this were likely to be a 15 line patch, I'd think differently. I'm even more worried about the possible reliability problems (ie introduction of bugs, which are likely to be of the data-eating variety) that such changes are likely to incur. So I tend to agree with Andres that this is probably not a good idea. regards, tom lane -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
[HACKERS] Re: [HACKERS] Re: [HACKERS] GSoC 2015 proposal: Improve the performance of “ALTER TABLE .. SET LOGGED / UNLOGGED” statement
On Wed, Jul 8, 2015 at 11:37 AM, Andres Freund and...@anarazel.de wrote: On 2015-07-08 10:58:51 -0300, Fabrízio de Royes Mello wrote: Think in an ETL job that can be use an unlogged table to improve the load performance, but this job create a large table and to guarantee the data consistency you need to transform it into a regular table, and with the current implementation rewrite the entire heap, toast and indexes. Don't buy that. The final target relation will usually already have content. Also everything but wal_level=minimal will force you to WAL log the contents anyway. If the wal_level=minimal we don't need to force the wal log of the contents. If the wal_level != minimal we need just to xlog all the pages, but in both cases we don't need the extra job to create a new datafiles and copy the contents between them. So we'll improve performance, or am I missing something? Regards, -- Fabrízio de Royes Mello Consultoria/Coaching PostgreSQL Timbira: http://www.timbira.com.br Blog: http://fabriziomello.github.io Linkedin: http://br.linkedin.com/in/fabriziomello Twitter: http://twitter.com/fabriziomello Github: http://github.com/fabriziomello
Re: [HACKERS] Re: [HACKERS] GSoC 2015 proposal: Improve the performance of “ALTER TABLE .. SET LOGGED / UNLOGGED” statement
On 2015-07-03 08:18:09 -0300, Fabrízio de Royes Mello wrote: *** ALTER TABLE changes ... Without going into the specifics of this change: Are we actually sure this feature warrants the complexity it'll introduce? I'm really rather doubtful. -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
[HACKERS] Re: [HACKERS] Re: [HACKERS] GSoC 2015 proposal: Improve the performance of “ALTER TABLE .. SET LOGGED / UNLOGGED” statement
On Wed, Jul 8, 2015 at 10:53 AM, Andres Freund and...@anarazel.de wrote: On 2015-07-03 08:18:09 -0300, Fabrízio de Royes Mello wrote: *** ALTER TABLE changes ... Without going into the specifics of this change: Are we actually sure this feature warrants the complexity it'll introduce? I'm really rather doubtful. Think in an ETL job that can be use an unlogged table to improve the load performance, but this job create a large table and to guarantee the data consistency you need to transform it into a regular table, and with the current implementation rewrite the entire heap, toast and indexes. Regards, -- Fabrízio de Royes Mello Consultoria/Coaching PostgreSQL Timbira: http://www.timbira.com.br Blog: http://fabriziomello.github.io Linkedin: http://br.linkedin.com/in/fabriziomello Twitter: http://twitter.com/fabriziomello Github: http://github.com/fabriziomello
Re: [HACKERS] Re: [HACKERS] GSoC 2015 proposal: Improve the performance of “ALTER TABLE .. SET LOGGED / UNLOGGED” statement
On 2015-07-08 10:58:51 -0300, Fabrízio de Royes Mello wrote: Think in an ETL job that can be use an unlogged table to improve the load performance, but this job create a large table and to guarantee the data consistency you need to transform it into a regular table, and with the current implementation rewrite the entire heap, toast and indexes. Don't buy that. The final target relation will usually already have content. Also everything but wal_level=minimal will force you to WAL log the contents anyway. -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
