Thread: Re: Plans for 8.2?
>> >> Too bad - I think that will keep a lot of potential users from >> evaluating Pg as a serious alternative. Good or bad, decide for >> yourself :) >Why on earth should that be? Citing Baldur Norddahl (http://archives.postgresql.org/pgsql-general/2006-01/msg00597.php): "I will also point out that none of the replication solutions have the same solid reputation as postgresql. As long the postgresqlteam will not endorse a replication solution, you can not expect people to put the same trust in these solutionsas we put into postgresql itself." /Mikael
Sigh. This old chestnut. On Fri, Jan 13, 2006 at 03:15:06PM +0100, Mikael Carneholm wrote: > > Citing Baldur Norddahl > (http://archives.postgresql.org/pgsql-general/2006-01/msg00597.php): > > "I will also point out that none of the replication solutions have > the same solid reputation as postgresql. As long the postgresql > team will not endorse a replication solution, you can not expect > people to put the same trust in these solutions as we put into > postgresql itself." 1. Define "postgresql team". What, CORE members? Like Jan, perhaps? Or maybe you mean active developers? Like the ones doing the coding on the project? 2. Define "endorse". Does that mean "in the backend"? So everyone has to pay the performance penalty even though they won't all use it? Even though no other database system makes you make that compromise? 3. Define "solid reputation". Does this mean "has been around 10 years?" I gots news for you: until the pg_timemachine module is ready, we're not going to be able to add things retrospectively to the history of Postgresql. And does this mean that WAL isn't solid, because it's not 10 years old? Hmm. What about PITR? 4. Define "trust". You mean that people are just picking up Postgres and using it because it has a good reputation, and doing no testing? Well, I wish them lots of luck, whatever system they use. If you think there is something wrong with some bit of code, in Slony or any other project, it would be _really nice_ to see a proposal to fix it, rather than claim that there's some magic "in the core" bit that is somehow the thing which makes code reliable. Or, if you think that there is some all-others-killing set of packages that needs to be put together as a world-domination solution, please start a project on pgfoundry to build the integrated distribution. That's what free software is for. If the complaint is instead, "Company O and Company I have big giant marketing departments that turn 85 modules into 85*85 different products! How come we don't?" then I suggest you have failed to grasp exactly where the strengths of community based development lie. (I'm also more than a little impatient with people who moan about how this or that replication system isn't a substitute for Oracle's RAC orgrid computing or whatever. I guess we oughta get to work, eh? A Mere Matter of Programming, that one. But I think I've ranted enough.) A -- Andrew Sullivan | ajs@crankycanuck.ca When my information changes, I alter my conclusions. What do you do sir? --attr. John Maynard Keynes
On Fri, Jan 13, 2006 at 03:15:06PM +0100, Mikael Carneholm wrote: > >> > >> Too bad - I think that will keep a lot of potential users from > >> evaluating Pg as a serious alternative. Good or bad, decide for > >> yourself :) > > >Why on earth should that be? > > Citing Baldur Norddahl (http://archives.postgresql.org/pgsql-general/2006-01/msg00597.php): > > "I will also point out that none of the replication solutions have the same solid reputation as postgresql. As long thepostgresql team will not endorse a replication solution, you can not expect people to put the same trust in these solutionsas we put into postgresql itself." I'm not really sure what you're looking for here. None of the replication solutions have the same reputation as PostgreSQL itself because they're both newer than PostgreSQL itself and used by a much smaller number of people. If you want to increase the reputation of a replication solution, it's going to take something other than trying to get core to put out some kind of endorsement. Case studies of real-world users is something that would help. Showing what kind of test coverage there is wouldn't hurt. Performance tests would be good. In other words, if promoting replication is important to you, there's plenty of things you can do that will help on that front. But as others have said, the various replication solutions are going to have to stand (or fall) on their own merits. -- Jim C. Nasby, Sr. Engineering Consultant jnasby@pervasive.com Pervasive Software http://pervasive.com work: 512-231-6117 vcard: http://jim.nasby.net/pervasive.vcf cell: 512-569-9461
