Thread: Re: [COMMITTERS] pgsql: Remove replicaiton FAQ item.

Re: [COMMITTERS] pgsql: Remove replicaiton FAQ item.

From
Bruce Momjian
Date:
Marc G. Fournier wrote:
> On Tue, 16 Nov 2004, Robert Treat wrote:
>
> > On Sunday 14 November 2004 16:33, Bruce Momjian wrote:
> >> Log Message:
> >> -----------
> >> Remove replicaiton FAQ item.
> >>
> >> Modified Files:
> >> --------------
> >>     pgsql/doc:
> >>         FAQ (r1.269 -> r1.270)
> >>
> >> (http://developer.postgresql.org/cvsweb.cgi/pgsql/doc/FAQ.diff?r1=1.269&r2=
> >> 1.270) pgsql/doc/src/FAQ:
> >>         FAQ.html (r1.227 -> r1.228)
> >>
> >> (http://developer.postgresql.org/cvsweb.cgi/pgsql/doc/src/FAQ/FAQ.html.diff
> >> ?r1=1.227&r2=1.228)
> >
> > While the answer may have been a little dated, this still seems to be an FAQ.
> > Was there a discussion on removing this?  If so I'll go read that, but ISTM
> > this question should remain.
>
> agreed, this definitly does fall under a "Very FAQ" kinda thing ...
> pointers to the various methods that are currently in use, at a minimum,
> would be good ...

Yep, big mistake my removing the replication FAQ item. I have re-added a
new one:

    4.21) What replication solutions are available?

    Though "replication" is a single term, there are two major replication
    technologies. Multi-master replication allows read/write queries to be
    sent to multiple replicated computers. Using this capability has a
    serious performance impact. Pgcluster is the most popular such solution
    available for PostgreSQL. Master-slave replication allows a single
    master to receive read/write queries, with slaves able to accept only
    read/SELECT queries. The most popular master-slave PostgreSQL
    replication solution is Slony. There are also commercial and
    hardware-based replication solutions available.

--
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073

Re: [COMMITTERS] pgsql: Remove replicaiton FAQ item.

From
Jeff Davis
Date:
Should the text include a mention of synchronous vs. asynchronous? Or
does master-master imply synchronous?

Regards,
    Jeff Davis

On Sun, 2005-04-24 at 08:01 -0400, Bruce Momjian wrote:
> Marc G. Fournier wrote:
> > On Tue, 16 Nov 2004, Robert Treat wrote:
> >
> > > On Sunday 14 November 2004 16:33, Bruce Momjian wrote:
> > >> Log Message:
> > >> -----------
> > >> Remove replicaiton FAQ item.
> > >>
> > >> Modified Files:
> > >> --------------
> > >>     pgsql/doc:
> > >>         FAQ (r1.269 -> r1.270)
> > >>
> > >> (http://developer.postgresql.org/cvsweb.cgi/pgsql/doc/FAQ.diff?r1=1.269&r2=
> > >> 1.270) pgsql/doc/src/FAQ:
> > >>         FAQ.html (r1.227 -> r1.228)
> > >>
> > >> (http://developer.postgresql.org/cvsweb.cgi/pgsql/doc/src/FAQ/FAQ.html.diff
> > >> ?r1=1.227&r2=1.228)
> > >
> > > While the answer may have been a little dated, this still seems to be an FAQ.
> > > Was there a discussion on removing this?  If so I'll go read that, but ISTM
> > > this question should remain.
> >
> > agreed, this definitly does fall under a "Very FAQ" kinda thing ...
> > pointers to the various methods that are currently in use, at a minimum,
> > would be good ...
>
> Yep, big mistake my removing the replication FAQ item. I have re-added a
> new one:
>
>     4.21) What replication solutions are available?
>
>     Though "replication" is a single term, there are two major replication
>     technologies. Multi-master replication allows read/write queries to be
>     sent to multiple replicated computers. Using this capability has a
>     serious performance impact. Pgcluster is the most popular such solution
>     available for PostgreSQL. Master-slave replication allows a single
>     master to receive read/write queries, with slaves able to accept only
>     read/SELECT queries. The most popular master-slave PostgreSQL
>     replication solution is Slony. There are also commercial and
>     hardware-based replication solutions available.
>


Re: [COMMITTERS] pgsql: Remove replicaiton FAQ item.

From
"Joshua D. Drake"
Date:
Jeff Davis wrote:
> Should the text include a mention of synchronous vs. asynchronous? Or
> does master-master imply synchronous?

It shouldn't. It is possible to have synchronous replication without it
being master<->master.

