No idea if this is an option for you or not, but if you make tables
unlogged they won't generate WAL and therefore won't be replicated:
http://rhaas.blogspot.ca/2010/05/global-temporary-and-unlogged-tables.html
Of course unlogged tables have several drawbacks to carefully consider.
On 03/26/2015 08:32 AM, Sven Geggus wrote:
Hello,
I have a question regarding log-shipping replication, but let me first
explain what I currently do.
Usually Openstreetmap Tile Servers are currently set up using a
PostgreSQL/Postgis Database which stores the data need for rendering raster
map-tiles.
After the initial import of a complete OSM dataset it is possible to
replicate changes to this database using cyclic calls to openstreetmaps own
tool called osm2pgsql.
Unfortunately there is one major drawback with this approach:
We need to somehow save the state of the osm database to be able to apply
future changes. This is currently done using a few tables in the target
database. However these tables are not needed for map rendering and are
consuming by far the most disk space (still somewhat expensive on SSD)!
So here is my question:
Would it be possible to have a setup, where one master data database will
act in the above matter (still running osm2pgsl) but will also provide
publicly availabe data for log-shipping standby servers?
We would need to be able to explicitely specify the tables to be replicated
or the other way round explicitely exclude a couple of them. If this helps
it would be possible to separate them by tablespaces.
Regards
Sven
--
Mike
--
Sent via pgsql-general mailing list (pgsql-general@xxxxxxxxxxxxxx)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general