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Easy everyone. Let's not bite the newcomers too hard here. > 2. Define "endorse". Does that mean "in the backend"? So > everyone has to pay the performance penalty even though they won't > all use it? Even though no other database system makes you make that > compromise? I would presume that at least packaged with PG (in the contrib section) would be a good start. A prominent, east to find link to Slony on the website would help too. I just did a test to see what comes up when I typed "replication" in the search box at postgresql.org. Got a 503 error. We really need to work on that. Bad enough we don't use Postgres to do the searching. I'd better stop here before I start ranting myself. I didn't expect that 503 error when I started this letter. - -- Greg Sabino Mullane greg@turnstep.com PGP Key: 0x14964AC8 200601131734 http://biglumber.com/x/web?pk=2529DF6AB8F79407E94445B4BC9B906714964AC8 -----BEGIN PGP SIGNATURE----- iD8DBQFDyCt4vJuQZxSWSsgRAkXZAJ4hvwlENtOxGPh1x+vNu3++izLQCQCgsqCa rW1MUxPxDqYFbdgontgxuwk= =ZlIa -----END PGP SIGNATURE-----
On Fri, Jan 13, 2006 at 10:39:06PM -0000, Greg Sabino Mullane wrote: > > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > > Easy everyone. Let's not bite the newcomers too hard here. > > > 2. Define "endorse". Does that mean "in the backend"? So > > everyone has to pay the performance penalty even though they won't > > all use it? Even though no other database system makes you make that > > compromise? > > I would presume that at least packaged with PG (in the contrib section) > would be a good start. A prominent, east to find link to Slony on > the website would help too. Why just Slony? There's at least 2 other free replication solutions I can think of off the top of my head, and I'm sure I'm missing some. And there was rather extensive discussion about contrib on -hackers about 6 months ago. IIRC the decision was that the only reason to put something in contrib was if it was either dependant on specific backend code or if it was targeted for inclusion into the backend. -- Jim C. Nasby, Sr. Engineering Consultant jnasby@pervasive.com Pervasive Software http://pervasive.com work: 512-231-6117 vcard: http://jim.nasby.net/pervasive.vcf cell: 512-569-9461
Jim C. Nasby wrote: > On Fri, Jan 13, 2006 at 10:39:06PM -0000, Greg Sabino Mullane wrote: >> -----BEGIN PGP SIGNED MESSAGE----- >> Hash: SHA1 >> >> >> Easy everyone. Let's not bite the newcomers too hard here. >> >>> 2. Define "endorse". Does that mean "in the backend"? So >>> everyone has to pay the performance penalty even though they won't >>> all use it? Even though no other database system makes you make that >>> compromise? >> I would presume that at least packaged with PG (in the contrib section) >> would be a good start. A prominent, east to find link to Slony on >> the website would help too. > > Why just Slony? There's at least 2 other free replication solutions I > can think of off the top of my head, and I'm sure I'm missing some. Slony is the only free OSS postgreSQL replication solution that I would ever suggest to someone. However if the project is going to start suggesting replication solutions it should suggest all of them. Sincerely, Joshua D. Drake -- The PostgreSQL Company - Command Prompt, Inc. 1.503.667.4564 PostgreSQL Replication, Consulting, Custom Development, 24x7 support Managed Services, Shared and Dedicated Hosting Co-Authors: plPHP, plPerlNG - http://www.commandprompt.com/
On Fri, Jan 13, 2006 at 10:39:06PM -0000, Greg Sabino Mullane wrote:
>
> I would presume that at least packaged with PG (in the contrib section)
> would be a good start. A prominent, east to find link to Slony on
But in Slony's case, that'd be a regression, not an improvement. It
is designed, on purpose, as a bolt-on. That's a feature, not a bug.
It allows you to do version upgrades with just a few minutes'
switchover time, to begin with, which is something that we don't
otherwise have.
If we want to fix the "in the tarball, or it's not real", we need to
continue to make packages easy to install. Nobody thinks that the
DBI is some sort of stupid tacky not-ready tool just because every
installation of Perl doesn't have it automatically.
A
--
Andrew Sullivan | ajs@crankycanuck.ca
A certain description of men are for getting out of debt, yet are
against all taxes for raising money to pay it off.
