Обсуждение: Replication

Seems that there are a number of simultaneous project in trying to implement replication
for postgres.
 
Is anyone out there using this feature currently? If so , which package & OS environment.
 
the latest news was eRServer v1.0.   Is this the same as the v1.2 that Postgres INC announced
early this year ?
 
Is there a master-master capability or only master-multi-slave ?

> Is anyone out there using this feature currently? If so , which package
> & OS environment.

>  the latest news was eRServer v1.0.  Is this the same as the v1.2 that
> Postgres INC announced early this year ?

yes, exactly the same source tree ...

> Is there a master-master capability or only master-multi-slave ?

only single master, multi-slave ...

Martha Stewart called it a Good Thing whenianh@hubspan.com ("Ian Huynh")wrote:
> Seems that there are a number of simultaneous project in trying to implement replication
> for postgres.
>
> Is anyone out there using this feature currently? If so , which package & OS environment.
>
> the latest news was eRServer v1.0.   Is this the same as the v1.2 that Postgres INC announced
> early this year ?
>
> Is there a master-master capability or only master-multi-slave ?

The version of eRServer recently released as free software is not the
latest "production" release, but it is a direct predecessor.

Its operational model is that of "master/possible multiple slaves,"
NOT multimaster.

Multimaster involves some thorny issues surrounding synchronization,
as it is possible for one master to receive a set of changes
concurrently with another master receiving _incompatible_ changes.

PalmOS users see this if they change a particular record on both
desktop and palmtop; the next time they sync, the conflicting record
normally gets duplicated and the user gets to decide how to fix it.

Doing that in a big production database context may not be
possible/acceptable.  You only should do multimaster if you REALLY
know what you're doing.
--
select 'aa454' || '@' || 'freenet.carleton.ca';
http://www.ntlug.org/~cbbrowne/sap.html
Do you know where your towel is?

Thanks for all who replied.
 
There are certain cases where multi-master rep are very implementable such as where
the datasets being rep are fairly disjointed but need to be replicated for completeness in
terms of reporting. But you are absolutely 110% correct. Multimaster synchro is a nightmare.
 
 

    -----Original Message----- 
    From: Christopher Browne [mailto:cbbrowne@acm.org] 
    Sent: Sat 8/30/2003 4:47 PM 
    To: pgsql-admin@postgresql.org 
    Cc: 
    Subject: Re: [ADMIN] Replication
    
    

    Martha Stewart called it a Good Thing whenianh@hubspan.com ("Ian Huynh")wrote:
    > Seems that there are a number of simultaneous project in trying to implement replication
    > for postgres.
    > 
    > Is anyone out there using this feature currently? If so , which package & OS environment.
    > 
    > the latest news was eRServer v1.0.   Is this the same as the v1.2 that Postgres INC announced
    > early this year ?
    > 
    > Is there a master-master capability or only master-multi-slave ?
    
    The version of eRServer recently released as free software is not the
    latest "production" release, but it is a direct predecessor.
    
    Its operational model is that of "master/possible multiple slaves,"
    NOT multimaster.
    
    Multimaster involves some thorny issues surrounding synchronization,
    as it is possible for one master to receive a set of changes
    concurrently with another master receiving _incompatible_ changes.
    
    PalmOS users see this if they change a particular record on both
    desktop and palmtop; the next time they sync, the conflicting record
    normally gets duplicated and the user gets to decide how to fix it.
    
    Doing that in a big production database context may not be
    possible/acceptable.  You only should do multimaster if you REALLY
    know what you're doing.
    --
    select 'aa454' || '@' || 'freenet.carleton.ca';
    http://www.ntlug.org/~cbbrowne/sap.html
    Do you know where your towel is?
    
    ---------------------------(end of broadcast)---------------------------
    TIP 4: Don't 'kill -9' the postmaster
    

On Sun, Aug 31, 2003 at 08:33:33PM -0700, Ian Huynh wrote:
> There are certain cases where multi-master rep are very
> implementable such as where the datasets being rep are fairly
> disjointed but need to be replicated for completeness in terms of
> reporting. But you are absolutely 110% correct. Multimaster synchro
> is a nightmare.

If you had two completely separate databases, you could merge them
using erserver.

If you had some nifty tricks with unique indexes, &c., you could also
use erserver to merge them.  Or at least, I think you could.  Imagine
making all your unique indexes on two columns: say, the row id (just
a type serial) plus a server id (which was static for everything in
that database),

There are plenty of practical problems with such a scheme, and it
seems to me a little fragile, but it'd work.

A

--
----
Andrew Sullivan                         204-4141 Yonge Street
Liberty RMS                           Toronto, Ontario Canada
<andrew@libertyrms.info>                              M2P 2A8
                                         +1 416 646 3304 x110