Re: postgresql replication

Vlad wrote:

>Hello,
>
>in need to increase reliability of the service (and perhaps eventually
>offload main DB server) we are looking to setup replication for the
>database server. I found two solutions:
>
>Slony ( http://gborg.postgresql.org/project/slony1/projdisplay.php )
>PGCluster ( http://pgfoundry.org/projects/pgcluster )
>
>I found that PgCluster supports multi-muster mode, which we can
>benefit from, but it's not required for "backup" which is #1 goal at
>the moment.
>
>If anyone used those solutions, compared performance, reliability, etc
>- please share your experience / thoughts.
>
>
There is a 3rd option - DBMirror, which comes as part of the Postgres
distribution (look in the contrib/dbmirror directory).

I looked at Slony, which seems to be a current favourite -but I couldn't
get it working on my database (claimed my tables didn't have relevant
keys - which they do). Slony-I had almost non-existent documentation
which I always find prety unacceptable unless everything goes very
smoothly..

I then found DBMirror - which I'm now using to replicate replicate from
master to slave database, and then cascade replicating my slave to a 3rd
off-site server.

Performance of DBMirror.pl (a Perl script that does the actual
replication) was very poor - so I re-wrote it in C++. Some details at:
http://www.whitebeam.org/library/guide/TechNotes/replicate.rhtm

DBmirror doesn't replicate schema changes, which wasn't a problem for me
since our schema changes very infrequently.

Hope that helps
Pete
--
http://www.yellowhawk.co.uk
http://www.whitebeam.org
-----