--Alexander Hamilton
On Jan 13, 2006, at 4:00 PM, Joshua D. Drake wrote: > >> Why just Slony? There's at least 2 other free replication solutions I >> can think of off the top of my head, and I'm sure I'm missing some. > > Slony is the only free OSS postgreSQL replication solution that I > would ever suggest to someone. > > However if the project is going to start suggesting replication > solutions it should suggest all of them. > > Sincerely, > > Joshua D. Drake Why would you only recommend Slony? How does it compare to Sequoia or p/cluster? I have to admit that reading about Slony II sounded very good, but it's apparently far off from reality. What's the best solution that would work on OS X Server? Thanks, ____________________________________________________________________ Brendan Duddridge | CTO | 403-277-5591 x24 | brendan@clickspace.com ClickSpace Interactive Inc. Suite L100, 239 - 10th Ave. SE Calgary, AB T2G 0V9 http://www.clickspace.com
Attachment
>> Sincerely, >> >> Joshua D. Drake > > > Why would you only recommend Slony? How does it compare to Sequoia or > p/cluster? Well p/cluster is not OSS. Sequioa is but is query based and doesn't correctly deal with things like now(). I was speaking directly about OSS replication. Sincerely, Joshua D. Drake -- The PostgreSQL Company - Command Prompt, Inc. 1.503.667.4564 PostgreSQL Replication, Consulting, Custom Development, 24x7 support Managed Services, Shared and Dedicated Hosting Co-Authors: plPHP, plPerlNG - http://www.commandprompt.com/
> I just did a test to see what comes up when I typed > "replication" in the search box at postgresql.org. Got a 503 > error. We really need to work on that. Bad enough we don't > use Postgres to do the searching. Yes we do. We certainly don't back aspseek with mysql... (No, we don't use the standard version, the one we use is fairly extensively modified) Oh, and I beleive the 503 error is solved. And yes, that definitly could've been handled better. //Magnus
> Sure, but it still means installing an external tool, which requires > PHP, which isn't trivial to install in it's own right. And afaik there's > still no way to find out how much IO each query did, how much CPU was > spent, if any sorts overflowed, etc., etc. Exactly my point, and thus the reason why I'd like to see such functionality implemented. Since I haven't hacked Pg one singlebit (apart from increasing the max identifier name length way back by changing some #define, iirc), I don't reallythink I'm the right person to do it. :) /Mikael
On Mon, Jan 16, 2006 at 09:21:55AM +0100, H.J. Sanders wrote: > Thanks for you answer. I've looked for the stats but I could not find > it in our 7.4 , but I will look further (perhaps SUSE didn't > install them). How did you look? You shouldn't have to install anything extra; statistics collection is built-in to the backend. It's generally not enabled by default, however, so you might have to modify your server configuration (postgresql.conf). See "Statistics Collection Configuration" in the documentation: http://www.postgresql.org/docs/7.4/interactive/monitoring-stats.html -- Michael Fuhr
On Fri, Jan 13, 2006 at 05:30:49PM -0800, Joshua D. Drake wrote: > >Why would you only recommend Slony? How does it compare to Sequoia or > >p/cluster? > > Well p/cluster is not OSS. Sequioa is but is query based and doesn't > correctly deal with things like now(). > > I was speaking directly about OSS replication. Correct me if I'm wrong, but AFAIK Slony is the only OSS replication that isn't statement based, which as Josh mentioned has some serious ramifications. -- Jim C. Nasby, Sr. Engineering Consultant jnasby@pervasive.com Pervasive Software http://pervasive.com work: 512-231-6117 vcard: http://jim.nasby.net/pervasive.vcf cell: 512-569-9461
>> Well p/cluster is not OSS. Sequioa is but is query based and doesn't >> correctly deal with things like now(). >> >> I was speaking directly about OSS replication. >> > > Correct me if I'm wrong, but AFAIK Slony is the only OSS replication > that isn't statement based, which as Josh mentioned has some serious > ramifications. > Yes that is correct to my knowledge. Sincerely, Joshua D. Drake -- The PostgreSQL Company - Command Prompt, Inc. 1.503.667.4564 PostgreSQL Replication, Consulting, Custom Development, 24x7 support Managed Services, Shared and Dedicated Hosting Co-Authors: PLphp, PLperl - http://www.commandprompt.com/