Sincerely,

Joshua D. Drake


>
> Regards,
>     Jeff Davis
>
> On Sun, 2005-04-24 at 08:01 -0400, Bruce Momjian wrote:
>
>>Marc G. Fournier wrote:
>>
>>>On Tue, 16 Nov 2004, Robert Treat wrote:
>>>
>>>
>>>>On Sunday 14 November 2004 16:33, Bruce Momjian wrote:
>>>>
>>>>>Log Message:
>>>>>-----------
>>>>>Remove replicaiton FAQ item.
>>>>>
>>>>>Modified Files:
>>>>>--------------
>>>>>    pgsql/doc:
>>>>>        FAQ (r1.269 -> r1.270)
>>>>>
>>>>>(http://developer.postgresql.org/cvsweb.cgi/pgsql/doc/FAQ.diff?r1=1.269&r2=
>>>>>1.270) pgsql/doc/src/FAQ:
>>>>>        FAQ.html (r1.227 -> r1.228)
>>>>>
>>>>>(http://developer.postgresql.org/cvsweb.cgi/pgsql/doc/src/FAQ/FAQ.html.diff
>>>>>?r1=1.227&r2=1.228)
>>>>
>>>>While the answer may have been a little dated, this still seems to be an FAQ.
>>>>Was there a discussion on removing this?  If so I'll go read that, but ISTM
>>>>this question should remain.
>>>
>>>agreed, this definitly does fall under a "Very FAQ" kinda thing ...
>>>pointers to the various methods that are currently in use, at a minimum,
>>>would be good ...
>>
>>Yep, big mistake my removing the replication FAQ item. I have re-added a
>>new one:
>>
>>    4.21) What replication solutions are available?
>>
>>    Though "replication" is a single term, there are two major replication
>>    technologies. Multi-master replication allows read/write queries to be
>>    sent to multiple replicated computers. Using this capability has a
>>    serious performance impact. Pgcluster is the most popular such solution
>>    available for PostgreSQL. Master-slave replication allows a single
>>    master to receive read/write queries, with slaves able to accept only
>>    read/SELECT queries. The most popular master-slave PostgreSQL
>>    replication solution is Slony. There are also commercial and
>>    hardware-based replication solutions available.
>>
>
>
>
> ---------------------------(end of broadcast)---------------------------
> TIP 8: explain analyze is your friend


Re: [COMMITTERS] pgsql: Remove replicaiton FAQ item.

From
Bruce Momjian
Date:
Joshua D. Drake wrote:
> Jeff Davis wrote:
> > Should the text include a mention of synchronous vs. asynchronous? Or
> > does master-master imply synchronous?
>
> It shouldn't. It is possible to have synchronous replication without it
> being master<->master.

We could mention async vs sync but at the FAQ stage I think the
multi-master/master-slave disinction is enough. Let's see what questions
we get.

--
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073

Re: [COMMITTERS] pgsql: Remove replicaiton FAQ item.

From
Mischa Sandberg
Date:
Quoting Bruce Momjian <pgman@candle.pha.pa.us>:

> Joshua D. Drake wrote:
> > Jeff Davis wrote:
> > > Should the text include a mention of synchronous vs.
> asynchronous? Or
> > > does master-master imply synchronous?
> >
> > It shouldn't. It is possible to have synchronous replication
> without it
> > being master<->master.
>
> We could mention async vs sync but at the FAQ stage I think the
> multi-master/master-slave disinction is enough. Let's see what
> questions
> we get.

About all you can say is, it is not possible to implement synchronous
master-master replication, because of conflict resolution of
simultaneous transactions.
--
"Dreams come true, not free." -- S.Sondheim, ITW


Re: [COMMITTERS] pgsql: Remove replicaiton FAQ item.

From
Bruce Momjian
Date:
Mischa Sandberg wrote:
> Quoting Bruce Momjian <pgman@candle.pha.pa.us>:
>
> > Joshua D. Drake wrote:
> > > Jeff Davis wrote:
> > > > Should the text include a mention of synchronous vs.
> > asynchronous? Or
> > > > does master-master imply synchronous?
> > >
> > > It shouldn't. It is possible to have synchronous replication
> > without it
> > > being master<->master.
> >
> > We could mention async vs sync but at the FAQ stage I think the
> > multi-master/master-slave disinction is enough. Let's see what
> > questions
> > we get.
>
> About all you can say is, it is not possible to implement synchronous
> master-master replication, because of conflict resolution of
> simultaneous transactions.

You mean asynchronous master-master?  There are some implementations
(Sybase?, pgreplicator) that do this and have conflict resolution.

--
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073

Re: [COMMITTERS] pgsql: Remove replicaiton FAQ item.

From
Mischa Sandberg
Date:
Quoting Bruce Momjian <pgman@candle.pha.pa.us>:

> Mischa Sandberg wrote:
> > About all you can say is, it is not possible to implement
> synchronous
> > master-master replication, because of conflict resolution of
> > simultaneous transactions.
>
> You mean asynchronous master-master?  There are some implementations
> (Sybase?, pgreplicator) that do this and have conflict resolution.

Forgive my kibitz, then; perhaps my head has been stuck in sand a
while. You're right, I forgot about the nicer bits of Sybase
RepServer. Synchronous replication was scratched early in the project,
because two-phase commit took too long across a dozen db servers.

