Thread: Postgres and geographically diverse replication

Postgres and geographically diverse replication

From
"Drew Myers"
Date:
Hi,

I've been given a task to build a couple of geographically separate
servers, which are capable of replicating data between each other.

I've surfed through various google results, and most of what I've found
seems to be a bit dated, so I thought I'd pose my question here, perhaps
for more detailed and more up-to-date info.

Is this normally done in a push/pull scenario within the postgres
installations themselves, or is additional software required? What are
the various replication capabilities?

I apologize for the general nature of my questions, I'm new to postgres
and to geographically separate replication. Any tips, books, whitepapers
or other resources you might be able to point me to is most appreciated.

Thanks,

Drew Myers

Re: Postgres and geographically diverse replication

From
Ben
Date:
To answer your question meaningfully, you need to provide more details.
What are you trying to get out of "replication"? High availability? Load
sharing?

On Wed, 18 Apr 2007, Drew Myers wrote:

>
> Hi,
>
> I've been given a task to build a couple of geographically separate
> servers, which are capable of replicating data between each other.
>
> I've surfed through various google results, and most of what I've found
> seems to be a bit dated, so I thought I'd pose my question here, perhaps
> for more detailed and more up-to-date info.
>
> Is this normally done in a push/pull scenario within the postgres
> installations themselves, or is additional software required? What are
> the various replication capabilities?
>
> I apologize for the general nature of my questions, I'm new to postgres
> and to geographically separate replication. Any tips, books, whitepapers
> or other resources you might be able to point me to is most appreciated.
>
> Thanks,
>
> Drew Myers
>
> ---------------------------(end of broadcast)---------------------------
> TIP 2: Don't 'kill -9' the postmaster
>

Re: Postgres and geographically diverse replication

From
Bill Moran
Date:
In response to "Drew Myers" <drew.myers@innerwireless.com>:
>
> I've been given a task to build a couple of geographically separate
> servers, which are capable of replicating data between each other.
>
> I've surfed through various google results, and most of what I've found
> seems to be a bit dated, so I thought I'd pose my question here, perhaps
> for more detailed and more up-to-date info.
>
> Is this normally done in a push/pull scenario within the postgres
> installations themselves, or is additional software required? What are
> the various replication capabilities?
>
> I apologize for the general nature of my questions, I'm new to postgres
> and to geographically separate replication. Any tips, books, whitepapers
> or other resources you might be able to point me to is most appreciated.

Generally speaking, when you're talking geographically separate, Slony
is your best bet.  We're using it to maintain data on opposites sides of
the US with good success.

http://slony.info

--
Bill Moran
http://www.potentialtech.com

Re: Postgres and geographically diverse replication

From
Brad Nicholson
Date:
On Wed, 2007-04-18 at 16:43 -0400, Bill Moran wrote:
> In response to "Drew Myers" <drew.myers@innerwireless.com>:
> >
> > I've been given a task to build a couple of geographically separate
> > servers, which are capable of replicating data between each other.
> >
> > I've surfed through various google results, and most of what I've found
> > seems to be a bit dated, so I thought I'd pose my question here, perhaps
> > for more detailed and more up-to-date info.
> >
> > Is this normally done in a push/pull scenario within the postgres
> > installations themselves, or is additional software required? What are
> > the various replication capabilities?
> >
> > I apologize for the general nature of my questions, I'm new to postgres
> > and to geographically separate replication. Any tips, books, whitepapers
> > or other resources you might be able to point me to is most appreciated.
>
> Generally speaking, when you're talking geographically separate, Slony
> is your best bet.  We're using it to maintain data on opposites sides of
> the US with good success.

Successfully using slony over a wide area is going to depend on how much
data you are replicating, how fast the connection between the two sites
is, and how stable the connection between the two sites is.

--
Brad Nicholson  416-673-4106
Database Administrator, Afilias Canada Corp.


Re: Postgres and geographically diverse replication

From
Bill Moran
Date:
In response to Brad Nicholson <bnichols@ca.afilias.info>:

> On Wed, 2007-04-18 at 16:43 -0400, Bill Moran wrote:
> > In response to "Drew Myers" <drew.myers@innerwireless.com>:
> > >
> > > I've been given a task to build a couple of geographically separate
> > > servers, which are capable of replicating data between each other.
> > >
> > > I've surfed through various google results, and most of what I've found
> > > seems to be a bit dated, so I thought I'd pose my question here, perhaps
> > > for more detailed and more up-to-date info.
> > >
> > > Is this normally done in a push/pull scenario within the postgres
> > > installations themselves, or is additional software required? What are
> > > the various replication capabilities?
> > >
> > > I apologize for the general nature of my questions, I'm new to postgres
> > > and to geographically separate replication. Any tips, books, whitepapers
> > > or other resources you might be able to point me to is most appreciated.
> >
> > Generally speaking, when you're talking geographically separate, Slony
> > is your best bet.  We're using it to maintain data on opposites sides of
> > the US with good success.
>
> Successfully using slony over a wide area is going to depend on how much
> data you are replicating, how fast the connection between the two sites
> is, and how stable the connection between the two sites is.

That's all true.  It's also true of _any_ geographically diverse replication.
Except, of course, Microsoft's.  According to Microsoft's marketing materials,
their replication system is able to exceed the speed of light.

So far, we've found Slony to do an excellent job of gracefully handing
intermittent network problems and occasional transaction spikes that
temporarily exceeded our available bandwidth.  Obviously, excessive
instances of either of these situations are going to make Slony fail,
but they'll do that to _any_ replication system, except those that can
exceed the speed of light ...

--
Bill Moran
http://www.potentialtech.com