Pgreplicator advertises itself as a "store-and-forward asynch replica
(sic) engine." (quoth http://pgreplicator.sourceforge.net/ )
--
"Dreams come true, not free." -- S.Sondheim, ITW


Re: [COMMITTERS] pgsql: Remove replicaiton FAQ item.

From
Christopher Browne
Date:
Martha Stewart called it a Good Thing when pgman@candle.pha.pa.us (Bruce Momjian) wrote:
> Yep, big mistake my removing the replication FAQ item. I have re-added a
> new one:
>
>     4.21) What replication solutions are available?
>
>     Though "replication" is a single term, there are two major replication
>     technologies. Multi-master replication allows read/write queries to be
>     sent to multiple replicated computers. Using this capability has a
>     serious performance impact. Pgcluster is the most popular such solution
>     available for PostgreSQL. Master-slave replication allows a single
>     master to receive read/write queries, with slaves able to accept only
>     read/SELECT queries. The most popular master-slave PostgreSQL
>     replication solution is Slony. There are also commercial and
>     hardware-based replication solutions available.

A suggestion for rewording:

Though "replication" is a single term, there are a number of
approaches and technologies to handle it, and thus, for different
purposes, a number of replication systems are available.

Master/slave replication allows a single master to receive read/write
queries, where slaves can only accept read/SELECT queries.  The most
popular freely available master-slave PostgreSQL replication solution
is Slony-I.

Multi-master replication allows read/write queries to be sent to
multiple replicated computers.  Using this capability generaly has a
fairly severe impact on performance due to the need to synchronize
changes between servers.  Pgcluster is the most popular such solution
freely available for PostgreSQL.

There are also commercial and hardware-based replication solutions
available supporting a variety of replication models.
--
(format nil "~S@~S" "cbbrowne" "acm.org")
http://linuxdatabases.info/info/slony.html
"It   can be   shown   that for any  nutty  theory,  beyond-the-fringe
political view or  strange religion there  exists  a proponent  on the
Net. The proof is left as an exercise for your kill-file."
-- Bertil Jonell

Re: [COMMITTERS] pgsql: Remove replicaiton FAQ item.

From
Bruce Momjian
Date:
OK, updated:

    <P>Though "replication" is a single term, there are several technologies
    for doing replication, with advantages and disadvantages for each.</P>

    <P>Master/slave replication allows a single master to receive read/write
    queries, while slaves can only accept read/<SMALL>SELECT</SMALL> queries.
    The most popular freely available master-slave PostgreSQL replication
    solution is <A href="http://gborg.postgresql.org/project/slony1/projdisplay$
    Slony-I</A>.</P>

    <P>Multi-master replication allows read/write queries to be sent to
    multiple replicated computers.  This capability also has a severe
    impact on performance due to the need to synchronize changes between
    servers.  <A href="http://pgfoundry.org/projects/pgcluster/">Pgcluster</a>
    is the most popular such solution freely available for PostgreSQL.</P>

    <P>There are also commercial and hardware-based replication solutions
    available supporting a variety of replication models.</P>


---------------------------------------------------------------------------

Christopher Browne wrote:
> Martha Stewart called it a Good Thing when pgman@candle.pha.pa.us (Bruce Momjian) wrote:
> > Yep, big mistake my removing the replication FAQ item. I have re-added a
> > new one:
> >
> >     4.21) What replication solutions are available?
> >
> >     Though "replication" is a single term, there are two major replication
> >     technologies. Multi-master replication allows read/write queries to be
> >     sent to multiple replicated computers. Using this capability has a
> >     serious performance impact. Pgcluster is the most popular such solution
> >     available for PostgreSQL. Master-slave replication allows a single
> >     master to receive read/write queries, with slaves able to accept only
> >     read/SELECT queries. The most popular master-slave PostgreSQL
> >     replication solution is Slony. There are also commercial and
> >     hardware-based replication solutions available.
>
> A suggestion for rewording:
>
> Though "replication" is a single term, there are a number of
> approaches and technologies to handle it, and thus, for different
> purposes, a number of replication systems are available.
>
> Master/slave replication allows a single master to receive read/write
> queries, where slaves can only accept read/SELECT queries.  The most
> popular freely available master-slave PostgreSQL replication solution
> is Slony-I.
>
> Multi-master replication allows read/write queries to be sent to
> multiple replicated computers.  Using this capability generaly has a
> fairly severe impact on performance due to the need to synchronize
> changes between servers.  Pgcluster is the most popular such solution
> freely available for PostgreSQL.
>
> There are also commercial and hardware-based replication solutions
> available supporting a variety of replication models.
> --
> (format nil "~S@~S" "cbbrowne" "acm.org")
> http://linuxdatabases.info/info/slony.html
> "It   can be   shown   that for any  nutty  theory,  beyond-the-fringe
> political view or  strange religion there  exists  a proponent  on the
> Net. The proof is left as an exercise for your kill-file."
> -- Bertil Jonell
>
> ---------------------------(end of broadcast)---------------------------
> TIP 6: Have you searched our list archives?
>
>                http://archives.postgresql.org
>

--
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073